W0127 11:26:50.717000 3370624 site-packages/torch/distributed/run.py:803] 
W0127 11:26:50.717000 3370624 site-packages/torch/distributed/run.py:803] *****************************************
W0127 11:26:50.717000 3370624 site-packages/torch/distributed/run.py:803] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed. 
W0127 11:26:50.717000 3370624 site-packages/torch/distributed/run.py:803] *****************************************
2026-01-27 11:26:55,908 - WARNING - chunkwise_backward is incompatible with FSDP; disabling it.
2026-01-27 11:26:55,909 - INFO - ============================================================
2026-01-27 11:26:55,909 - INFO - Qwen3-4B + Titans training (DDP/FSDP)
2026-01-27 11:26:55,909 - INFO - ============================================================
2026-01-27 11:26:55,909 - INFO - distributed=True, world_size=4, use_fsdp=True
2026-01-27 11:26:55,909 - INFO - mode=TRAIN
2026-01-27 11:26:55,909 - INFO - model_path=/data/huangyifei/huggingface_cache/hub/models--Qwen--Qwen3-4B-Instruct-2507/snapshots/cdbee75f17c01a7cc42f958dc650907174af0554
2026-01-27 11:26:55,909 - INFO - data_path=/data/yty/BABILong/babilong-train-5k-samples/data/qa1/32k.json
2026-01-27 11:26:55,909 - INFO - output_dir=./outputs/qwen_babilong_no_memory
2026-01-27 11:26:55,909 - INFO - max_samples=2500
2026-01-27 11:26:55,909 - INFO - max_length=32768
2026-01-27 11:26:55,909 - INFO - chunk_size=4096
2026-01-27 11:26:55,909 - INFO - use_memory=False
2026-01-27 11:26:55,909 - INFO - chunkwise_backward=False
2026-01-27 11:26:55,909 - INFO - label_prefix_tokens=0
2026-01-27 11:26:55,909 - INFO - detach_mem_state=True

Loading checkpoint shards:   0%|          | 0/3 [00:00<?, ?it/s]
Loading checkpoint shards:   0%|          | 0/3 [00:00<?, ?it/s]
Loading checkpoint shards:   0%|          | 0/3 [00:00<?, ?it/s]
Loading checkpoint shards:   0%|          | 0/3 [00:00<?, ?it/s]
Loading checkpoint shards:  33%|███▎      | 1/3 [00:00<00:01,  1.16it/s]
Loading checkpoint shards:  33%|███▎      | 1/3 [00:00<00:01,  1.09it/s]
Loading checkpoint shards:  33%|███▎      | 1/3 [00:00<00:01,  1.20it/s]
Loading checkpoint shards:  33%|███▎      | 1/3 [00:00<00:01,  1.27it/s]
Loading checkpoint shards:  67%|██████▋   | 2/3 [00:01<00:00,  1.07it/s]
Loading checkpoint shards:  67%|██████▋   | 2/3 [00:01<00:00,  1.04it/s]
Loading checkpoint shards:  67%|██████▋   | 2/3 [00:01<00:00,  1.08it/s]
Loading checkpoint shards:  67%|██████▋   | 2/3 [00:01<00:00,  1.10it/s]
Loading checkpoint shards: 100%|██████████| 3/3 [00:01<00:00,  1.60it/s]
Loading checkpoint shards: 100%|██████████| 3/3 [00:01<00:00,  1.55it/s]


Loading checkpoint shards: 100%|██████████| 3/3 [00:01<00:00,  1.62it/s]
Loading checkpoint shards: 100%|██████████| 3/3 [00:01<00:00,  1.66it/s]

2026-01-27 11:27:28,081 - INFO - Loading dataset: /data/yty/BABILong/babilong-train-5k-samples/data/qa1/32k.json
2026-01-27 11:27:28,095 - INFO - Loading dataset: /data/yty/BABILong/babilong-train-5k-samples/data/qa1/32k.json
2026-01-27 11:27:28,096 - INFO - Loading dataset: /data/yty/BABILong/babilong-train-5k-samples/data/qa1/32k.json
2026-01-27 11:27:28,097 - INFO - Loading dataset: /data/yty/BABILong/babilong-train-5k-samples/data/qa1/32k.json
2026-01-27 11:27:32,272 - INFO - Dataset size: 2500
2026-01-27 11:27:32,273 - INFO - [QwenTitansForBABILong] Initialized (memory disabled)
2026-01-27 11:27:32,273 - INFO -   - hidden_size: 2560
2026-01-27 11:27:32,273 - INFO -   - chunk_size: 4096
2026-01-27 11:27:32,308 - INFO - Dataset size: 2500
2026-01-27 11:27:32,309 - INFO - [QwenTitansForBABILong] Initialized (memory disabled)
2026-01-27 11:27:32,309 - INFO -   - hidden_size: 2560
2026-01-27 11:27:32,309 - INFO -   - chunk_size: 4096
2026-01-27 11:27:32,347 - INFO - Param groups: memory=0, pretrained=398
/root/githubs/titans-pytorch/examples/train_qwen_titans_babilong.py:768: FutureWarning: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead.
  self.scaler = torch.cuda.amp.GradScaler(enabled=config.fp16)
2026-01-27 11:27:32,349 - INFO - Start training
2026-01-27 11:27:32,349 - INFO - Epoch 1/3

Epoch 1/3:   0%|          | 0/563 [00:00<?, ?it/s]2026-01-27 11:27:32,382 - INFO - Param groups: memory=0, pretrained=398
/root/githubs/titans-pytorch/examples/train_qwen_titans_babilong.py:768: FutureWarning: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead.
  self.scaler = torch.cuda.amp.GradScaler(enabled=config.fp16)
2026-01-27 11:27:32,442 - INFO - Dataset size: 2500
2026-01-27 11:27:32,444 - INFO - [QwenTitansForBABILong] Initialized (memory disabled)
2026-01-27 11:27:32,444 - INFO -   - hidden_size: 2560
2026-01-27 11:27:32,444 - INFO -   - chunk_size: 4096
2026-01-27 11:27:32,462 - INFO - Dataset size: 2500
2026-01-27 11:27:32,466 - INFO - [QwenTitansForBABILong] Initialized (memory disabled)
2026-01-27 11:27:32,466 - INFO -   - hidden_size: 2560
2026-01-27 11:27:32,466 - INFO -   - chunk_size: 4096
2026-01-27 11:27:32,519 - INFO - Param groups: memory=0, pretrained=398
/root/githubs/titans-pytorch/examples/train_qwen_titans_babilong.py:768: FutureWarning: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead.
  self.scaler = torch.cuda.amp.GradScaler(enabled=config.fp16)
2026-01-27 11:27:32,544 - INFO - Param groups: memory=0, pretrained=398
/root/githubs/titans-pytorch/examples/train_qwen_titans_babilong.py:768: FutureWarning: `torch.cuda.amp.GradScaler(args...)` is deprecated. Please use `torch.amp.GradScaler('cuda', args...)` instead.
  self.scaler = torch.cuda.amp.GradScaler(enabled=config.fp16)

Epoch 1/3:   0%|          | 1/563 [00:04<42:40,  4.56s/it]
Epoch 1/3:   0%|          | 2/563 [00:08<36:50,  3.94s/it]
Epoch 1/3:   1%|          | 3/563 [00:11<35:19,  3.78s/it]
Epoch 1/3:   1%|          | 4/563 [00:15<34:16,  3.68s/it]
Epoch 1/3:   1%|          | 5/563 [00:18<33:39,  3.62s/it]
Epoch 1/3:   1%|          | 6/563 [00:22<33:35,  3.62s/it]
Epoch 1/3:   1%|          | 7/563 [00:25<33:14,  3.59s/it]
Epoch 1/3:   1%|▏         | 8/563 [00:29<33:15,  3.60s/it]
Epoch 1/3:   2%|▏         | 9/563 [00:32<33:00,  3.57s/it]
Epoch 1/3:   2%|▏         | 10/563 [00:36<32:49,  3.56s/it]
Epoch 1/3:   2%|▏         | 11/563 [00:40<32:40,  3.55s/it]
Epoch 1/3:   2%|▏         | 12/563 [00:43<32:33,  3.55s/it]
Epoch 1/3:   2%|▏         | 13/563 [00:47<32:27,  3.54s/it]
Epoch 1/3:   2%|▏         | 14/563 [00:50<32:22,  3.54s/it]
Epoch 1/3:   3%|▎         | 15/563 [00:54<32:17,  3.54s/it]
Epoch 1/3:   3%|▎         | 16/563 [00:57<32:14,  3.54s/it]
Epoch 1/3:   3%|▎         | 17/563 [01:01<32:10,  3.53s/it]
Epoch 1/3:   3%|▎         | 18/563 [01:04<32:06,  3.53s/it]
Epoch 1/3:   3%|▎         | 19/563 [01:08<32:02,  3.53s/it]
Epoch 1/3:   4%|▎         | 20/563 [01:11<31:59,  3.54s/it]
Epoch 1/3:   4%|▎         | 21/563 [01:15<31:55,  3.53s/it]
Epoch 1/3:   4%|▍         | 22/563 [01:18<31:51,  3.53s/it]
Epoch 1/3:   4%|▍         | 23/563 [01:22<32:02,  3.56s/it]
Epoch 1/3:   4%|▍         | 24/563 [01:26<32:08,  3.58s/it]
Epoch 1/3:   4%|▍         | 25/563 [01:29<32:30,  3.63s/it]
Epoch 1/3:   5%|▍         | 26/563 [01:33<32:27,  3.63s/it]
Epoch 1/3:   5%|▍         | 27/563 [01:37<32:40,  3.66s/it]
Epoch 1/3:   5%|▍         | 28/563 [01:40<32:46,  3.68s/it]
Epoch 1/3:   5%|▌         | 29/563 [01:44<32:20,  3.63s/it]
Epoch 1/3:   5%|▌         | 30/563 [01:48<32:01,  3.60s/it]
Epoch 1/3:   6%|▌         | 31/563 [01:51<31:46,  3.58s/it]
Epoch 1/3:   6%|▌         | 31/563 [01:55<31:46,  3.58s/it, gstep=1, loss=18.7221]
Epoch 1/3:   6%|▌         | 32/563 [01:55<31:53,  3.60s/it, gstep=1, loss=18.7221]
Epoch 1/3:   6%|▌         | 33/563 [01:58<31:40,  3.59s/it, gstep=1, loss=18.7221]
Epoch 1/3:   6%|▌         | 34/563 [02:02<31:27,  3.57s/it, gstep=1, loss=18.7221]
Epoch 1/3:   6%|▌         | 35/563 [02:05<31:18,  3.56s/it, gstep=1, loss=18.7221]
Epoch 1/3:   6%|▋         | 36/563 [02:09<31:11,  3.55s/it, gstep=1, loss=18.7221]
Epoch 1/3:   7%|▋         | 37/563 [02:12<31:04,  3.55s/it, gstep=1, loss=18.7221]
Epoch 1/3:   7%|▋         | 38/563 [02:16<31:13,  3.57s/it, gstep=1, loss=18.7221]
Epoch 1/3:   7%|▋         | 39/563 [02:20<31:04,  3.56s/it, gstep=1, loss=18.7221]
Epoch 1/3:   7%|▋         | 40/563 [02:23<30:56,  3.55s/it, gstep=1, loss=18.7221]
Epoch 1/3:   7%|▋         | 41/563 [02:27<30:51,  3.55s/it, gstep=1, loss=18.7221]
Epoch 1/3:   7%|▋         | 42/563 [02:30<30:44,  3.54s/it, gstep=1, loss=18.7221]
Epoch 1/3:   8%|▊         | 43/563 [02:34<30:52,  3.56s/it, gstep=1, loss=18.7221]
Epoch 1/3:   8%|▊         | 44/563 [02:37<30:58,  3.58s/it, gstep=1, loss=18.7221]
Epoch 1/3:   8%|▊         | 45/563 [02:41<30:46,  3.57s/it, gstep=1, loss=18.7221]
Epoch 1/3:   8%|▊         | 46/563 [02:44<30:38,  3.56s/it, gstep=1, loss=18.7221]
Epoch 1/3:   8%|▊         | 47/563 [02:48<30:31,  3.55s/it, gstep=1, loss=18.7221]
Epoch 1/3:   9%|▊         | 48/563 [02:52<30:25,  3.54s/it, gstep=1, loss=18.7221]
Epoch 1/3:   9%|▊         | 49/563 [02:55<30:33,  3.57s/it, gstep=1, loss=18.7221]
Epoch 1/3:   9%|▉         | 50/563 [02:59<30:24,  3.56s/it, gstep=1, loss=18.7221]
Epoch 1/3:   9%|▉         | 51/563 [03:02<30:17,  3.55s/it, gstep=1, loss=18.7221]
Epoch 1/3:   9%|▉         | 52/563 [03:06<30:11,  3.54s/it, gstep=1, loss=18.7221]
Epoch 1/3:   9%|▉         | 53/563 [03:09<30:05,  3.54s/it, gstep=1, loss=18.7221]
Epoch 1/3:  10%|▉         | 54/563 [03:13<30:01,  3.54s/it, gstep=1, loss=18.7221]
Epoch 1/3:  10%|▉         | 55/563 [03:16<29:56,  3.54s/it, gstep=1, loss=18.7221]
Epoch 1/3:  10%|▉         | 56/563 [03:20<29:51,  3.53s/it, gstep=1, loss=18.7221]
Epoch 1/3:  10%|█         | 57/563 [03:23<29:48,  3.53s/it, gstep=1, loss=18.7221]
Epoch 1/3:  10%|█         | 58/563 [03:27<29:44,  3.53s/it, gstep=1, loss=18.7221]
Epoch 1/3:  10%|█         | 59/563 [03:31<29:54,  3.56s/it, gstep=1, loss=18.7221]
Epoch 1/3:  11%|█         | 60/563 [03:34<29:47,  3.55s/it, gstep=1, loss=18.7221]
Epoch 1/3:  11%|█         | 61/563 [03:38<29:39,  3.55s/it, gstep=1, loss=18.7221]
Epoch 1/3:  11%|█         | 62/563 [03:41<29:34,  3.54s/it, gstep=1, loss=18.7221]
Epoch 1/3:  11%|█         | 63/563 [03:45<29:28,  3.54s/it, gstep=1, loss=18.7221]
Epoch 1/3:  11%|█         | 63/563 [03:48<29:28,  3.54s/it, gstep=2, loss=16.7293]
Epoch 1/3:  11%|█▏        | 64/563 [03:48<29:32,  3.55s/it, gstep=2, loss=16.7293]
Epoch 1/3:  12%|█▏        | 65/563 [03:52<29:35,  3.57s/it, gstep=2, loss=16.7293]
Epoch 1/3:  12%|█▏        | 66/563 [03:55<29:41,  3.59s/it, gstep=2, loss=16.7293]
Epoch 1/3:  12%|█▏        | 67/563 [03:59<29:30,  3.57s/it, gstep=2, loss=16.7293]
Epoch 1/3:  12%|█▏        | 68/563 [04:03<29:35,  3.59s/it, gstep=2, loss=16.7293]
Epoch 1/3:  12%|█▏        | 69/563 [04:06<29:36,  3.60s/it, gstep=2, loss=16.7293]
Epoch 1/3:  12%|█▏        | 70/563 [04:10<29:23,  3.58s/it, gstep=2, loss=16.7293]
Epoch 1/3:  13%|█▎        | 71/563 [04:13<29:26,  3.59s/it, gstep=2, loss=16.7293]
Epoch 1/3:  13%|█▎        | 72/563 [04:17<29:14,  3.57s/it, gstep=2, loss=16.7293]
Epoch 1/3:  13%|█▎        | 73/563 [04:20<29:04,  3.56s/it, gstep=2, loss=16.7293]
Epoch 1/3:  13%|█▎        | 74/563 [04:24<29:09,  3.58s/it, gstep=2, loss=16.7293]
Epoch 1/3:  13%|█▎        | 75/563 [04:28<29:13,  3.59s/it, gstep=2, loss=16.7293]
Epoch 1/3:  13%|█▎        | 76/563 [04:31<29:00,  3.57s/it, gstep=2, loss=16.7293]
Epoch 1/3:  14%|█▎        | 77/563 [04:35<29:04,  3.59s/it, gstep=2, loss=16.7293]
Epoch 1/3:  14%|█▍        | 78/563 [04:38<28:52,  3.57s/it, gstep=2, loss=16.7293]
Epoch 1/3:  14%|█▍        | 79/563 [04:42<28:43,  3.56s/it, gstep=2, loss=16.7293]
Epoch 1/3:  14%|█▍        | 80/563 [04:45<28:35,  3.55s/it, gstep=2, loss=16.7293]
Epoch 1/3:  14%|█▍        | 81/563 [04:49<28:29,  3.55s/it, gstep=2, loss=16.7293]
Epoch 1/3:  15%|█▍        | 82/563 [04:53<28:23,  3.54s/it, gstep=2, loss=16.7293]
Epoch 1/3:  15%|█▍        | 83/563 [04:56<28:44,  3.59s/it, gstep=2, loss=16.7293]
Epoch 1/3:  15%|█▍        | 84/563 [05:00<28:32,  3.58s/it, gstep=2, loss=16.7293]
Epoch 1/3:  15%|█▌        | 85/563 [05:03<28:22,  3.56s/it, gstep=2, loss=16.7293]
Epoch 1/3:  15%|█▌        | 86/563 [05:07<28:13,  3.55s/it, gstep=2, loss=16.7293]
Epoch 1/3:  15%|█▌        | 87/563 [05:10<28:21,  3.57s/it, gstep=2, loss=16.7293]
Epoch 1/3:  16%|█▌        | 88/563 [05:14<28:11,  3.56s/it, gstep=2, loss=16.7293]
Epoch 1/3:  16%|█▌        | 89/563 [05:18<28:03,  3.55s/it, gstep=2, loss=16.7293]
Epoch 1/3:  16%|█▌        | 90/563 [05:21<28:09,  3.57s/it, gstep=2, loss=16.7293]
Epoch 1/3:  16%|█▌        | 91/563 [05:25<28:12,  3.59s/it, gstep=2, loss=16.7293]
Epoch 1/3:  16%|█▋        | 92/563 [05:28<28:01,  3.57s/it, gstep=2, loss=16.7293]
Epoch 1/3:  17%|█▋        | 93/563 [05:32<27:52,  3.56s/it, gstep=2, loss=16.7293]
Epoch 1/3:  17%|█▋        | 94/563 [05:35<27:44,  3.55s/it, gstep=2, loss=16.7293]
Epoch 1/3:  17%|█▋        | 95/563 [05:39<27:38,  3.54s/it, gstep=2, loss=16.7293]
Epoch 1/3:  17%|█▋        | 95/563 [05:42<27:38,  3.54s/it, gstep=3, loss=14.6711]2026-01-27 11:33:15,331 - INFO - Step 3 | loss=14.6711 | lr=4.96e-06

Epoch 1/3:  17%|█▋        | 96/563 [05:42<27:41,  3.56s/it, gstep=3, loss=14.6711]
Epoch 1/3:  17%|█▋        | 97/563 [05:46<27:32,  3.55s/it, gstep=3, loss=14.6711]
Epoch 1/3:  17%|█▋        | 98/563 [05:50<27:39,  3.57s/it, gstep=3, loss=14.6711]
Epoch 1/3:  18%|█▊        | 99/563 [05:53<27:30,  3.56s/it, gstep=3, loss=14.6711]
Epoch 1/3:  18%|█▊        | 100/563 [05:57<27:22,  3.55s/it, gstep=3, loss=14.6711]
Epoch 1/3:  18%|█▊        | 101/563 [06:00<27:42,  3.60s/it, gstep=3, loss=14.6711]
Epoch 1/3:  18%|█▊        | 102/563 [06:04<27:29,  3.58s/it, gstep=3, loss=14.6711]
Epoch 1/3:  18%|█▊        | 103/563 [06:08<27:31,  3.59s/it, gstep=3, loss=14.6711]
Epoch 1/3:  18%|█▊        | 104/563 [06:11<27:44,  3.63s/it, gstep=3, loss=14.6711]
Epoch 1/3:  19%|█▊        | 105/563 [06:15<27:27,  3.60s/it, gstep=3, loss=14.6711]
Epoch 1/3:  19%|█▉        | 106/563 [06:18<27:14,  3.58s/it, gstep=3, loss=14.6711]
Epoch 1/3:  19%|█▉        | 107/563 [06:22<27:04,  3.56s/it, gstep=3, loss=14.6711]
Epoch 1/3:  19%|█▉        | 108/563 [06:25<26:56,  3.55s/it, gstep=3, loss=14.6711]
Epoch 1/3:  19%|█▉        | 109/563 [06:29<26:49,  3.55s/it, gstep=3, loss=14.6711]
Epoch 1/3:  20%|█▉        | 110/563 [06:32<26:44,  3.54s/it, gstep=3, loss=14.6711]
Epoch 1/3:  20%|█▉        | 111/563 [06:36<26:39,  3.54s/it, gstep=3, loss=14.6711]
Epoch 1/3:  20%|█▉        | 112/563 [06:39<26:35,  3.54s/it, gstep=3, loss=14.6711]
Epoch 1/3:  20%|██        | 113/563 [06:43<26:42,  3.56s/it, gstep=3, loss=14.6711]
Epoch 1/3:  20%|██        | 114/563 [06:47<26:34,  3.55s/it, gstep=3, loss=14.6711]
Epoch 1/3:  20%|██        | 115/563 [06:50<26:52,  3.60s/it, gstep=3, loss=14.6711]
Epoch 1/3:  21%|██        | 116/563 [06:54<27:03,  3.63s/it, gstep=3, loss=14.6711]
Epoch 1/3:  21%|██        | 117/563 [06:58<26:45,  3.60s/it, gstep=3, loss=14.6711]
Epoch 1/3:  21%|██        | 118/563 [07:01<26:32,  3.58s/it, gstep=3, loss=14.6711]
Epoch 1/3:  21%|██        | 119/563 [07:05<26:46,  3.62s/it, gstep=3, loss=14.6711]
Epoch 1/3:  21%|██▏       | 120/563 [07:08<26:32,  3.59s/it, gstep=3, loss=14.6711]
Epoch 1/3:  21%|██▏       | 121/563 [07:12<26:44,  3.63s/it, gstep=3, loss=14.6711]
Epoch 1/3:  22%|██▏       | 122/563 [07:16<26:28,  3.60s/it, gstep=3, loss=14.6711]
Epoch 1/3:  22%|██▏       | 123/563 [07:19<26:15,  3.58s/it, gstep=3, loss=14.6711]
Epoch 1/3:  22%|██▏       | 124/563 [07:23<26:18,  3.60s/it, gstep=3, loss=14.6711]
Epoch 1/3:  22%|██▏       | 125/563 [07:26<26:19,  3.61s/it, gstep=3, loss=14.6711]
Epoch 1/3:  22%|██▏       | 126/563 [07:30<26:06,  3.58s/it, gstep=3, loss=14.6711]
Epoch 1/3:  23%|██▎       | 127/563 [07:34<26:06,  3.59s/it, gstep=3, loss=14.6711]
Epoch 1/3:  23%|██▎       | 127/563 [07:37<26:06,  3.59s/it, gstep=4, loss=13.1929]
Epoch 1/3:  23%|██▎       | 128/563 [07:37<26:02,  3.59s/it, gstep=4, loss=13.1929]
Epoch 1/3:  23%|██▎       | 129/563 [07:41<25:49,  3.57s/it, gstep=4, loss=13.1929]
Epoch 1/3:  23%|██▎       | 130/563 [07:44<25:42,  3.56s/it, gstep=4, loss=13.1929]
Epoch 1/3:  23%|██▎       | 131/563 [07:48<25:35,  3.55s/it, gstep=4, loss=13.1929]
Epoch 1/3:  23%|██▎       | 132/563 [07:51<25:29,  3.55s/it, gstep=4, loss=13.1929]
Epoch 1/3:  24%|██▎       | 133/563 [07:55<25:35,  3.57s/it, gstep=4, loss=13.1929]
Epoch 1/3:  24%|██▍       | 134/563 [07:58<25:27,  3.56s/it, gstep=4, loss=13.1929]
Epoch 1/3:  24%|██▍       | 135/563 [08:02<25:32,  3.58s/it, gstep=4, loss=13.1929]
Epoch 1/3:  24%|██▍       | 136/563 [08:06<25:23,  3.57s/it, gstep=4, loss=13.1929]
Epoch 1/3:  24%|██▍       | 137/563 [08:09<25:15,  3.56s/it, gstep=4, loss=13.1929]
Epoch 1/3:  25%|██▍       | 138/563 [08:13<25:10,  3.55s/it, gstep=4, loss=13.1929]
Epoch 1/3:  25%|██▍       | 139/563 [08:16<25:05,  3.55s/it, gstep=4, loss=13.1929]
Epoch 1/3:  25%|██▍       | 140/563 [08:20<24:59,  3.54s/it, gstep=4, loss=13.1929]
Epoch 1/3:  25%|██▌       | 141/563 [08:23<24:54,  3.54s/it, gstep=4, loss=13.1929]
Epoch 1/3:  25%|██▌       | 142/563 [08:27<24:49,  3.54s/it, gstep=4, loss=13.1929]
Epoch 1/3:  25%|██▌       | 143/563 [08:30<24:45,  3.54s/it, gstep=4, loss=13.1929]
Epoch 1/3:  26%|██▌       | 144/563 [08:34<24:40,  3.53s/it, gstep=4, loss=13.1929]
Epoch 1/3:  26%|██▌       | 145/563 [08:37<24:37,  3.54s/it, gstep=4, loss=13.1929]
Epoch 1/3:  26%|██▌       | 146/563 [08:41<24:33,  3.53s/it, gstep=4, loss=13.1929]
Epoch 1/3:  26%|██▌       | 147/563 [08:45<24:41,  3.56s/it, gstep=4, loss=13.1929]
Epoch 1/3:  26%|██▋       | 148/563 [08:48<24:35,  3.56s/it, gstep=4, loss=13.1929]
Epoch 1/3:  26%|██▋       | 149/563 [08:52<24:40,  3.57s/it, gstep=4, loss=13.1929]
Epoch 1/3:  27%|██▋       | 150/563 [08:55<24:32,  3.56s/it, gstep=4, loss=13.1929]
Epoch 1/3:  27%|██▋       | 151/563 [08:59<24:25,  3.56s/it, gstep=4, loss=13.1929]
Epoch 1/3:  27%|██▋       | 152/563 [09:02<24:18,  3.55s/it, gstep=4, loss=13.1929]
Epoch 1/3:  27%|██▋       | 153/563 [09:06<24:25,  3.57s/it, gstep=4, loss=13.1929]
Epoch 1/3:  27%|██▋       | 154/563 [09:10<24:17,  3.56s/it, gstep=4, loss=13.1929]
Epoch 1/3:  28%|██▊       | 155/563 [09:13<24:10,  3.56s/it, gstep=4, loss=13.1929]
Epoch 1/3:  28%|██▊       | 156/563 [09:17<24:04,  3.55s/it, gstep=4, loss=13.1929]
Epoch 1/3:  28%|██▊       | 157/563 [09:20<24:11,  3.57s/it, gstep=4, loss=13.1929]
Epoch 1/3:  28%|██▊       | 158/563 [09:24<24:13,  3.59s/it, gstep=4, loss=13.1929]
Epoch 1/3:  28%|██▊       | 159/563 [09:27<24:03,  3.57s/it, gstep=4, loss=13.1929]
Epoch 1/3:  28%|██▊       | 159/563 [09:31<24:03,  3.57s/it, gstep=5, loss=11.7782]
Epoch 1/3:  28%|██▊       | 160/563 [09:31<24:13,  3.61s/it, gstep=5, loss=11.7782]
Epoch 1/3:  29%|██▊       | 161/563 [09:35<23:58,  3.58s/it, gstep=5, loss=11.7782]
Epoch 1/3:  29%|██▉       | 162/563 [09:38<24:01,  3.59s/it, gstep=5, loss=11.7782]
Epoch 1/3:  29%|██▉       | 163/563 [09:42<23:50,  3.58s/it, gstep=5, loss=11.7782]
Epoch 1/3:  29%|██▉       | 164/563 [09:45<23:52,  3.59s/it, gstep=5, loss=11.7782]
Epoch 1/3:  29%|██▉       | 165/563 [09:49<23:52,  3.60s/it, gstep=5, loss=11.7782]
Epoch 1/3:  29%|██▉       | 166/563 [09:53<23:51,  3.61s/it, gstep=5, loss=11.7782]
Epoch 1/3:  30%|██▉       | 167/563 [09:56<23:53,  3.62s/it, gstep=5, loss=11.7782]
Epoch 1/3:  30%|██▉       | 168/563 [10:00<23:50,  3.62s/it, gstep=5, loss=11.7782]
Epoch 1/3:  30%|███       | 169/563 [10:03<23:36,  3.60s/it, gstep=5, loss=11.7782]
Epoch 1/3:  30%|███       | 170/563 [10:07<23:26,  3.58s/it, gstep=5, loss=11.7782]
Epoch 1/3:  30%|███       | 171/563 [10:10<23:17,  3.57s/it, gstep=5, loss=11.7782]
Epoch 1/3:  31%|███       | 172/563 [10:14<23:20,  3.58s/it, gstep=5, loss=11.7782]
Epoch 1/3:  31%|███       | 173/563 [10:18<23:22,  3.60s/it, gstep=5, loss=11.7782]
Epoch 1/3:  31%|███       | 174/563 [10:21<23:22,  3.60s/it, gstep=5, loss=11.7782]
Epoch 1/3:  31%|███       | 175/563 [10:25<23:09,  3.58s/it, gstep=5, loss=11.7782]
Epoch 1/3:  31%|███▏      | 176/563 [10:28<23:00,  3.57s/it, gstep=5, loss=11.7782]
Epoch 1/3:  31%|███▏      | 177/563 [10:32<22:53,  3.56s/it, gstep=5, loss=11.7782]
Epoch 1/3:  32%|███▏      | 178/563 [10:36<22:47,  3.55s/it, gstep=5, loss=11.7782]
Epoch 1/3:  32%|███▏      | 179/563 [10:39<22:41,  3.55s/it, gstep=5, loss=11.7782]
Epoch 1/3:  32%|███▏      | 180/563 [10:43<22:36,  3.54s/it, gstep=5, loss=11.7782]
Epoch 1/3:  32%|███▏      | 181/563 [10:46<22:43,  3.57s/it, gstep=5, loss=11.7782]
Epoch 1/3:  32%|███▏      | 182/563 [10:50<22:46,  3.59s/it, gstep=5, loss=11.7782]
Epoch 1/3:  33%|███▎      | 183/563 [10:53<22:46,  3.60s/it, gstep=5, loss=11.7782]
Epoch 1/3:  33%|███▎      | 184/563 [10:57<22:36,  3.58s/it, gstep=5, loss=11.7782]
Epoch 1/3:  33%|███▎      | 185/563 [11:01<22:37,  3.59s/it, gstep=5, loss=11.7782]
Epoch 1/3:  33%|███▎      | 186/563 [11:04<22:26,  3.57s/it, gstep=5, loss=11.7782]
Epoch 1/3:  33%|███▎      | 187/563 [11:08<22:19,  3.56s/it, gstep=5, loss=11.7782]
Epoch 1/3:  33%|███▎      | 188/563 [11:11<22:12,  3.55s/it, gstep=5, loss=11.7782]
Epoch 1/3:  34%|███▎      | 189/563 [11:15<22:07,  3.55s/it, gstep=5, loss=11.7782]
Epoch 1/3:  34%|███▎      | 190/563 [11:18<22:01,  3.54s/it, gstep=5, loss=11.7782]
Epoch 1/3:  34%|███▍      | 191/563 [11:22<21:58,  3.54s/it, gstep=5, loss=11.7782]
Epoch 1/3:  34%|███▍      | 191/563 [11:25<21:58,  3.54s/it, gstep=6, loss=10.8305]2026-01-27 11:38:58,261 - INFO - Step 6 | loss=10.8305 | lr=4.85e-06

Epoch 1/3:  34%|███▍      | 192/563 [11:25<21:59,  3.56s/it, gstep=6, loss=10.8305]
Epoch 1/3:  34%|███▍      | 193/563 [11:29<21:51,  3.54s/it, gstep=6, loss=10.8305]
Epoch 1/3:  34%|███▍      | 194/563 [11:32<21:47,  3.54s/it, gstep=6, loss=10.8305]
Epoch 1/3:  35%|███▍      | 195/563 [11:36<21:42,  3.54s/it, gstep=6, loss=10.8305]
Epoch 1/3:  35%|███▍      | 196/563 [11:40<21:38,  3.54s/it, gstep=6, loss=10.8305]
Epoch 1/3:  35%|███▍      | 197/563 [11:43<21:54,  3.59s/it, gstep=6, loss=10.8305]
Epoch 1/3:  35%|███▌      | 198/563 [11:47<21:44,  3.57s/it, gstep=6, loss=10.8305]
Epoch 1/3:  35%|███▌      | 199/563 [11:50<21:36,  3.56s/it, gstep=6, loss=10.8305]
Epoch 1/3:  36%|███▌      | 200/563 [11:54<21:29,  3.55s/it, gstep=6, loss=10.8305]
Epoch 1/3:  36%|███▌      | 201/563 [11:57<21:34,  3.58s/it, gstep=6, loss=10.8305]
Epoch 1/3:  36%|███▌      | 202/563 [12:01<21:45,  3.62s/it, gstep=6, loss=10.8305]
Epoch 1/3:  36%|███▌      | 203/563 [12:05<21:33,  3.59s/it, gstep=6, loss=10.8305]
Epoch 1/3:  36%|███▌      | 204/563 [12:08<21:23,  3.57s/it, gstep=6, loss=10.8305]
Epoch 1/3:  36%|███▋      | 205/563 [12:12<21:33,  3.61s/it, gstep=6, loss=10.8305]
Epoch 1/3:  37%|███▋      | 206/563 [12:16<21:40,  3.64s/it, gstep=6, loss=10.8305]
Epoch 1/3:  37%|███▋      | 207/563 [12:19<21:34,  3.63s/it, gstep=6, loss=10.8305]
Epoch 1/3:  37%|███▋      | 208/563 [12:23<21:19,  3.60s/it, gstep=6, loss=10.8305]
Epoch 1/3:  37%|███▋      | 209/563 [12:26<21:08,  3.58s/it, gstep=6, loss=10.8305]
Epoch 1/3:  37%|███▋      | 210/563 [12:30<21:18,  3.62s/it, gstep=6, loss=10.8305]
Epoch 1/3:  37%|███▋      | 211/563 [12:34<21:24,  3.65s/it, gstep=6, loss=10.8305]
Epoch 1/3:  38%|███▊      | 212/563 [12:37<21:09,  3.62s/it, gstep=6, loss=10.8305]
Epoch 1/3:  38%|███▊      | 213/563 [12:41<20:57,  3.59s/it, gstep=6, loss=10.8305]
Epoch 1/3:  38%|███▊      | 214/563 [12:44<20:48,  3.58s/it, gstep=6, loss=10.8305]
Epoch 1/3:  38%|███▊      | 215/563 [12:48<20:39,  3.56s/it, gstep=6, loss=10.8305]
Epoch 1/3:  38%|███▊      | 216/563 [12:51<20:33,  3.55s/it, gstep=6, loss=10.8305]
Epoch 1/3:  39%|███▊      | 217/563 [12:55<20:27,  3.55s/it, gstep=6, loss=10.8305]
Epoch 1/3:  39%|███▊      | 218/563 [12:59<20:23,  3.55s/it, gstep=6, loss=10.8305]
Epoch 1/3:  39%|███▉      | 219/563 [13:02<20:28,  3.57s/it, gstep=6, loss=10.8305]
Epoch 1/3:  39%|███▉      | 220/563 [13:06<20:20,  3.56s/it, gstep=6, loss=10.8305]
Epoch 1/3:  39%|███▉      | 221/563 [13:09<20:14,  3.55s/it, gstep=6, loss=10.8305]
Epoch 1/3:  39%|███▉      | 222/563 [13:13<20:08,  3.54s/it, gstep=6, loss=10.8305]
Epoch 1/3:  40%|███▉      | 223/563 [13:16<20:04,  3.54s/it, gstep=6, loss=10.8305]
Epoch 1/3:  40%|███▉      | 223/563 [13:20<20:04,  3.54s/it, gstep=7, loss=9.9936] 
Epoch 1/3:  40%|███▉      | 224/563 [13:20<20:14,  3.58s/it, gstep=7, loss=9.9936]
Epoch 1/3:  40%|███▉      | 225/563 [13:23<20:04,  3.56s/it, gstep=7, loss=9.9936]
Epoch 1/3:  40%|████      | 226/563 [13:27<19:57,  3.55s/it, gstep=7, loss=9.9936]
Epoch 1/3:  40%|████      | 227/563 [13:31<20:10,  3.60s/it, gstep=7, loss=9.9936]
Epoch 1/3:  40%|████      | 228/563 [13:34<19:59,  3.58s/it, gstep=7, loss=9.9936]
Epoch 1/3:  41%|████      | 229/563 [13:38<19:51,  3.57s/it, gstep=7, loss=9.9936]
Epoch 1/3:  41%|████      | 230/563 [13:41<19:43,  3.56s/it, gstep=7, loss=9.9936]
Epoch 1/3:  41%|████      | 231/563 [13:45<19:37,  3.55s/it, gstep=7, loss=9.9936]
Epoch 1/3:  41%|████      | 232/563 [13:48<19:32,  3.54s/it, gstep=7, loss=9.9936]
Epoch 1/3:  41%|████▏     | 233/563 [13:52<19:27,  3.54s/it, gstep=7, loss=9.9936]
Epoch 1/3:  42%|████▏     | 234/563 [13:55<19:23,  3.54s/it, gstep=7, loss=9.9936]
Epoch 1/3:  42%|████▏     | 235/563 [13:59<19:19,  3.53s/it, gstep=7, loss=9.9936]
Epoch 1/3:  42%|████▏     | 236/563 [14:03<19:15,  3.53s/it, gstep=7, loss=9.9936]
Epoch 1/3:  42%|████▏     | 237/563 [14:06<19:11,  3.53s/it, gstep=7, loss=9.9936]
Epoch 1/3:  42%|████▏     | 238/563 [14:10<19:08,  3.53s/it, gstep=7, loss=9.9936]
Epoch 1/3:  42%|████▏     | 239/563 [14:13<19:04,  3.53s/it, gstep=7, loss=9.9936]
Epoch 1/3:  43%|████▎     | 240/563 [14:17<19:00,  3.53s/it, gstep=7, loss=9.9936]W0128 06:35:14.418000 3370624 site-packages/torch/distributed/elastic/agent/server/api.py:725] Received 15 death signal, shutting down workers
Traceback (most recent call last):
  File "/root/miniforge/lib/python3.12/subprocess.py", line 1990, in _internal_poll
    pid, sts = _waitpid(self.pid, _WNOHANG)
               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/root/miniforge/lib/python3.12/site-packages/torch/distributed/elastic/multiprocessing/api.py", line 85, in _terminate_process_handler
    sigval = signal.Signals(signum)
    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
torch.distributed.elastic.multiprocessing.api.SignalException: Process 3370624 got signal: 1

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/root/miniforge/bin/torchrun", line 7, in <module>
    sys.exit(main())
             ^^^^^^
  File "/root/miniforge/lib/python3.12/site-packages/torch/distributed/elastic/multiprocessing/errors/__init__.py", line 357, in wrapper
    @wraps(f)
              
  File "/root/miniforge/lib/python3.12/site-packages/torch/distributed/run.py", line 936, in main
    if args.module:
^^^^^
  File "/root/miniforge/lib/python3.12/site-packages/torch/distributed/run.py", line 927, in run
    cmd_args.append(args.training_script)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/root/miniforge/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 156, in __call__
    # entrypoint is a command and ``script.py`` is the python module.
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/root/miniforge/lib/python3.12/site-packages/torch/distributed/launcher/api.py", line 284, in launch_agent
    rdzv_handler=rdzv_registry.get_rendezvous_handler(rdzv_parameters),
             ^^^^^^^^^^^
  File "/root/miniforge/lib/python3.12/site-packages/torch/distributed/elastic/metrics/api.py", line 138, in wrapper
    raise
          
  File "/root/miniforge/lib/python3.12/site-packages/torch/distributed/elastic/agent/server/api.py", line 717, in run
    """Restart (stops, rendezvous, starts) all local workers in the group."""
                 ^^^^^^^^^^^^^^^^^^^^^^
  File "/root/miniforge/lib/python3.12/site-packages/torch/distributed/elastic/agent/server/api.py", line 882, in _invoke_run
    flakiness = 100.0
                 ^^^^^
  File "/root/miniforge/lib/python3.12/site-packages/torch/distributed/elastic/metrics/api.py", line 138, in wrapper
    raise
          
  File "/root/miniforge/lib/python3.12/site-packages/torch/distributed/elastic/agent/server/local_elastic_agent.py", line 389, in _monitor_workers
    
  File "/root/miniforge/lib/python3.12/site-packages/torch/distributed/elastic/multiprocessing/api.py", line 537, in wait
    "TORCHELASTIC_SIGNALS_TO_HANDLE", "SIGTERM,SIGINT,SIGHUP,SIGQUIT"
       ^^^^^^^^^^^^
  File "/root/miniforge/lib/python3.12/site-packages/torch/distributed/elastic/multiprocessing/api.py", line 858, in _poll
    def pids(self) -> dict[int, int]:
                       ^^^^^^^^^^^^^^^
  File "/root/miniforge/lib/python3.12/subprocess.py", line 1236, in poll
    return self._internal_poll()
           ^^^^^^^^^^^^^^^^^^^^^
  File "/root/miniforge/lib/python3.12/subprocess.py", line 2004, in _internal_poll
    self._waitpid_lock.release()
  File "/root/miniforge/lib/python3.12/site-packages/torch/distributed/elastic/multiprocessing/api.py", line 85, in _terminate_process_handler
    sigval = signal.Signals(signum)
    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
torch.distributed.elastic.multiprocessing.api.SignalException: Process 3370624 got signal: 15
[W128 06:35:14.145474637 AllocatorConfig.cpp:28] Warning: PYTORCH_CUDA_ALLOC_CONF is deprecated, use PYTORCH_ALLOC_CONF instead (function operator())