/opt/conda/lib/python3.11/site-packages/torch/utils/_pytree.py:185: FutureWarning: optree is installed but the version is too old to support PyTorch Dynamo in C++ pytree. C++ pytree support is disabled. Please consider upgrading optree using `python3 -m pip install --upgrade 'optree>=0.13.0'`.
  warnings.warn(
============================================================
  OPEN-ENDED EVAL: Base vs SFT (Multi-GPU)
  Base model: /workspace/rl4phyx/models/Qwen2.5-VL-3B-Instruct
  SFT model:  /workspace/rl4phyx/RL4Phyx/SFT/checkpoints/sft_qwen25vl_3b_fullft_phyx/final
  Base GPUs:  [0, 1, 2, 3]
  SFT GPUs:   [0, 1, 2, 3]
============================================================

Loaded 1533 test samples
  Mechanics: 276
  Electromagnetism: 275
  Thermodynamics: 255
  Waves/Acoustics: 253
  Optics: 252
  Modern Physics: 222

>>> Starting SFT model inference...
/opt/conda/lib/python3.11/site-packages/torch/utils/_pytree.py:185: FutureWarning: optree is installed but the version is too old to support PyTorch Dynamo in C++ pytree. C++ pytree support is disabled. Please consider upgrading optree using `python3 -m pip install --upgrade 'optree>=0.13.0'`.
  warnings.warn(
/opt/conda/lib/python3.11/site-packages/torch/utils/_pytree.py:185: FutureWarning: optree is installed but the version is too old to support PyTorch Dynamo in C++ pytree. C++ pytree support is disabled. Please consider upgrading optree using `python3 -m pip install --upgrade 'optree>=0.13.0'`.
  warnings.warn(
[sft][GPU 0] Loading model...
/opt/conda/lib/python3.11/site-packages/torch/utils/_pytree.py:185: FutureWarning: optree is installed but the version is too old to support PyTorch Dynamo in C++ pytree. C++ pytree support is disabled. Please consider upgrading optree using `python3 -m pip install --upgrade 'optree>=0.13.0'`.
  warnings.warn(
The tokenizer you are loading from '/workspace/rl4phyx/RL4Phyx/SFT/checkpoints/sft_qwen25vl_3b_fullft_phyx/final' with an incorrect regex pattern: https://huggingface.co/mistralai/Mistral-Small-3.1-24B-Instruct-2503/discussions/84#69121093e8b480e709447d5e. This will lead to incorrect tokenization. You should set the `fix_mistral_regex=True` flag when loading this tokenizer to fix this issue.
`torch_dtype` is deprecated! Use `dtype` instead!
/opt/conda/lib/python3.11/site-packages/torch/utils/_pytree.py:185: FutureWarning: optree is installed but the version is too old to support PyTorch Dynamo in C++ pytree. C++ pytree support is disabled. Please consider upgrading optree using `python3 -m pip install --upgrade 'optree>=0.13.0'`.
  warnings.warn(
[sft][GPU 1] Loading model...
The tokenizer you are loading from '/workspace/rl4phyx/RL4Phyx/SFT/checkpoints/sft_qwen25vl_3b_fullft_phyx/final' with an incorrect regex pattern: https://huggingface.co/mistralai/Mistral-Small-3.1-24B-Instruct-2503/discussions/84#69121093e8b480e709447d5e. This will lead to incorrect tokenization. You should set the `fix_mistral_regex=True` flag when loading this tokenizer to fix this issue.
`torch_dtype` is deprecated! Use `dtype` instead!
[sft][GPU 2] Loading model...
The tokenizer you are loading from '/workspace/rl4phyx/RL4Phyx/SFT/checkpoints/sft_qwen25vl_3b_fullft_phyx/final' with an incorrect regex pattern: https://huggingface.co/mistralai/Mistral-Small-3.1-24B-Instruct-2503/discussions/84#69121093e8b480e709447d5e. This will lead to incorrect tokenization. You should set the `fix_mistral_regex=True` flag when loading this tokenizer to fix this issue.
`torch_dtype` is deprecated! Use `dtype` instead!
[sft][GPU 3] Loading model...
The tokenizer you are loading from '/workspace/rl4phyx/RL4Phyx/SFT/checkpoints/sft_qwen25vl_3b_fullft_phyx/final' with an incorrect regex pattern: https://huggingface.co/mistralai/Mistral-Small-3.1-24B-Instruct-2503/discussions/84#69121093e8b480e709447d5e. This will lead to incorrect tokenization. You should set the `fix_mistral_regex=True` flag when loading this tokenizer to fix this issue.
`torch_dtype` is deprecated! Use `dtype` instead!

Loading checkpoint shards:   0%|          | 0/2 [00:00<?, ?it/s]
Loading checkpoint shards:   0%|          | 0/2 [00:00<?, ?it/s]
Loading checkpoint shards:   0%|          | 0/2 [00:00<?, ?it/s]
Loading checkpoint shards:   0%|          | 0/2 [00:00<?, ?it/s]
Loading checkpoint shards:  50%|█████     | 1/2 [00:01<00:01,  1.54s/it]
Loading checkpoint shards:  50%|█████     | 1/2 [00:01<00:01,  1.24s/it]
Loading checkpoint shards:  50%|█████     | 1/2 [00:01<00:01,  1.10s/it]
Loading checkpoint shards: 100%|██████████| 2/2 [00:02<00:00,  1.01it/s]
Loading checkpoint shards: 100%|██████████| 2/2 [00:02<00:00,  1.07s/it]
[sft][GPU 0] Model loaded. Processing 384 samples.

Loading checkpoint shards:  50%|█████     | 1/2 [00:01<00:01,  1.28s/it]
Loading checkpoint shards: 100%|██████████| 2/2 [00:01<00:00,  1.28it/s]
Loading checkpoint shards: 100%|██████████| 2/2 [00:01<00:00,  1.20it/s]

Loading checkpoint shards: 100%|██████████| 2/2 [00:01<00:00,  1.12it/s]
Loading checkpoint shards: 100%|██████████| 2/2 [00:01<00:00,  1.06it/s]
[sft][GPU 1] Model loaded. Processing 384 samples.
[sft][GPU 2] Model loaded. Processing 384 samples.

Loading checkpoint shards: 100%|██████████| 2/2 [00:02<00:00,  1.05it/s]
Loading checkpoint shards: 100%|██████████| 2/2 [00:02<00:00,  1.01s/it]
[sft][GPU 3] Model loaded. Processing 381 samples.
[sft][GPU 0] 20/384 done
[sft][GPU 3] 20/381 done
[sft][GPU 2] 20/384 done
[sft][GPU 1] 20/384 done
[sft][GPU 0] 40/384 done
[sft][GPU 3] 40/381 done
[sft][GPU 2] 40/384 done
[sft][GPU 1] 40/384 done
[sft][GPU 0] 60/384 done
[sft][GPU 2] 60/384 done
[sft][GPU 3] 60/381 done
[sft][GPU 0] 80/384 done
[sft][GPU 1] 60/384 done
[sft][GPU 2] 80/384 done
[sft][GPU 0] 100/384 done
[sft][GPU 1] 80/384 done
[sft][GPU 3] 80/381 done
[sft][GPU 2] 100/384 done
[sft][GPU 3] 100/381 done
[sft][GPU 0] 120/384 done
[sft][GPU 1] 100/384 done
[sft][GPU 2] 120/384 done
[sft][GPU 3] 120/381 done
[sft][GPU 0] 140/384 done
[sft][GPU 1] 120/384 done
[sft][GPU 2] 140/384 done
[sft][GPU 0] 160/384 done
[sft][GPU 3] 140/381 done
[sft][GPU 1] 140/384 done
[sft][GPU 2] 160/384 done
[sft][GPU 0] 180/384 done
[sft][GPU 1] 160/384 done
[sft][GPU 3] 160/381 done
[sft][GPU 2] 180/384 done
[sft][GPU 0] 200/384 done
[sft][GPU 3] 180/381 done
[sft][GPU 1] 180/384 done
[sft][GPU 2] 200/384 done
[sft][GPU 1] 200/384 done
[sft][GPU 3] 200/381 done
[sft][GPU 0] 220/384 done
[sft][GPU 2] 220/384 done
[sft][GPU 3] 220/381 done
[sft][GPU 1] 220/384 done
[sft][GPU 0] 240/384 done
[sft][GPU 2] 240/384 done
[sft][GPU 3] 240/381 done
[sft][GPU 1] 240/384 done
[sft][GPU 3] 260/381 done
[sft][GPU 0] 260/384 done
[sft][GPU 2] 260/384 done
[sft][GPU 1] 260/384 done
[sft][GPU 0] 280/384 done
[sft][GPU 3] 280/381 done
[sft][GPU 2] 280/384 done
[sft][GPU 0] 300/384 done
[sft][GPU 1] 280/384 done
[sft][GPU 2] 300/384 done
[sft][GPU 1] 300/384 done
[sft][GPU 3] 300/381 done
[sft][GPU 0] 320/384 done
[sft][GPU 2] 320/384 done
[sft][GPU 1] 320/384 done
[sft][GPU 0] 340/384 done
[sft][GPU 2] 340/384 done
[sft][GPU 3] 320/381 done
[sft][GPU 1] 340/384 done
[sft][GPU 0] 360/384 done
[sft][GPU 2] 360/384 done
[sft][GPU 1] 360/384 done
[sft][GPU 2] 380/384 done
[sft][GPU 3] 340/381 done
[sft][GPU 0] 380/384 done
[sft][GPU 2] 384/384 done
[sft][GPU 2] Saved 384 results to /workspace/rl4phyx/RL4Phyx/SFT/sft_eval_footprint/inference_results_phyx_gpu2.jsonl
[sft][GPU 0] 384/384 done
[sft][GPU 0] Saved 384 results to /workspace/rl4phyx/RL4Phyx/SFT/sft_eval_footprint/inference_results_phyx_gpu0.jsonl
[sft][GPU 3] 360/381 done
[sft][GPU 1] 380/384 done
[sft][GPU 1] 384/384 done
[sft][GPU 1] Saved 384 results to /workspace/rl4phyx/RL4Phyx/SFT/sft_eval_footprint/inference_results_phyx_gpu1.jsonl
[sft][GPU 3] 380/381 done
[sft][GPU 3] 381/381 done
[sft][GPU 3] Saved 381 results to /workspace/rl4phyx/RL4Phyx/SFT/sft_eval_footprint/inference_results_phyx_gpu3.jsonl

============================================================
  INFERENCE COMPLETE in 94.1 min
  Base results: 0 → /workspace/rl4phyx/RL4Phyx/SFT/sft_eval_footprint/inference_results_base.jsonl
  SFT results:  1533 → /workspace/rl4phyx/RL4Phyx/SFT/sft_eval_footprint/inference_results_phyx.jsonl
============================================================