Upload rl/autoslm-1781754511-314127dc/seed0/console_rl.txt with huggingface_hub

rl/autoslm-1781754511-314127dc/seed0/console_rl.txt

@@ -0,0 +1,35 @@
+HEARTBEAT {"stage": "boot", "ts": 1781754602.8021507, "run_id": "autoslm-1781754511-314127dc", "mode": "rl", "seed": 0}
+/usr/local/lib/python3.12/dist-packages/trl/generation/__init__.py:22: UserWarning: TRL currently supports vLLM versions from 0.12.0 to 0.19.0. You have version 0.19.1 installed. We recommend installing a supported version to avoid compatibility issues.
+  if is_vllm_available():
+/usr/local/lib/python3.12/dist-packages/trl/generation/vllm_client.py:40: UserWarning: TRL currently supports vLLM versions from 0.12.0 to 0.19.0. You have version 0.19.1 installed. We recommend installing a supported version to avoid compatibility issues.
+  if is_vllm_available():
+/usr/local/lib/python3.12/dist-packages/trl/generation/vllm_generation.py:42: UserWarning: TRL currently supports vLLM versions from 0.12.0 to 0.19.0. You have version 0.19.1 installed. We recommend installing a supported version to avoid compatibility issues.
+  if is_vllm_available():
+HEARTBEAT {"stage": "rl_start", "ts": 1781754609.8739352, "run_id": "autoslm-1781754511-314127dc", "mode": "rl", "seed": 0}
+GPU ready after 0 retries: NVIDIA GeForce RTX 5090
+
+Fetching 13 files:   0%|          | 0/13 [00:00<?, ?it/s]
+Fetching 13 files:   8%|▊         | 1/13 [00:00<00:04,  2.54it/s]
+Fetching 13 files:  23%|██▎       | 3/13 [00:00<00:01,  6.40it/s]
+Fetching 13 files:  54%|█████▍    | 7/13 [00:02<00:02,  2.15it/s]
+Fetching 13 files: 100%|██████████| 13/13 [00:02<00:00,  4.35it/s]
+HEARTBEAT {"stage": "model_prefetched", "ts": 1781754613.7379317, "run_id": "autoslm-1781754511-314127dc", "mode": "rl", "seed": 0, "model": "Qwen/Qwen3.5-0.8B", "download_seconds": 3.1, "hf_transfer": "1"}
+[rl] rollout backend: colocated vLLM
+[rl] dropped 10 prompts over the 768-token prompt budget (engine 1024 - completion 256)
+Traceback (most recent call last):
+  File "/runcode/code/autoslm/engine/worker.py", line 1698, in main
+    handler()
+  File "/runcode/code/autoslm/engine/worker.py", line 1244, in run_rl
+    raise ValueError(
+ValueError: every training prompt exceeds the 768-token prompt budget (engine 1024 - completion 256); raise RL_VLLM_MAX_LEN, lower RL_MAX_COMPLETION, or shorten the environment's prompts
+HEARTBEAT {"stage": "error_rl", "ts": 1781754616.741913, "run_id": "autoslm-1781754511-314127dc", "mode": "rl", "seed": 0, "error": "every training prompt exceeds the 768-token prompt budget (engine 1024 - completion 256); raise RL_VLLM_MAX_LEN, lower RL_MAX_COMPLETION, or shorten the environment's prompts", "diag": {"torch": "2.10.0+cu128", "torch_cuda": "12.8", "cuda_available": true, "device_count": 1, "device_name": "NVIDIA GeForce RTX 5090", "nvidia_smi": "580.159.03, NVIDIA GeForce RTX 5090, 32607 MiB"}}
+Traceback (most recent call last):
+  File "<frozen runpy>", line 198, in _run_module_as_main
+  File "<frozen runpy>", line 88, in _run_code
+  File "/runcode/code/autoslm/engine/worker.py", line 1733, in <module>
+    main()
+  File "/runcode/code/autoslm/engine/worker.py", line 1698, in main
+    handler()
+  File "/runcode/code/autoslm/engine/worker.py", line 1244, in run_rl
+    raise ValueError(
+ValueError: every training prompt exceeds the 768-token prompt budget (engine 1024 - completion 256); raise RL_VLLM_MAX_LEN, lower RL_MAX_COMPLETION, or shorten the environment's prompts