Files
ModelHub XC 8c7e5aea19 初始化项目,由ModelHub XC社区提供模型
Model: logan7000/llm-math345-gt-phi4mini-endpoint
Source: Original Platform
2026-11-13 06:52:18 +08:00

2056 lines
346 KiB
Plaintext

W0812 22:27:10.568000 3510464 torch/distributed/run.py:852]
W0812 22:27:10.568000 3510464 torch/distributed/run.py:852] *****************************************
W0812 22:27:10.568000 3510464 torch/distributed/run.py:852] 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.
W0812 22:27:10.568000 3510464 torch/distributed/run.py:852] *****************************************
================================================================================
RUN CONFIGURATION
================================================================================
================================================================================WandB Run Name: gt__microsoft_Phi-4-mini-instruct__20260812_222659_lr3e-6_bs1536
Output Directory: /weka/scratch/jhu/dssg2026-ext-rghani1/yyang331/trl-projects-llm/work_dirs/stage/gt__microsoft_Phi-4-mini-instruct__20260812_222659
Num Generations: 12
Temperature: 1.0
Max Completion Length: 3072
RUN CONFIGURATION================================================================================
================================================================================
================================================================================
WandB Run Name: gt__microsoft_Phi-4-mini-instruct__20260812_222659_lr3e-6_bs1536Loading model with dtype: float32
Output Directory: /weka/scratch/jhu/dssg2026-ext-rghani1/yyang331/trl-projects-llm/work_dirs/stage/gt__microsoft_Phi-4-mini-instruct__20260812_222659Using attention implementation: flash_attention_2
Num Generations: 12================================================================================
Temperature: 1.0
Max Completion Length: 3072
================================================================================
================================================================================
Loading model with dtype: float32
Using attention implementation: flash_attention_2
================================================================================
================================================================================
RUN CONFIGURATION
================================================================================
WandB Run Name: gt__microsoft_Phi-4-mini-instruct__20260812_222659_lr3e-6_bs1536
Output Directory: /weka/scratch/jhu/dssg2026-ext-rghani1/yyang331/trl-projects-llm/work_dirs/stage/gt__microsoft_Phi-4-mini-instruct__20260812_222659
Num Generations: 12
Temperature: 1.0
Max Completion Length: 3072
================================================================================
================================================================================
Loading model with dtype: float32
Using attention implementation: flash_attention_2
================================================================================
================================================================================
RUN CONFIGURATION
================================================================================
WandB Run Name: gt__microsoft_Phi-4-mini-instruct__20260812_222659_lr3e-6_bs1536
Output Directory: /weka/scratch/jhu/dssg2026-ext-rghani1/yyang331/trl-projects-llm/work_dirs/stage/gt__microsoft_Phi-4-mini-instruct__20260812_222659
Num Generations: 12
Temperature: 1.0
Max Completion Length: 3072
================================================================================
================================================================================
Loading model with dtype: float32
Using attention implementation: flash_attention_2
================================================================================
================================================================================
RUN CONFIGURATION
================================================================================
WandB Run Name: gt__microsoft_Phi-4-mini-instruct__20260812_222659_lr3e-6_bs1536
Output Directory: /weka/scratch/jhu/dssg2026-ext-rghani1/yyang331/trl-projects-llm/work_dirs/stage/gt__microsoft_Phi-4-mini-instruct__20260812_222659
Num Generations: 12
Temperature: 1.0
Max Completion Length: 3072
================================================================================
================================================================================
Loading model with dtype: float32
Using attention implementation: flash_attention_2
================================================================================
================================================================================
RUN CONFIGURATION
================================================================================
WandB Run Name: gt__microsoft_Phi-4-mini-instruct__20260812_222659_lr3e-6_bs1536
Output Directory: /weka/scratch/jhu/dssg2026-ext-rghani1/yyang331/trl-projects-llm/work_dirs/stage/gt__microsoft_Phi-4-mini-instruct__20260812_222659
Num Generations: 12
Temperature: 1.0
Max Completion Length: 3072
================================================================================
wandb: [wandb.login()] Loaded credentials for https://api.wandb.ai from WANDB_API_KEY.
wandb: Currently logged in as: logan-yang2002 (logan-yang2002-johns-hopkins-university) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
wandb: Tracking run with wandb version 0.25.1
wandb: Run data is saved locally in /weka/home/jhu/dssg2026-ext-yyang331/raw/tmp/trl-projects/wandb/run-20260812_222730-5cwwrnh0
wandb: Run `wandb offline` to turn off syncing.
wandb: Syncing run gt__microsoft_Phi-4-mini-instruct__20260812_222659_lr3e-6_bs1536
wandb: ⭐️ View project at https://wandb.ai/logan-yang2002-johns-hopkins-university/grpo-training
wandb: 🚀 View run at https://wandb.ai/logan-yang2002-johns-hopkins-university/grpo-training/runs/5cwwrnh0
wandb: Detected [openai, verifiers] in use.
wandb: Use W&B Weave for improved LLM call tracing. Install Weave with `pip install weave` then add `import weave` to the top of your script.
wandb: For more information, check out the docs at: https://weave-docs.wandb.ai/
================================================================================
Loading model with dtype: float32
Using attention implementation: flash_attention_2
================================================================================
`torch_dtype` is deprecated! Use `dtype` instead!
`torch_dtype` is deprecated! Use `dtype` instead!
`torch_dtype` is deprecated! Use `dtype` instead!
`torch_dtype` is deprecated! Use `dtype` instead!
`torch_dtype` is deprecated! Use `dtype` instead!
`torch_dtype` is deprecated! Use `dtype` instead!
Fetching 2 files: 0%| | 0/2 [00:00<?, ?it/s]
Fetching 2 files: 0%| | 0/2 [00:00<?, ?it/s]
Fetching 2 files: 0%| | 0/2 [00:00<?, ?it/s]
Fetching 2 files: 0%| | 0/2 [00:00<?, ?it/s]
Fetching 2 files: 0%| | 0/2 [00:00<?, ?it/s]
Fetching 2 files: 0%| | 0/2 [00:00<?, ?it/s]
Fetching 2 files: 50%|█████ | 1/2 [00:08<00:08, 8.10s/it]
Fetching 2 files: 50%|█████ | 1/2 [00:08<00:08, 8.06s/it]
Fetching 2 files: 50%|█████ | 1/2 [00:08<00:08, 8.07s/it]
Fetching 2 files: 50%|█████ | 1/2 [00:08<00:08, 8.09s/it]
Fetching 2 files: 50%|█████ | 1/2 [00:08<00:08, 8.04s/it]
Fetching 2 files: 50%|█████ | 1/2 [00:08<00:08, 8.11s/it]
Fetching 2 files: 100%|██████████| 2/2 [00:09<00:00, 4.76s/it]
Fetching 2 files: 100%|██████████| 2/2 [00:09<00:00, 4.76s/it]
Fetching 2 files: 100%|██████████| 2/2 [00:09<00:00, 4.74s/it]
Fetching 2 files: 100%|██████████| 2/2 [00:09<00:00, 4.75s/it]
Fetching 2 files: 100%|██████████| 2/2 [00:09<00:00, 4.75s/it]
Fetching 2 files: 100%|██████████| 2/2 [00:09<00:00, 4.75s/it]
Flash Attention 2 only supports torch.float16 and torch.bfloat16 dtypes, but the current dype in Phi3ForCausalLM is torch.float32. You should run training or inference using Automatic Mixed-Precision via the `with torch.autocast(device_type='torch_device'):` decorator, or load the model with the `dtype` argument. Example: `model = AutoModel.from_pretrained("openai/whisper-tiny", attn_implementation="flash_attention_2", dtype=torch.float16)`
Flash Attention 2 only supports torch.float16 and torch.bfloat16 dtypes, but the current dype in Phi3Model is torch.float32. You should run training or inference using Automatic Mixed-Precision via the `with torch.autocast(device_type='torch_device'):` decorator, or load the model with the `dtype` argument. Example: `model = AutoModel.from_pretrained("openai/whisper-tiny", attn_implementation="flash_attention_2", dtype=torch.float16)`
Flash Attention 2 only supports torch.float16 and torch.bfloat16 dtypes, but the current dype in Phi3ForCausalLM is torch.float32. You should run training or inference using Automatic Mixed-Precision via the `with torch.autocast(device_type='torch_device'):` decorator, or load the model with the `dtype` argument. Example: `model = AutoModel.from_pretrained("openai/whisper-tiny", attn_implementation="flash_attention_2", dtype=torch.float16)`
Flash Attention 2 only supports torch.float16 and torch.bfloat16 dtypes, but the current dype in Phi3Model is torch.float32. You should run training or inference using Automatic Mixed-Precision via the `with torch.autocast(device_type='torch_device'):` decorator, or load the model with the `dtype` argument. Example: `model = AutoModel.from_pretrained("openai/whisper-tiny", attn_implementation="flash_attention_2", dtype=torch.float16)`
Fetching 2 files: 100%|██████████| 2/2 [00:09<00:00, 4.76s/it]
Fetching 2 files: 100%|██████████| 2/2 [00:09<00:00, 4.76s/it]
Flash Attention 2 only supports torch.float16 and torch.bfloat16 dtypes, but the current dype in Phi3ForCausalLM is torch.float32. You should run training or inference using Automatic Mixed-Precision via the `with torch.autocast(device_type='torch_device'):` decorator, or load the model with the `dtype` argument. Example: `model = AutoModel.from_pretrained("openai/whisper-tiny", attn_implementation="flash_attention_2", dtype=torch.float16)`
Fetching 2 files: 100%|██████████| 2/2 [00:09<00:00, 4.74s/it]
Fetching 2 files: 100%|██████████| 2/2 [00:09<00:00, 4.74s/it]
Flash Attention 2 only supports torch.float16 and torch.bfloat16 dtypes, but the current dype in Phi3Model is torch.float32. You should run training or inference using Automatic Mixed-Precision via the `with torch.autocast(device_type='torch_device'):` decorator, or load the model with the `dtype` argument. Example: `model = AutoModel.from_pretrained("openai/whisper-tiny", attn_implementation="flash_attention_2", dtype=torch.float16)`
Flash Attention 2 only supports torch.float16 and torch.bfloat16 dtypes, but the current dype in Phi3ForCausalLM is torch.float32. You should run training or inference using Automatic Mixed-Precision via the `with torch.autocast(device_type='torch_device'):` decorator, or load the model with the `dtype` argument. Example: `model = AutoModel.from_pretrained("openai/whisper-tiny", attn_implementation="flash_attention_2", dtype=torch.float16)`
Flash Attention 2 only supports torch.float16 and torch.bfloat16 dtypes, but the current dype in Phi3Model is torch.float32. You should run training or inference using Automatic Mixed-Precision via the `with torch.autocast(device_type='torch_device'):` decorator, or load the model with the `dtype` argument. Example: `model = AutoModel.from_pretrained("openai/whisper-tiny", attn_implementation="flash_attention_2", dtype=torch.float16)`
Fetching 2 files: 100%|██████████| 2/2 [00:09<00:00, 4.78s/it]
Fetching 2 files: 100%|██████████| 2/2 [00:09<00:00, 4.78s/it]
Flash Attention 2 only supports torch.float16 and torch.bfloat16 dtypes, but the current dype in Phi3ForCausalLM is torch.float32. You should run training or inference using Automatic Mixed-Precision via the `with torch.autocast(device_type='torch_device'):` decorator, or load the model with the `dtype` argument. Example: `model = AutoModel.from_pretrained("openai/whisper-tiny", attn_implementation="flash_attention_2", dtype=torch.float16)`
Flash Attention 2 only supports torch.float16 and torch.bfloat16 dtypes, but the current dype in Phi3Model is torch.float32. You should run training or inference using Automatic Mixed-Precision via the `with torch.autocast(device_type='torch_device'):` decorator, or load the model with the `dtype` argument. Example: `model = AutoModel.from_pretrained("openai/whisper-tiny", attn_implementation="flash_attention_2", dtype=torch.float16)`
Flash Attention 2 only supports torch.float16 and torch.bfloat16 dtypes, but the current dype in Phi3ForCausalLM is torch.float32. You should run training or inference using Automatic Mixed-Precision via the `with torch.autocast(device_type='torch_device'):` decorator, or load the model with the `dtype` argument. Example: `model = AutoModel.from_pretrained("openai/whisper-tiny", attn_implementation="flash_attention_2", dtype=torch.float16)`
Flash Attention 2 only supports torch.float16 and torch.bfloat16 dtypes, but the current dype in Phi3Model is torch.float32. You should run training or inference using Automatic Mixed-Precision via the `with torch.autocast(device_type='torch_device'):` decorator, or load the model with the `dtype` argument. Example: `model = AutoModel.from_pretrained("openai/whisper-tiny", attn_implementation="flash_attention_2", dtype=torch.float16)`
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: 0%| | 0/2 [00:00<?, ?it/s]
Loading checkpoint shards: 0%| | 0/2 [00:00<?, ?it/s]
Loading checkpoint shards: 50%|█████ | 1/2 [00:03<00:03, 3.05s/it]
Loading checkpoint shards: 50%|█████ | 1/2 [00:03<00:03, 3.05s/it]
Loading checkpoint shards: 50%|█████ | 1/2 [00:03<00:03, 3.06s/it]
Loading checkpoint shards: 50%|█████ | 1/2 [00:03<00:03, 3.05s/it]
Loading checkpoint shards: 50%|█████ | 1/2 [00:03<00:03, 3.05s/it]
Loading checkpoint shards: 50%|█████ | 1/2 [00:03<00:03, 3.18s/it]
Loading checkpoint shards: 100%|██████████| 2/2 [00:05<00:00, 2.43s/it]
Loading checkpoint shards: 100%|██████████| 2/2 [00:05<00:00, 2.43s/it]
Loading checkpoint shards: 100%|██████████| 2/2 [00:05<00:00, 2.43s/it]
Loading checkpoint shards: 100%|██████████| 2/2 [00:05<00:00, 2.43s/it]
Loading checkpoint shards: 100%|██████████| 2/2 [00:05<00:00, 2.43s/it]
Loading checkpoint shards: 100%|██████████| 2/2 [00:05<00:00, 2.52s/it]
Loading checkpoint shards: 100%|██████████| 2/2 [00:05<00:00, 2.52s/it]
Loading checkpoint shards: 100%|██████████| 2/2 [00:05<00:00, 2.52s/it]
Loading checkpoint shards: 100%|██████████| 2/2 [00:05<00:00, 2.52s/it]
Loading checkpoint shards: 100%|██████████| 2/2 [00:05<00:00, 2.52s/it]
Loading checkpoint shards: 100%|██████████| 2/2 [00:05<00:00, 2.45s/it]
Loading checkpoint shards: 100%|██████████| 2/2 [00:05<00:00, 2.56s/it]
The argument `trust_remote_code` is to be used with Auto classes. It has no effect here and is ignored.
The argument `trust_remote_code` is to be used with Auto classes. It has no effect here and is ignored.
The argument `trust_remote_code` is to be used with Auto classes. It has no effect here and is ignored.
The argument `trust_remote_code` is to be used with Auto classes. It has no effect here and is ignored.
The argument `trust_remote_code` is to be used with Auto classes. It has no effect here and is ignored.
The argument `trust_remote_code` is to be used with Auto classes. It has no effect here and is ignored.
The argument `trust_remote_code` is to be used with Auto classes. It has no effect here and is ignored.
The argument `trust_remote_code` is to be used with Auto classes. It has no effect here and is ignored.
The argument `trust_remote_code` is to be used with Auto classes. It has no effect here and is ignored.
The argument `trust_remote_code` is to be used with Auto classes. It has no effect here and is ignored.
The argument `trust_remote_code` is to be used with Auto classes. It has no effect here and is ignored.
The argument `trust_remote_code` is to be used with Auto classes. It has no effect here and is ignored.
A new version of the following files was downloaded from https://huggingface.co/microsoft/Phi-4-mini-instruct:
- configuration_phi3.py
. Make sure to double-check they do not contain any added malicious code. To avoid downloading new versions of the code file, you can pin a revision.
A new version of the following files was downloaded from https://huggingface.co/microsoft/Phi-4-mini-instruct:
- configuration_phi3.py
. Make sure to double-check they do not contain any added malicious code. To avoid downloading new versions of the code file, you can pin a revision.
A new version of the following files was downloaded from https://huggingface.co/microsoft/Phi-4-mini-instruct:
- configuration_phi3.py
. Make sure to double-check they do not contain any added malicious code. To avoid downloading new versions of the code file, you can pin a revision.
A new version of the following files was downloaded from https://huggingface.co/microsoft/Phi-4-mini-instruct:
- configuration_phi3.py
. Make sure to double-check they do not contain any added malicious code. To avoid downloading new versions of the code file, you can pin a revision.
A new version of the following files was downloaded from https://huggingface.co/microsoft/Phi-4-mini-instruct:
- configuration_phi3.py
. Make sure to double-check they do not contain any added malicious code. To avoid downloading new versions of the code file, you can pin a revision.
A new version of the following files was downloaded from https://huggingface.co/microsoft/Phi-4-mini-instruct:
- configuration_phi3.py
. Make sure to double-check they do not contain any added malicious code. To avoid downloading new versions of the code file, you can pin a revision.
<frozen importlib._bootstrap_external>:1297: FutureWarning: The cuda.cudart module is deprecated and will be removed in a future release, please switch to use the cuda.bindings.runtime module instead.
<frozen importlib._bootstrap_external>:1297: FutureWarning: The cuda.cudart module is deprecated and will be removed in a future release, please switch to use the cuda.bindings.runtime module instead.
<frozen importlib._bootstrap_external>:1297: FutureWarning: The cuda.cudart module is deprecated and will be removed in a future release, please switch to use the cuda.bindings.runtime module instead.
<frozen importlib._bootstrap_external>:1297: FutureWarning: The cuda.cudart module is deprecated and will be removed in a future release, please switch to use the cuda.bindings.runtime module instead.
<frozen importlib._bootstrap_external>:1297: FutureWarning: The cuda.cudart module is deprecated and will be removed in a future release, please switch to use the cuda.bindings.runtime module instead.
<frozen importlib._bootstrap_external>:1297: FutureWarning: The cuda.nvrtc module is deprecated and will be removed in a future release, please switch to use the cuda.bindings.nvrtc module instead.
<frozen importlib._bootstrap_external>:1297: FutureWarning: The cuda.nvrtc module is deprecated and will be removed in a future release, please switch to use the cuda.bindings.nvrtc module instead.
<frozen importlib._bootstrap_external>:1297: FutureWarning: The cuda.nvrtc module is deprecated and will be removed in a future release, please switch to use the cuda.bindings.nvrtc module instead.
<frozen importlib._bootstrap_external>:1297: FutureWarning: The cuda.nvrtc module is deprecated and will be removed in a future release, please switch to use the cuda.bindings.nvrtc module instead.
<frozen importlib._bootstrap_external>:1297: FutureWarning: The cuda.nvrtc module is deprecated and will be removed in a future release, please switch to use the cuda.bindings.nvrtc module instead.
<frozen importlib._bootstrap_external>:1297: FutureWarning: The cuda.cudart module is deprecated and will be removed in a future release, please switch to use the cuda.bindings.runtime module instead.
<frozen importlib._bootstrap_external>:1297: FutureWarning: The cuda.nvrtc module is deprecated and will be removed in a future release, please switch to use the cuda.bindings.nvrtc module instead.
Loading safetensors checkpoint shards: 0% Completed | 0/2 [00:00<?, ?it/s]
Loading safetensors checkpoint shards: 50% Completed | 1/2 [00:00<00:00, 1.61it/s]
Loading safetensors checkpoint shards: 100% Completed | 2/2 [00:01<00:00, 2.07it/s]
Loading safetensors checkpoint shards: 100% Completed | 2/2 [00:01<00:00, 1.98it/s]
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 0%| | 0/51 [00:00<?, ?it/s]
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 4%|▍ | 2/51 [00:00<00:02, 18.75it/s]
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 10%|▉ | 5/51 [00:00<00:02, 21.26it/s]
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 16%|█▌ | 8/51 [00:00<00:01, 22.67it/s]
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 22%|██▏ | 11/51 [00:00<00:01, 24.51it/s]
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 27%|██▋ | 14/51 [00:00<00:01, 25.87it/s]
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 33%|███▎ | 17/51 [00:00<00:01, 26.55it/s]
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 39%|███▉ | 20/51 [00:00<00:01, 27.36it/s]
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 45%|████▌ | 23/51 [00:00<00:01, 27.70it/s]
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 51%|█████ | 26/51 [00:00<00:00, 28.21it/s]
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 57%|█████▋ | 29/51 [00:01<00:00, 28.43it/s]
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 63%|██████▎ | 32/51 [00:01<00:00, 28.86it/s]
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 69%|██████▊ | 35/51 [00:01<00:00, 28.88it/s]
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 75%|███████▍ | 38/51 [00:01<00:00, 28.98it/s]
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 80%|████████ | 41/51 [00:01<00:00, 29.20it/s]
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 86%|████████▋ | 44/51 [00:01<00:00, 29.29it/s]
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 92%|█████████▏| 47/51 [00:01<00:00, 29.40it/s]
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 98%|█████████▊| 50/51 [00:01<00:00, 27.95it/s]
Capturing CUDA graphs (mixed prefill-decode, PIECEWISE): 100%|██████████| 51/51 [00:01<00:00, 27.23it/s]
Capturing CUDA graphs (decode, FULL): 0%| | 0/35 [00:00<?, ?it/s]
Capturing CUDA graphs (decode, FULL): 11%|█▏ | 4/35 [00:00<00:00, 31.98it/s]
Capturing CUDA graphs (decode, FULL): 23%|██▎ | 8/35 [00:00<00:00, 32.35it/s]
Capturing CUDA graphs (decode, FULL): 34%|███▍ | 12/35 [00:00<00:00, 33.33it/s]
Capturing CUDA graphs (decode, FULL): 46%|████▌ | 16/35 [00:00<00:00, 34.10it/s]
Capturing CUDA graphs (decode, FULL): 57%|█████▋ | 20/35 [00:00<00:00, 34.91it/s]
Capturing CUDA graphs (decode, FULL): 69%|██████▊ | 24/35 [00:00<00:00, 35.92it/s]
Capturing CUDA graphs (decode, FULL): 80%|████████ | 28/35 [00:00<00:00, 36.64it/s]
Capturing CUDA graphs (decode, FULL): 91%|█████████▏| 32/35 [00:00<00:00, 36.93it/s]
Capturing CUDA graphs (decode, FULL): 100%|██████████| 35/35 [00:00<00:00, 35.53it/s]
0%| | 0/136 [00:00<?, ?it/s]
0%| | 0/84 [00:00<?, ?it/s]
2%|▏ | 2/84 [00:07<05:26, 3.99s/it]
4%|▎ | 3/84 [00:12<05:46, 4.28s/it]
5%|▍ | 4/84 [00:19<07:09, 5.37s/it]
6%|▌ | 5/84 [00:26<07:37, 5.79s/it]
7%|▋ | 6/84 [00:32<07:41, 5.91s/it]
8%|▊ | 7/84 [00:41<08:40, 6.76s/it]
10%|▉ | 8/84 [00:45<07:29, 5.91s/it]
11%|█ | 9/84 [00:53<08:28, 6.78s/it]
12%|█▏ | 10/84 [00:59<07:54, 6.41s/it]
13%|█▎ | 11/84 [01:23<14:09, 11.63s/it]
14%|█▍ | 12/84 [01:29<12:09, 10.13s/it]
15%|█▌ | 13/84 [01:36<10:42, 9.05s/it]
17%|█▋ | 14/84 [01:43<10:01, 8.60s/it]
18%|█▊ | 15/84 [01:50<09:23, 8.16s/it]
19%|█▉ | 16/84 [02:14<14:30, 12.80s/it]
20%|██ | 17/84 [02:20<11:51, 10.63s/it]
21%|██▏ | 18/84 [02:24<09:44, 8.86s/it]
23%|██▎ | 19/84 [02:32<09:18, 8.59s/it]
24%|██▍ | 20/84 [02:43<09:57, 9.34s/it]
25%|██▌ | 21/84 [02:51<09:07, 8.69s/it]
26%|██▌ | 22/84 [02:56<07:59, 7.74s/it]
27%|██▋ | 23/84 [03:05<08:09, 8.02s/it]
29%|██▊ | 24/84 [03:11<07:32, 7.54s/it]
30%|██▉ | 25/84 [03:18<07:08, 7.26s/it]
31%|███ | 26/84 [03:21<05:57, 6.16s/it]
32%|███▏ | 27/84 [03:45<10:48, 11.38s/it]
33%|███▎ | 28/84 [03:56<10:38, 11.39s/it]
35%|███▍ | 29/84 [04:02<08:55, 9.75s/it]
36%|███▌ | 30/84 [04:08<07:47, 8.66s/it]
37%|███▋ | 31/84 [04:15<07:12, 8.16s/it]
38%|███▊ | 32/84 [04:20<06:13, 7.18s/it]
39%|███▉ | 33/84 [04:26<05:36, 6.59s/it]
40%|████ | 34/84 [04:33<05:39, 6.78s/it]
42%|████▏ | 35/84 [04:57<09:43, 11.90s/it]
43%|████▎ | 36/84 [05:04<08:28, 10.59s/it]
44%|████▍ | 37/84 [05:10<07:11, 9.17s/it]
45%|████▌ | 38/84 [05:34<10:20, 13.49s/it]
46%|████▋ | 39/84 [05:40<08:28, 11.29s/it]
48%|████▊ | 40/84 [05:50<08:07, 11.07s/it]
49%|████▉ | 41/84 [05:55<06:36, 9.22s/it]
50%|█████ | 42/84 [06:05<06:31, 9.33s/it]
51%|█████ | 43/84 [06:11<05:42, 8.36s/it]
52%|█████▏ | 44/84 [06:15<04:48, 7.20s/it]
54%|█████▎ | 45/84 [06:22<04:30, 6.95s/it]
55%|█████▍ | 46/84 [06:28<04:18, 6.81s/it]
56%|█████▌ | 47/84 [06:34<03:56, 6.38s/it]
57%|█████▋ | 48/84 [06:42<04:14, 7.08s/it]
58%|█████▊ | 49/84 [06:49<03:58, 6.82s/it]
60%|█████▉ | 50/84 [06:58<04:19, 7.64s/it]
61%|██████ | 51/84 [07:04<03:50, 6.99s/it]
62%|██████▏ | 52/84 [07:13<04:03, 7.60s/it]
63%|██████▎ | 53/84 [07:17<03:27, 6.68s/it]
64%|██████▍ | 54/84 [07:23<03:09, 6.32s/it]
65%|██████▌ | 55/84 [07:33<03:40, 7.61s/it]
67%|██████▋ | 56/84 [07:35<02:48, 6.02s/it]
68%|██████▊ | 57/84 [07:41<02:40, 5.94s/it]
69%|██████▉ | 58/84 [07:45<02:19, 5.37s/it]
70%|███████ | 59/84 [07:52<02:25, 5.82s/it]
71%|███████▏ | 60/84 [07:57<02:15, 5.64s/it]
73%|███████▎ | 61/84 [08:03<02:06, 5.50s/it]
74%|███████▍ | 62/84 [08:09<02:07, 5.80s/it]
75%|███████▌ | 63/84 [08:19<02:30, 7.17s/it]
76%|███████▌ | 64/84 [08:27<02:28, 7.41s/it]
77%|███████▋ | 65/84 [08:43<03:05, 9.79s/it]
79%|███████▊ | 66/84 [08:46<02:23, 7.95s/it]
80%|███████▉ | 67/84 [08:53<02:10, 7.65s/it]
81%|████████ | 68/84 [09:03<02:11, 8.25s/it]
82%|████████▏ | 69/84 [09:09<01:55, 7.68s/it]
83%|████████▎ | 70/84 [09:16<01:41, 7.25s/it]
85%|████████▍ | 71/84 [09:20<01:22, 6.36s/it]
86%|████████▌ | 72/84 [09:26<01:14, 6.20s/it]
87%|████████▋ | 73/84 [09:32<01:09, 6.30s/it]
88%|████████▊ | 74/84 [09:35<00:53, 5.30s/it]
89%|████████▉ | 75/84 [09:41<00:48, 5.44s/it]
90%|█████████ | 76/84 [09:50<00:52, 6.61s/it]
92%|█████████▏| 77/84 [09:55<00:42, 6.11s/it]
93%|█████████▎| 78/84 [10:03<00:40, 6.69s/it]
94%|█████████▍| 79/84 [10:09<00:31, 6.33s/it]
95%|█████████▌| 80/84 [10:14<00:23, 5.91s/it]
96%|█████████▋| 81/84 [10:19<00:17, 5.84s/it]
98%|█████████▊| 82/84 [10:27<00:12, 6.34s/it]
99%|█████████▉| 83/84 [10:33<00:06, 6.30s/it]
100%|██████████| 84/84 [10:38<00:00, 5.78s/it]
{'eval_loss': 0.0, 'eval_runtime': 645.7404, 'eval_samples_per_second': 0.774, 'eval_steps_per_second': 0.13, 'eval_num_tokens': 0.0, 'eval_completions/mean_length': 556.7480287097749, 'eval_completions/min_length': 248.51190476190476, 'eval_completions/max_length': 1050.8809523809523, 'eval_completions/clipped_ratio': 0.00992063521629288, 'eval_completions/mean_terminated_length': 531.9472343808129, 'eval_completions/min_terminated_length': 248.51190476190476, 'eval_completions/max_terminated_length': 925.8571428571429, 'eval_rewards/reward_correctness/mean': 0.6587301629285017, 'eval_rewards/reward_correctness/std': 0.4356326439550945, 'eval_reward': 0.6587301735721883, 'eval_reward_std': 0.43563264750299, 'eval_frac_reward_zero_std': 1.0, 'eval_sampling/sampling_logp_difference/mean': 0.06213199950399853, 'eval_sampling/sampling_logp_difference/max': 2.6867869666644504, 'eval_sampling/importance_sampling_ratio/min': 7.668724304032257e-06, 'eval_sampling/importance_sampling_ratio/mean': 0.01892138532918845, 'eval_sampling/importance_sampling_ratio/max': 0.10470169515777646, 'eval_entropy': 0.2654567967568125, 'eval_clip_ratio/low_mean': 0.0, 'eval_clip_ratio/low_min': 0.0, 'eval_clip_ratio/high_mean': 0.0, 'eval_clip_ratio/high_max': 0.0, 'eval_clip_ratio/region_mean': 0.0, 'epoch': 0}
0%| | 0/136 [10:45<?, ?it/s]
100%|██████████| 84/84 [10:38<00:00, 5.78s/it]
[2026-08-12 22:40:56] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,0)
[2026-08-12 22:40:56] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro """ @(1,22)
Open LaTeX blocks:
@(1,22) arguments of macro """
[2026-08-12 22:40:56] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,1)
[2026-08-12 22:40:56] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Bad \end macro: expected {environmentname} @(1,98)
[2026-08-12 22:40:57] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,7)
Could not estimate the number of tokens of the input, floating-point operations will not be computed
Could not estimate the number of tokens of the input, floating-point operations will not be computed
Could not estimate the number of tokens of the input, floating-point operations will not be computed
Could not estimate the number of tokens of the input, floating-point operations will not be computed
Could not estimate the number of tokens of the input, floating-point operations will not be computed
Could not estimate the number of tokens of the input, floating-point operations will not be computed
1%| | 1/136 [15:59<35:59:56, 959.98s/it]
{'loss': -0.0115, 'grad_norm': 0.0848995053910046, 'learning_rate': 0.0, 'num_tokens': 1152974.0, 'completions/mean_length': 649.4779052734375, 'completions/min_length': 85.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.0240885429084301, 'completions/mean_terminated_length': 589.7718505859375, 'completions/min_terminated_length': 85.0, 'completions/max_terminated_length': 2792.0, 'rewards/reward_correctness/mean': 0.4778645932674408, 'rewards/reward_correctness/std': 0.49967244267463684, 'reward': 0.4778645932674408, 'reward_std': 0.49967244267463684, 'frac_reward_zero_std': 0.4765625, 'sampling/sampling_logp_difference/mean': 0.03527523949742317, 'sampling/sampling_logp_difference/max': 13.99999713897705, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.42810118198394775, 'sampling/importance_sampling_ratio/max': 2.935307502746582, 'entropy': 0.5836714202305302, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 313.13801932055503, 'epoch': 0.01}
1%| | 1/136 [15:59<35:59:56, 959.98s/it][2026-08-12 22:46:01] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,110)
[2026-08-12 22:46:01] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,29)
[2026-08-12 22:46:01] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,17)
1%|▏ | 2/136 [20:33<20:42:38, 556.40s/it]
{'loss': 0.0013, 'grad_norm': 0.06942535958988857, 'learning_rate': 6.000000000000001e-07, 'num_tokens': 2271322.0, 'completions/mean_length': 633.5443115234375, 'completions/min_length': 88.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.02278645895421505, 'completions/mean_terminated_length': 576.7142333984375, 'completions/min_terminated_length': 88.0, 'completions/max_terminated_length': 2986.0, 'rewards/reward_correctness/mean': 0.4720052182674408, 'rewards/reward_correctness/std': 0.4993782639503479, 'reward': 0.4720052182674408, 'reward_std': 0.4993782341480255, 'frac_reward_zero_std': 0.40625, 'sampling/sampling_logp_difference/mean': 0.023212533444166183, 'sampling/sampling_logp_difference/max': 1.144627571105957, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5243185758590698, 'sampling/importance_sampling_ratio/max': 2.9959943294525146, 'entropy': 0.578207426937297, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 272.8161257347092, 'epoch': 0.03}
1%|▏ | 2/136 [20:33<20:42:38, 556.40s/it][2026-08-12 22:50:37] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,1)
[2026-08-12 22:50:37] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Bad \end macro: expected {environmentname} @(1,40)
[2026-08-12 22:50:37] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,1)
[2026-08-12 22:50:38] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,1)
2%|▏ | 3/136 [25:02<15:41:43, 424.83s/it]
{'loss': -0.0343, 'grad_norm': 0.09186588323837147, 'learning_rate': 1.2000000000000002e-06, 'num_tokens': 3370264.0, 'completions/mean_length': 632.01953125, 'completions/min_length': 32.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.02473958395421505, 'completions/mean_terminated_length': 570.1242065429688, 'completions/min_terminated_length': 32.0, 'completions/max_terminated_length': 2565.0, 'rewards/reward_correctness/mean': 0.5319010615348816, 'rewards/reward_correctness/std': 0.4991438090801239, 'reward': 0.5319010615348816, 'reward_std': 0.4991437494754791, 'frac_reward_zero_std': 0.3515625, 'sampling/sampling_logp_difference/mean': 0.02268599346280098, 'sampling/sampling_logp_difference/max': 0.9646859169006348, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5294168591499329, 'sampling/importance_sampling_ratio/max': 2.989497184753418, 'entropy': 0.5463347954209894, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 267.1632728436962, 'epoch': 0.04}
2%|▏ | 3/136 [25:02<15:41:43, 424.83s/it][2026-08-12 22:55:06] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,15)
[2026-08-12 22:55:06] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,8)
[2026-08-12 22:55:06] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,1)
[2026-08-12 22:55:06] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Bad \end macro: expected {environmentname} @(1,138)
[2026-08-12 22:55:06] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,1)
[2026-08-12 22:55:06] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Bad \end macro: expected {environmentname} @(1,138)
[2026-08-12 22:55:06] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,14)
3%|▎ | 4/136 [29:36<13:23:36, 365.28s/it]
{'loss': -0.0568, 'grad_norm': 0.12364894599624972, 'learning_rate': 1.8e-06, 'num_tokens': 4516584.0, 'completions/mean_length': 649.9661865234375, 'completions/min_length': 7.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.02669270895421505, 'completions/mean_terminated_length': 583.54248046875, 'completions/min_terminated_length': 7.0, 'completions/max_terminated_length': 2661.0, 'rewards/reward_correctness/mean': 0.423828125, 'rewards/reward_correctness/std': 0.4943247139453888, 'reward': 0.423828125, 'reward_std': 0.4943247139453888, 'frac_reward_zero_std': 0.34375, 'sampling/sampling_logp_difference/mean': 0.02456444501876831, 'sampling/sampling_logp_difference/max': 0.9797148704528809, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5203623175621033, 'sampling/importance_sampling_ratio/max': 2.9678268432617188, 'entropy': 0.6193987773731351, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 272.84204353392124, 'epoch': 0.06}
3%|▎ | 4/136 [29:36<13:23:36, 365.28s/it][2026-08-12 22:59:39] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\)' @(1,12)
[2026-08-12 22:59:39] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,11)
[2026-08-12 22:59:39] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,302)
[2026-08-12 22:59:39] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,164)
[2026-08-12 22:59:39] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,7)
[2026-08-12 22:59:39] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,19)
4%|▎ | 5/136 [34:15<12:09:44, 334.23s/it]
{'loss': -0.0142, 'grad_norm': 0.08114140165877926, 'learning_rate': 2.4000000000000003e-06, 'num_tokens': 5663145.0, 'completions/mean_length': 646.568359375, 'completions/min_length': 51.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.0240885429084301, 'completions/mean_terminated_length': 586.701171875, 'completions/min_terminated_length': 51.0, 'completions/max_terminated_length': 2579.0, 'rewards/reward_correctness/mean': 0.4537760317325592, 'rewards/reward_correctness/std': 0.4980209469795227, 'reward': 0.4537760615348816, 'reward_std': 0.49802088737487793, 'frac_reward_zero_std': 0.421875, 'sampling/sampling_logp_difference/mean': 0.023626327514648438, 'sampling/sampling_logp_difference/max': 1.7712206840515137, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5120000839233398, 'sampling/importance_sampling_ratio/max': 2.9854531288146973, 'entropy': 0.6028780657798052, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 278.0259652659297, 'epoch': 0.07}
4%|▎ | 5/136 [34:15<12:09:44, 334.23s/it][2026-08-12 23:04:20] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,2)
[2026-08-12 23:04:20] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,6)
[2026-08-12 23:04:20] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "frac" @(1,1)
Open LaTeX blocks:
@(1,1) arguments of macro "frac"
[2026-08-12 23:04:20] WARNING __init__.py:1389: WARNING: Error in configuration: macro '\frac' failed its substitution!
4%|▍ | 6/136 [38:50<11:20:55, 314.27s/it]
{'loss': -0.0204, 'grad_norm': 0.05935708343875522, 'learning_rate': 3e-06, 'num_tokens': 6771561.0, 'completions/mean_length': 635.78125, 'completions/min_length': 142.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.02669270895421505, 'completions/mean_terminated_length': 568.9685668945312, 'completions/min_terminated_length': 142.0, 'completions/max_terminated_length': 2117.0, 'rewards/reward_correctness/mean': 0.5423176884651184, 'rewards/reward_correctness/std': 0.4983682334423065, 'reward': 0.5423177480697632, 'reward_std': 0.49836820363998413, 'frac_reward_zero_std': 0.484375, 'sampling/sampling_logp_difference/mean': 0.023153454065322876, 'sampling/sampling_logp_difference/max': 2.0358147621154785, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5353550314903259, 'sampling/importance_sampling_ratio/max': 2.988870859146118, 'entropy': 0.5801784613868222, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 274.2749971947633, 'epoch': 0.09}
4%|▍ | 6/136 [38:50<11:20:55, 314.27s/it][2026-08-12 23:08:54] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,12)
[2026-08-12 23:08:54] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,33)
[2026-08-12 23:08:54] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,40)
5%|▌ | 7/136 [43:26<10:48:36, 301.67s/it]
{'loss': -0.0121, 'grad_norm': 0.06858649907867036, 'learning_rate': 2.9996118137817615e-06, 'num_tokens': 7907301.0, 'completions/mean_length': 651.171875, 'completions/min_length': 106.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.0325520858168602, 'completions/mean_terminated_length': 569.7173461914062, 'completions/min_terminated_length': 106.0, 'completions/max_terminated_length': 1737.0, 'rewards/reward_correctness/mean': 0.55078125, 'rewards/reward_correctness/std': 0.4975765645503998, 'reward': 0.55078125, 'reward_std': 0.4975765645503998, 'frac_reward_zero_std': 0.546875, 'sampling/sampling_logp_difference/mean': 0.023694580420851707, 'sampling/sampling_logp_difference/max': 1.0719003677368164, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5241525173187256, 'sampling/importance_sampling_ratio/max': 2.9824862480163574, 'entropy': 0.570356258423999, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 274.51085175946355, 'epoch': 0.1}
5%|▌ | 7/136 [43:26<10:48:36, 301.67s/it][2026-08-12 23:13:31] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,17)
[2026-08-12 23:13:31] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,197)
[2026-08-12 23:13:31] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "frac" @(1,1)
Open LaTeX blocks:
@(1,1) arguments of macro "frac"
[2026-08-12 23:13:31] WARNING __init__.py:1389: WARNING: Error in configuration: macro '\frac' failed its substitution!
[2026-08-12 23:13:31] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,15)
[2026-08-12 23:13:32] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,3)
[2026-08-12 23:13:32] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,18)
[2026-08-12 23:13:32] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,109)
[2026-08-12 23:13:32] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,185)
6%|▌ | 8/136 [47:56<10:22:18, 291.71s/it]
{'loss': -0.0272, 'grad_norm': 0.06117333934464337, 'learning_rate': 2.998447478369329e-06, 'num_tokens': 9139437.0, 'completions/mean_length': 716.3203125, 'completions/min_length': 166.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.0338541679084301, 'completions/mean_terminated_length': 633.7762451171875, 'completions/min_terminated_length': 166.0, 'completions/max_terminated_length': 2899.0, 'rewards/reward_correctness/mean': 0.4895833432674408, 'rewards/reward_correctness/std': 0.5000542402267456, 'reward': 0.4895833432674408, 'reward_std': 0.5000542402267456, 'frac_reward_zero_std': 0.453125, 'sampling/sampling_logp_difference/mean': 0.021436309441924095, 'sampling/sampling_logp_difference/max': 1.0037541389465332, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5255301594734192, 'sampling/importance_sampling_ratio/max': 2.9775948524475098, 'entropy': 0.5644197696819901, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 269.14979780837893, 'epoch': 0.12}
6%|▌ | 8/136 [47:56<10:22:18, 291.71s/it][2026-08-12 23:18:01] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,43)
[2026-08-12 23:18:01] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,24)
[2026-08-12 23:18:01] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,25)
[2026-08-12 23:18:01] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "frac" @(1,4)
Open LaTeX blocks:
@(1,4) arguments of macro "frac"
[2026-08-12 23:18:01] WARNING __init__.py:1389: WARNING: Error in configuration: macro '\frac' failed its substitution!
7%|▋ | 9/136 [52:35<10:08:51, 287.65s/it]
{'loss': -0.002, 'grad_norm': 0.05537902992630806, 'learning_rate': 2.9965076633611604e-06, 'num_tokens': 10282293.0, 'completions/mean_length': 656.640625, 'completions/min_length': 109.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.021484375, 'completions/mean_terminated_length': 603.6087646484375, 'completions/min_terminated_length': 109.0, 'completions/max_terminated_length': 2486.0, 'rewards/reward_correctness/mean': 0.572265625, 'rewards/reward_correctness/std': 0.494911253452301, 'reward': 0.572265625, 'reward_std': 0.49491122364997864, 'frac_reward_zero_std': 0.53125, 'sampling/sampling_logp_difference/mean': 0.02303769439458847, 'sampling/sampling_logp_difference/max': 0.994542121887207, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5056574940681458, 'sampling/importance_sampling_ratio/max': 2.991780996322632, 'entropy': 0.5387037470936775, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 277.4853810737841, 'epoch': 0.13}
7%|▋ | 9/136 [52:35<10:08:51, 287.65s/it][2026-08-12 23:22:37] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,83)
[2026-08-12 23:22:38] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,122)
7%|▋ | 10/136 [57:13<9:57:39, 284.60s/it]
{'loss': -0.0034, 'grad_norm': 0.05498481378949289, 'learning_rate': 2.993793484326816e-06, 'num_tokens': 11421770.0, 'completions/mean_length': 643.1126708984375, 'completions/min_length': 155.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.02083333395421505, 'completions/mean_terminated_length': 591.442138671875, 'completions/min_terminated_length': 155.0, 'completions/max_terminated_length': 3008.0, 'rewards/reward_correctness/mean': 0.5618489384651184, 'rewards/reward_correctness/std': 0.49632152915000916, 'reward': 0.5618489980697632, 'reward_std': 0.49632149934768677, 'frac_reward_zero_std': 0.4765625, 'sampling/sampling_logp_difference/mean': 0.0213819220662117, 'sampling/sampling_logp_difference/max': 0.9893784523010254, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5399176478385925, 'sampling/importance_sampling_ratio/max': 2.9866268634796143, 'entropy': 0.47972284245770425, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 276.4947049263865, 'epoch': 0.15}
7%|▋ | 10/136 [57:13<9:57:39, 284.60s/it]
0%| | 0/84 [00:00<?, ?it/s]
2%|▏ | 2/84 [00:07<04:53, 3.58s/it]
4%|▎ | 3/84 [00:12<05:59, 4.43s/it]
5%|▍ | 4/84 [00:19<07:15, 5.45s/it]
6%|▌ | 5/84 [00:27<08:02, 6.11s/it]
7%|▋ | 6/84 [00:34<08:31, 6.56s/it]
8%|▊ | 7/84 [00:43<09:11, 7.16s/it]
10%|▉ | 8/84 [00:45<07:15, 5.73s/it]
11%|█ | 9/84 [00:58<09:56, 7.96s/it]
12%|█▏ | 10/84 [01:04<08:58, 7.28s/it]
13%|█▎ | 11/84 [01:10<08:21, 6.88s/it]
14%|█▍ | 12/84 [01:16<08:00, 6.67s/it]
15%|█▌ | 13/84 [01:24<08:17, 7.01s/it]
17%|█▋ | 14/84 [01:30<07:55, 6.80s/it]
18%|█▊ | 15/84 [01:41<09:16, 8.06s/it]
19%|█▉ | 16/84 [01:48<08:38, 7.63s/it]
20%|██ | 17/84 [01:55<08:19, 7.45s/it]
21%|██▏ | 18/84 [02:00<07:34, 6.88s/it]
23%|██▎ | 19/84 [02:08<07:34, 6.99s/it]
24%|██▍ | 20/84 [02:15<07:26, 6.97s/it]
25%|██▌ | 21/84 [02:21<06:59, 6.67s/it]
26%|██▌ | 22/84 [02:25<06:08, 5.95s/it]
27%|██▋ | 23/84 [02:31<06:01, 5.93s/it]
29%|██▊ | 24/84 [02:37<05:54, 5.91s/it]
30%|██▉ | 25/84 [02:41<05:21, 5.45s/it]
31%|███ | 26/84 [02:45<04:59, 5.16s/it]
32%|███▏ | 27/84 [02:52<05:10, 5.44s/it]
33%|███▎ | 28/84 [03:15<10:10, 10.90s/it]
35%|███▍ | 29/84 [03:21<08:36, 9.40s/it]
36%|███▌ | 30/84 [03:28<07:41, 8.55s/it]
37%|███▋ | 31/84 [03:34<07:03, 7.99s/it]
38%|███▊ | 32/84 [03:40<06:26, 7.44s/it]
39%|███▉ | 33/84 [03:48<06:18, 7.42s/it]
40%|████ | 34/84 [03:56<06:17, 7.56s/it]
42%|████▏ | 35/84 [04:04<06:14, 7.63s/it]
43%|████▎ | 36/84 [04:10<05:53, 7.36s/it]
44%|████▍ | 37/84 [04:14<04:55, 6.29s/it]
45%|████▌ | 38/84 [04:20<04:39, 6.07s/it]
46%|████▋ | 39/84 [04:29<05:16, 7.03s/it]
48%|████▊ | 40/84 [04:53<08:51, 12.09s/it]
49%|████▉ | 41/84 [04:58<07:07, 9.94s/it]
50%|█████ | 42/84 [05:21<09:48, 14.02s/it]
51%|█████ | 43/84 [05:29<08:12, 12.02s/it]
52%|█████▏ | 44/84 [05:33<06:23, 9.60s/it]
54%|█████▎ | 45/84 [05:39<05:35, 8.61s/it]
55%|█████▍ | 46/84 [05:45<05:02, 7.96s/it]
56%|█████▌ | 47/84 [05:52<04:36, 7.49s/it]
57%|█████▋ | 48/84 [05:59<04:30, 7.52s/it]
58%|█████▊ | 49/84 [06:06<04:14, 7.27s/it]
60%|█████▉ | 50/84 [06:13<04:04, 7.18s/it]
61%|██████ | 51/84 [06:22<04:11, 7.62s/it]
62%|██████▏ | 52/84 [06:30<04:14, 7.95s/it]
63%|██████▎ | 53/84 [06:35<03:40, 7.11s/it]
64%|██████▍ | 54/84 [06:40<03:13, 6.44s/it]
65%|██████▌ | 55/84 [06:53<04:03, 8.38s/it]
67%|██████▋ | 56/84 [06:56<03:11, 6.83s/it]
68%|██████▊ | 57/84 [07:03<03:02, 6.76s/it]
69%|██████▉ | 58/84 [07:10<02:55, 6.75s/it]
70%|███████ | 59/84 [07:19<03:04, 7.37s/it]
71%|███████▏ | 60/84 [07:24<02:40, 6.67s/it]
73%|███████▎ | 61/84 [07:31<02:38, 6.89s/it]
74%|███████▍ | 62/84 [07:37<02:27, 6.71s/it]
75%|███████▌ | 63/84 [07:47<02:41, 7.67s/it]
76%|███████▌ | 64/84 [08:11<04:11, 12.59s/it]
77%|███████▋ | 65/84 [08:19<03:33, 11.23s/it]
79%|███████▊ | 66/84 [08:25<02:51, 9.52s/it]
80%|███████▉ | 67/84 [08:33<02:35, 9.17s/it]
81%|████████ | 68/84 [08:42<02:25, 9.06s/it]
82%|████████▏ | 69/84 [08:49<02:08, 8.55s/it]
83%|████████▎ | 70/84 [08:57<01:56, 8.30s/it]
85%|████████▍ | 71/84 [09:01<01:31, 7.00s/it]
86%|████████▌ | 72/84 [09:08<01:24, 7.01s/it]
87%|████████▋ | 73/84 [09:15<01:17, 7.02s/it]
88%|████████▊ | 74/84 [09:22<01:09, 7.00s/it]
89%|████████▉ | 75/84 [09:31<01:07, 7.49s/it]
90%|█████████ | 76/84 [09:39<01:01, 7.69s/it]
92%|█████████▏| 77/84 [09:44<00:48, 6.92s/it]
93%|█████████▎| 78/84 [09:53<00:45, 7.62s/it]
94%|█████████▍| 79/84 [09:59<00:35, 7.05s/it]
95%|█████████▌| 80/84 [10:05<00:27, 6.87s/it]
96%|█████████▋| 81/84 [10:12<00:20, 6.85s/it]
98%|█████████▊| 82/84 [10:19<00:13, 6.80s/it]
99%|█████████▉| 83/84 [10:25<00:06, 6.67s/it]
100%|██████████| 84/84 [10:31<00:00, 6.37s/it]
{'eval_loss': 0.0, 'eval_runtime': 638.6604, 'eval_samples_per_second': 0.783, 'eval_steps_per_second': 0.132, 'eval_num_tokens': 11421770.0, 'eval_completions/mean_length': 576.2758113316128, 'eval_completions/min_length': 281.51190476190476, 'eval_completions/max_length': 1043.8095238095239, 'eval_completions/clipped_ratio': 0.007936508173034304, 'eval_completions/mean_terminated_length': 556.6127137683686, 'eval_completions/min_terminated_length': 281.51190476190476, 'eval_completions/max_terminated_length': 943.4166666666666, 'eval_rewards/reward_correctness/mean': 0.6884920650294849, 'eval_rewards/reward_correctness/std': 0.4230475830180304, 'eval_reward': 0.6884920799306461, 'eval_reward_std': 0.42304758372760953, 'eval_frac_reward_zero_std': 1.0, 'eval_sampling/sampling_logp_difference/mean': 0.059989471831137224, 'eval_sampling/sampling_logp_difference/max': 2.7100065719513666, 'eval_sampling/importance_sampling_ratio/min': 3.407798286771119e-06, 'eval_sampling/importance_sampling_ratio/mean': 0.03580059619768522, 'eval_sampling/importance_sampling_ratio/max': 0.19307976385317444, 'eval_entropy': 0.2537455828416915, 'eval_clip_ratio/low_mean': 0.0, 'eval_clip_ratio/low_min': 0.0, 'eval_clip_ratio/high_mean': 0.0, 'eval_clip_ratio/high_max': 0.0, 'eval_clip_ratio/region_mean': 0.0, 'epoch': 0.15}
7%|▋ | 10/136 [1:07:52<9:57:39, 284.60s/it]
100%|██████████| 84/84 [10:31<00:00, 6.37s/it]

8%|▊ | 11/136 [1:12:20<16:29:44, 475.07s/it]
{'loss': -0.0187, 'grad_norm': 0.057721741043468014, 'learning_rate': 2.990306502165398e-06, 'num_tokens': 12545459.0, 'completions/mean_length': 639.107421875, 'completions/min_length': 136.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.02473958395421505, 'completions/mean_terminated_length': 577.3919067382812, 'completions/min_terminated_length': 136.0, 'completions/max_terminated_length': 2715.0, 'rewards/reward_correctness/mean': 0.57421875, 'rewards/reward_correctness/std': 0.4946219325065613, 'reward': 0.57421875, 'reward_std': 0.4946219027042389, 'frac_reward_zero_std': 0.421875, 'sampling/sampling_logp_difference/mean': 0.02195746824145317, 'sampling/sampling_logp_difference/max': 0.9523108005523682, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5547703504562378, 'sampling/importance_sampling_ratio/max': 2.979454755783081, 'entropy': 0.5096990133170038, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 258.60902352258563, 'epoch': 0.16}
8%|▊ | 11/136 [1:12:20<16:29:44, 475.07s/it][2026-08-12 23:42:22] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,34)
9%|▉ | 12/136 [1:16:49<14:12:12, 412.36s/it]
{'loss': -0.0058, 'grad_norm': 0.05418244520480129, 'learning_rate': 2.986048722207899e-06, 'num_tokens': 13729294.0, 'completions/mean_length': 672.0462646484375, 'completions/min_length': 133.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.0299479179084301, 'completions/mean_terminated_length': 597.9537353515625, 'completions/min_terminated_length': 133.0, 'completions/max_terminated_length': 2413.0, 'rewards/reward_correctness/mean': 0.5579426884651184, 'rewards/reward_correctness/std': 0.4967930316925049, 'reward': 0.5579427480697632, 'reward_std': 0.4967930018901825, 'frac_reward_zero_std': 0.5, 'sampling/sampling_logp_difference/mean': 0.02223217487335205, 'sampling/sampling_logp_difference/max': 1.4261195659637451, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5251661539077759, 'sampling/importance_sampling_ratio/max': 2.9794344902038574, 'entropy': 0.5439460552297533, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 267.7675925884396, 'epoch': 0.18}
9%|▉ | 12/136 [1:16:49<14:12:12, 412.36s/it][2026-08-12 23:47:04] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,7)
[2026-08-12 23:47:05] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,3)
[2026-08-12 23:47:05] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "frac" @(1,1)
Open LaTeX blocks:
@(1,1) arguments of macro "frac"
[2026-08-12 23:47:05] WARNING __init__.py:1389: WARNING: Error in configuration: macro '\frac' failed its substitution!
[2026-08-12 23:47:05] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,5)
10%|▉ | 13/136 [1:21:49<12:55:27, 378.27s/it]
{'loss': 0.0052, 'grad_norm': 0.05705325757219269, 'learning_rate': 2.981022593063946e-06, 'num_tokens': 14916792.0, 'completions/mean_length': 665.5872802734375, 'completions/min_length': 139.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.025390625, 'completions/mean_terminated_length': 603.4348754882812, 'completions/min_terminated_length': 139.0, 'completions/max_terminated_length': 2197.0, 'rewards/reward_correctness/mean': 0.53515625, 'rewards/reward_correctness/std': 0.49892494082450867, 'reward': 0.53515625, 'reward_std': 0.4989249110221863, 'frac_reward_zero_std': 0.46875, 'sampling/sampling_logp_difference/mean': 0.02207307144999504, 'sampling/sampling_logp_difference/max': 2.7724132537841797, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5438298583030701, 'sampling/importance_sampling_ratio/max': 2.997776746749878, 'entropy': 0.5305114695802331, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 298.6625449950807, 'epoch': 0.19}
10%|▉ | 13/136 [1:21:49<12:55:27, 378.27s/it][2026-08-12 23:51:54] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,0)
[2026-08-12 23:51:55] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,1)
[2026-08-12 23:51:55] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Bad \end macro: expected {environmentname} @(1,26)
10%|█ | 14/136 [1:26:25<11:46:26, 347.43s/it]
{'loss': 0.0141, 'grad_norm': 0.058814816392006475, 'learning_rate': 2.9752310052136353e-06, 'num_tokens': 16113926.0, 'completions/mean_length': 690.9935302734375, 'completions/min_length': 166.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.0279947929084301, 'completions/mean_terminated_length': 622.41796875, 'completions/min_terminated_length': 166.0, 'completions/max_terminated_length': 2265.0, 'rewards/reward_correctness/mean': 0.5716145634651184, 'rewards/reward_correctness/std': 0.4950059652328491, 'reward': 0.5716146230697632, 'reward_std': 0.49500590562820435, 'frac_reward_zero_std': 0.4375, 'sampling/sampling_logp_difference/mean': 0.02060725726187229, 'sampling/sampling_logp_difference/max': 1.378803014755249, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5442800521850586, 'sampling/importance_sampling_ratio/max': 2.9824302196502686, 'entropy': 0.49725982279051095, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 274.99928429629654, 'epoch': 0.21}
10%|█ | 14/136 [1:26:25<11:46:26, 347.43s/it][2026-08-12 23:56:34] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,18)
[2026-08-12 23:56:34] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro """ @(1,19)
Open LaTeX blocks:
@(1,19) arguments of macro """
[2026-08-12 23:56:35] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,2)
11%|█ | 15/136 [1:31:24<11:11:10, 332.81s/it]
{'loss': 0.0002, 'grad_norm': 0.05525897475793538, 'learning_rate': 2.968677289345242e-06, 'num_tokens': 17267821.0, 'completions/mean_length': 652.2025146484375, 'completions/min_length': 161.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.029296875, 'completions/mean_terminated_length': 579.4735107421875, 'completions/min_terminated_length': 161.0, 'completions/max_terminated_length': 1901.0, 'rewards/reward_correctness/mean': 0.5826823115348816, 'rewards/reward_correctness/std': 0.49327683448791504, 'reward': 0.5826823115348816, 'reward_std': 0.49327683448791504, 'frac_reward_zero_std': 0.5, 'sampling/sampling_logp_difference/mean': 0.02108415775001049, 'sampling/sampling_logp_difference/max': 1.0892324447631836, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5603505969047546, 'sampling/importance_sampling_ratio/max': 2.9618661403656006, 'entropy': 0.5102907492546365, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 297.76528844004497, 'epoch': 0.22}
11%|█ | 15/136 [1:31:24<11:11:10, 332.81s/it][2026-08-13 00:01:31] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,20)
[2026-08-13 00:01:31] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\)' @(1,167)
[2026-08-13 00:01:32] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,14)
[2026-08-13 00:01:32] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,15)
12%|█▏ | 16/136 [1:36:10<10:37:28, 318.74s/it]
{'loss': -0.0028, 'grad_norm': 0.05610150359953727, 'learning_rate': 2.9613652144397706e-06, 'num_tokens': 18491496.0, 'completions/mean_length': 692.1009521484375, 'completions/min_length': 113.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.0299479179084301, 'completions/mean_terminated_length': 618.6275634765625, 'completions/min_terminated_length': 113.0, 'completions/max_terminated_length': 2084.0, 'rewards/reward_correctness/mean': 0.5182291865348816, 'rewards/reward_correctness/std': 0.49983033537864685, 'reward': 0.5182291865348816, 'reward_std': 0.49983030557632446, 'frac_reward_zero_std': 0.46875, 'sampling/sampling_logp_difference/mean': 0.02187207341194153, 'sampling/sampling_logp_difference/max': 0.979703426361084, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5334643721580505, 'sampling/importance_sampling_ratio/max': 2.9970788955688477, 'entropy': 0.5383049458032474, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 284.81178701808676, 'epoch': 0.24}
12%|█▏ | 16/136 [1:36:10<10:37:28, 318.74s/it][2026-08-13 00:06:20] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "frac" @(1,1)
Open LaTeX blocks:
@(1,1) arguments of macro "frac"
[2026-08-13 00:06:20] WARNING __init__.py:1389: WARNING: Error in configuration: macro '\frac' failed its substitution!
[2026-08-13 00:06:20] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "frac" @(1,2)
Open LaTeX blocks:
@(1,2) arguments of macro "frac"
[2026-08-13 00:06:20] WARNING __init__.py:1389: WARNING: Error in configuration: macro '\frac' failed its substitution!
12%|█▎ | 17/136 [1:41:03<10:16:53, 311.03s/it]
{'loss': -0.0051, 'grad_norm': 0.04720701387579848, 'learning_rate': 2.9532989856034515e-06, 'num_tokens': 19722251.0, 'completions/mean_length': 712.1243896484375, 'completions/min_length': 153.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.0338541679084301, 'completions/mean_terminated_length': 629.4332885742188, 'completions/min_terminated_length': 153.0, 'completions/max_terminated_length': 2957.0, 'rewards/reward_correctness/mean': 0.556640625, 'rewards/reward_correctness/std': 0.49694326519966125, 'reward': 0.556640625, 'reward_std': 0.49694323539733887, 'frac_reward_zero_std': 0.5, 'sampling/sampling_logp_difference/mean': 0.022461896762251854, 'sampling/sampling_logp_difference/max': 0.9831252098083496, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5143538117408752, 'sampling/importance_sampling_ratio/max': 2.994859457015991, 'entropy': 0.5025974611053243, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 291.982781663537, 'epoch': 0.25}
12%|█▎ | 17/136 [1:41:03<10:16:53, 311.03s/it][2026-08-13 00:11:11] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,18)
[2026-08-13 00:11:11] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,14)
[2026-08-13 00:11:11] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,15)
13%|█▎ | 18/136 [1:45:47<9:55:45, 302.93s/it]
{'loss': -0.0191, 'grad_norm': 0.05438412892841907, 'learning_rate': 2.94448324164942e-06, 'num_tokens': 20875235.0, 'completions/mean_length': 651.921875, 'completions/min_length': 120.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.021484375, 'completions/mean_terminated_length': 598.7864379882812, 'completions/min_terminated_length': 120.0, 'completions/max_terminated_length': 2772.0, 'rewards/reward_correctness/mean': 0.580078125, 'rewards/reward_correctness/std': 0.4937065839767456, 'reward': 0.580078125, 'reward_std': 0.49370652437210083, 'frac_reward_zero_std': 0.421875, 'sampling/sampling_logp_difference/mean': 0.019891459494829178, 'sampling/sampling_logp_difference/max': 2.025559663772583, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5411784052848816, 'sampling/importance_sampling_ratio/max': 2.980726718902588, 'entropy': 0.4585979155963287, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 282.89908104995266, 'epoch': 0.26}
13%|█▎ | 18/136 [1:45:47<9:55:45, 302.93s/it]
14%|█▍ | 19/136 [1:50:31<9:39:52, 297.37s/it]
{'loss': -0.0131, 'grad_norm': 0.04706840896526157, 'learning_rate': 2.934923052429984e-06, 'num_tokens': 22092475.0, 'completions/mean_length': 699.4583740234375, 'completions/min_length': 172.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.0416666679084301, 'completions/mean_terminated_length': 596.3043823242188, 'completions/min_terminated_length': 172.0, 'completions/max_terminated_length': 2449.0, 'rewards/reward_correctness/mean': 0.59765625, 'rewards/reward_correctness/std': 0.49053022265434265, 'reward': 0.59765625, 'reward_std': 0.49053022265434265, 'frac_reward_zero_std': 0.46875, 'sampling/sampling_logp_difference/mean': 0.02276076376438141, 'sampling/sampling_logp_difference/max': 1.5002199411392212, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5492980480194092, 'sampling/importance_sampling_ratio/max': 2.9970383644104004, 'entropy': 0.5197840981418267, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 283.09366952162236, 'epoch': 0.28}
14%|█▍ | 19/136 [1:50:31<9:39:52, 297.37s/it][2026-08-13 00:20:39] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,168)
15%|█▍ | 20/136 [1:55:10<9:23:49, 291.63s/it]
{'loss': 0.0048, 'grad_norm': 0.05405444364791074, 'learning_rate': 2.924623915920992e-06, 'num_tokens': 23292318.0, 'completions/mean_length': 695.2103271484375, 'completions/min_length': 164.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.041015625, 'completions/mean_terminated_length': 593.5552978515625, 'completions/min_terminated_length': 164.0, 'completions/max_terminated_length': 1904.0, 'rewards/reward_correctness/mean': 0.5670573115348816, 'rewards/reward_correctness/std': 0.4956442713737488, 'reward': 0.5670573115348816, 'reward_std': 0.4956442713737488, 'frac_reward_zero_std': 0.4140625, 'sampling/sampling_logp_difference/mean': 0.023046687245368958, 'sampling/sampling_logp_difference/max': 1.4154164791107178, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5073182582855225, 'sampling/importance_sampling_ratio/max': 2.979520320892334, 'entropy': 0.5042589120566845, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 277.07896974822506, 'epoch': 0.29}
15%|█▍ | 20/136 [1:55:10<9:23:49, 291.63s/it]
0%| | 0/84 [00:00<?, ?it/s]
2%|▏ | 2/84 [00:06<04:20, 3.18s/it]
4%|▎ | 3/84 [00:10<05:00, 3.70s/it]
5%|▍ | 4/84 [00:18<06:49, 5.12s/it]
6%|▌ | 5/84 [00:24<07:27, 5.66s/it]
7%|▋ | 6/84 [00:30<07:12, 5.54s/it]
8%|▊ | 7/84 [00:39<08:48, 6.86s/it]
10%|▉ | 8/84 [00:42<07:09, 5.65s/it]
11%|█ | 9/84 [00:50<07:51, 6.29s/it]
12%|█▏ | 10/84 [00:56<07:27, 6.05s/it]
13%|█▎ | 11/84 [01:04<08:19, 6.85s/it]
14%|█▍ | 12/84 [01:10<07:56, 6.61s/it]
15%|█▌ | 13/84 [01:17<07:46, 6.58s/it]
17%|█▋ | 14/84 [01:23<07:39, 6.56s/it]
18%|█▊ | 15/84 [01:32<08:17, 7.21s/it]
19%|█▉ | 16/84 [01:36<07:06, 6.27s/it]
20%|██ | 17/84 [01:44<07:38, 6.85s/it]
21%|██▏ | 18/84 [01:49<06:50, 6.22s/it]
23%|██▎ | 19/84 [01:56<06:48, 6.28s/it]
24%|██▍ | 20/84 [02:03<07:12, 6.75s/it]
25%|██▌ | 21/84 [02:10<06:57, 6.63s/it]
26%|██▌ | 22/84 [02:14<06:01, 5.83s/it]
27%|██▋ | 23/84 [02:18<05:25, 5.34s/it]
29%|██▊ | 24/84 [02:25<05:55, 5.93s/it]
30%|██▉ | 25/84 [02:29<05:18, 5.40s/it]
31%|███ | 26/84 [02:36<05:40, 5.87s/it]
32%|███▏ | 27/84 [02:44<06:00, 6.32s/it]
33%|███▎ | 28/84 [02:49<05:31, 5.93s/it]
35%|███▍ | 29/84 [02:55<05:28, 5.98s/it]
36%|███▌ | 30/84 [03:02<05:47, 6.43s/it]
37%|███▋ | 31/84 [03:09<05:49, 6.60s/it]
38%|███▊ | 32/84 [03:14<05:09, 5.96s/it]
39%|███▉ | 33/84 [03:22<05:38, 6.63s/it]
40%|████ | 34/84 [03:30<05:51, 7.04s/it]
42%|████▏ | 35/84 [03:54<09:47, 11.98s/it]
43%|████▎ | 36/84 [03:59<07:58, 9.97s/it]
44%|████▍ | 37/84 [04:03<06:34, 8.39s/it]
45%|████▌ | 38/84 [04:14<06:58, 9.10s/it]
46%|████▋ | 39/84 [04:23<06:45, 9.01s/it]
48%|████▊ | 40/84 [04:47<09:52, 13.46s/it]
49%|████▉ | 41/84 [04:56<08:41, 12.12s/it]
50%|█████ | 42/84 [05:05<07:47, 11.13s/it]
51%|█████ | 43/84 [05:10<06:29, 9.49s/it]
52%|█████▏ | 44/84 [05:15<05:19, 7.98s/it]
54%|█████▎ | 45/84 [05:21<04:55, 7.58s/it]
55%|█████▍ | 46/84 [05:29<04:42, 7.44s/it]
56%|█████▌ | 47/84 [05:34<04:07, 6.69s/it]
57%|█████▋ | 48/84 [05:41<04:09, 6.93s/it]
58%|█████▊ | 49/84 [05:50<04:25, 7.60s/it]
60%|█████▉ | 50/84 [05:59<04:28, 7.90s/it]
61%|██████ | 51/84 [06:06<04:13, 7.67s/it]
62%|██████▏ | 52/84 [06:13<04:00, 7.50s/it]
63%|██████▎ | 53/84 [06:17<03:19, 6.44s/it]
64%|██████▍ | 54/84 [06:21<02:50, 5.69s/it]
65%|██████▌ | 55/84 [06:32<03:31, 7.28s/it]
67%|██████▋ | 56/84 [06:35<02:48, 6.03s/it]
68%|██████▊ | 57/84 [06:44<03:04, 6.83s/it]
69%|██████▉ | 58/84 [06:49<02:48, 6.50s/it]
70%|███████ | 59/84 [07:05<03:47, 9.11s/it]
71%|███████▏ | 60/84 [07:11<03:16, 8.19s/it]
73%|███████▎ | 61/84 [07:16<02:49, 7.37s/it]
74%|███████▍ | 62/84 [07:22<02:34, 7.02s/it]
75%|███████▌ | 63/84 [07:32<02:44, 7.82s/it]
76%|███████▌ | 64/84 [07:39<02:28, 7.42s/it]
77%|███████▋ | 65/84 [07:44<02:07, 6.72s/it]
79%|███████▊ | 66/84 [07:48<01:45, 5.87s/it]
80%|███████▉ | 67/84 [07:54<01:40, 5.91s/it]
81%|████████ | 68/84 [08:04<01:54, 7.17s/it]
82%|████████▏ | 69/84 [08:10<01:45, 7.02s/it]
83%|████████▎ | 70/84 [08:17<01:38, 7.02s/it]
85%|████████▍ | 71/84 [08:21<01:18, 6.06s/it]
86%|████████▌ | 72/84 [08:28<01:14, 6.23s/it]
87%|████████▋ | 73/84 [08:36<01:13, 6.69s/it]
88%|████████▊ | 74/84 [08:42<01:07, 6.77s/it]
89%|████████▉ | 75/84 [08:48<00:57, 6.34s/it]
90%|█████████ | 76/84 [08:56<00:55, 6.92s/it]
92%|█████████▏| 77/84 [09:02<00:45, 6.57s/it]
93%|█████████▎| 78/84 [09:08<00:38, 6.48s/it]
94%|█████████▍| 79/84 [09:14<00:31, 6.23s/it]
95%|█████████▌| 80/84 [09:20<00:24, 6.11s/it]
96%|█████████▋| 81/84 [09:27<00:19, 6.64s/it]
98%|█████████▊| 82/84 [09:38<00:15, 7.76s/it]
99%|█████████▉| 83/84 [09:43<00:07, 7.09s/it]
100%|██████████| 84/84 [09:48<00:00, 6.45s/it]
{'eval_loss': 0.0, 'eval_runtime': 593.7444, 'eval_samples_per_second': 0.842, 'eval_steps_per_second': 0.141, 'eval_num_tokens': 23292318.0, 'eval_completions/mean_length': 563.2658905755906, 'eval_completions/min_length': 285.7857142857143, 'eval_completions/max_length': 973.297619047619, 'eval_completions/clipped_ratio': 0.003968254086517152, 'eval_completions/mean_terminated_length': 554.1896990821475, 'eval_completions/min_terminated_length': 285.7857142857143, 'eval_completions/max_terminated_length': 935.7857142857143, 'eval_rewards/reward_correctness/mean': 0.6884920646746954, 'eval_rewards/reward_correctness/std': 0.4417421498468944, 'eval_reward': 0.6884920824141729, 'eval_reward_std': 0.44174214807294665, 'eval_frac_reward_zero_std': 1.0, 'eval_sampling/sampling_logp_difference/mean': 0.05756539217240754, 'eval_sampling/sampling_logp_difference/max': 2.7017440568833124, 'eval_sampling/importance_sampling_ratio/min': 8.914577132557873e-06, 'eval_sampling/importance_sampling_ratio/mean': 0.04097350465803001, 'eval_sampling/importance_sampling_ratio/max': 0.21161528248709752, 'eval_entropy': 0.24252484188902945, 'eval_clip_ratio/low_mean': 0.0, 'eval_clip_ratio/low_min': 0.0, 'eval_clip_ratio/high_mean': 0.0, 'eval_clip_ratio/high_max': 0.0, 'eval_clip_ratio/region_mean': 0.0, 'epoch': 0.29}
15%|█▍ | 20/136 [2:05:03<9:23:49, 291.63s/it]
100%|██████████| 84/84 [09:48<00:00, 6.45s/it]

15%|█▌ | 21/136 [2:09:59<15:02:59, 471.13s/it]
{'loss': -0.0281, 'grad_norm': 0.03804389825609189, 'learning_rate': 2.913591755060004e-06, 'num_tokens': 24588862.0, 'completions/mean_length': 745.3932495117188, 'completions/min_length': 168.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.0436197929084301, 'completions/mean_terminated_length': 639.3792114257812, 'completions/min_terminated_length': 168.0, 'completions/max_terminated_length': 1927.0, 'rewards/reward_correctness/mean': 0.544921875, 'rewards/reward_correctness/std': 0.4981401264667511, 'reward': 0.544921875, 'reward_std': 0.4981400966644287, 'frac_reward_zero_std': 0.515625, 'sampling/sampling_logp_difference/mean': 0.02183881774544716, 'sampling/sampling_logp_difference/max': 1.042482852935791, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5209087133407593, 'sampling/importance_sampling_ratio/max': 2.961940288543701, 'entropy': 0.5028883977793157, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 289.08046070858836, 'epoch': 0.31}
15%|█▌ | 21/136 [2:09:59<15:02:59, 471.13s/it][2026-08-13 00:40:09] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "frac" @(1,1)
Open LaTeX blocks:
@(1,1) arguments of macro "frac"
[2026-08-13 00:40:09] WARNING __init__.py:1389: WARNING: Error in configuration: macro '\frac' failed its substitution!
[2026-08-13 00:40:09] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,24)
16%|█▌ | 22/136 [2:14:40<13:06:49, 414.12s/it]
{'loss': -0.0177, 'grad_norm': 0.050298358349411876, 'learning_rate': 2.901832914340062e-06, 'num_tokens': 25890189.0, 'completions/mean_length': 757.6243896484375, 'completions/min_length': 177.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.0423177108168602, 'completions/mean_terminated_length': 655.3576049804688, 'completions/min_terminated_length': 177.0, 'completions/max_terminated_length': 2720.0, 'rewards/reward_correctness/mean': 0.53125, 'rewards/reward_correctness/std': 0.4991849958896637, 'reward': 0.53125, 'reward_std': 0.4991849958896637, 'frac_reward_zero_std': 0.421875, 'sampling/sampling_logp_difference/mean': 0.022565271705389023, 'sampling/sampling_logp_difference/max': 1.7315807342529297, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5123823881149292, 'sampling/importance_sampling_ratio/max': 2.954637289047241, 'entropy': 0.5200647414894775, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 280.0331052900292, 'epoch': 0.32}
16%|█▌ | 22/136 [2:14:40<13:06:49, 414.12s/it][2026-08-13 00:44:52] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,93)
17%|█▋ | 23/136 [2:19:35<11:52:35, 378.37s/it]
{'loss': -0.0047, 'grad_norm': 0.053729117053239864, 'learning_rate': 2.889354156161033e-06, 'num_tokens': 27140520.0, 'completions/mean_length': 717.806640625, 'completions/min_length': 198.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.0423177108168602, 'completions/mean_terminated_length': 614.0183715820312, 'completions/min_terminated_length': 198.0, 'completions/max_terminated_length': 2145.0, 'rewards/reward_correctness/mean': 0.591796875, 'rewards/reward_correctness/std': 0.4916611909866333, 'reward': 0.591796875, 'reward_std': 0.4916611313819885, 'frac_reward_zero_std': 0.515625, 'sampling/sampling_logp_difference/mean': 0.020792663097381592, 'sampling/sampling_logp_difference/max': 1.006546974182129, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5710696578025818, 'sampling/importance_sampling_ratio/max': 2.9909863471984863, 'entropy': 0.472823026124388, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 293.80172044876963, 'epoch': 0.34}
17%|█▋ | 23/136 [2:19:35<11:52:35, 378.37s/it][2026-08-13 00:49:45] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,3)
[2026-08-13 00:49:45] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Bad \end macro: expected {environmentname} @(1,34)
[2026-08-13 00:49:45] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,9)
18%|█▊ | 24/136 [2:24:33<11:01:07, 354.18s/it]
{'loss': 0.0024, 'grad_norm': 0.047428014800435944, 'learning_rate': 2.876162656940614e-06, 'num_tokens': 28369072.0, 'completions/mean_length': 702.1041870117188, 'completions/min_length': 176.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.0494791679084301, 'completions/mean_terminated_length': 578.73974609375, 'completions/min_terminated_length': 176.0, 'completions/max_terminated_length': 2233.0, 'rewards/reward_correctness/mean': 0.5735676884651184, 'rewards/reward_correctness/std': 0.4947192370891571, 'reward': 0.5735677480697632, 'reward_std': 0.4947192072868347, 'frac_reward_zero_std': 0.5078125, 'sampling/sampling_logp_difference/mean': 0.024461492896080017, 'sampling/sampling_logp_difference/max': 0.9955153465270996, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5351980328559875, 'sampling/importance_sampling_ratio/max': 2.939020872116089, 'entropy': 0.5469449365045875, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 296.57605882082134, 'epoch': 0.35}
18%|█▊ | 24/136 [2:24:33<11:01:07, 354.18s/it][2026-08-13 00:54:43] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,1)
[2026-08-13 00:54:43] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Bad \end macro: expected {environmentname} @(1,127)
[2026-08-13 00:54:43] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,5803)
18%|█▊ | 25/136 [2:29:20<10:18:00, 334.06s/it]
{'loss': -0.0119, 'grad_norm': 0.062490854299157274, 'learning_rate': 2.862266002987244e-06, 'num_tokens': 29582571.0, 'completions/mean_length': 700.2181396484375, 'completions/min_length': 166.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.041015625, 'completions/mean_terminated_length': 598.7772827148438, 'completions/min_terminated_length': 166.0, 'completions/max_terminated_length': 2020.0, 'rewards/reward_correctness/mean': 0.611328125, 'rewards/reward_correctness/std': 0.48760727047920227, 'reward': 0.611328125, 'reward_std': 0.4876072406768799, 'frac_reward_zero_std': 0.3984375, 'sampling/sampling_logp_difference/mean': 0.023312201723456383, 'sampling/sampling_logp_difference/max': 0.9733572006225586, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5363779067993164, 'sampling/importance_sampling_ratio/max': 2.9679155349731445, 'entropy': 0.4677532056812197, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 285.94121135165915, 'epoch': 0.37}
18%|█▊ | 25/136 [2:29:20<10:18:00, 334.06s/it][2026-08-13 00:59:38] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,91)
[2026-08-13 00:59:38] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,92)
[2026-08-13 00:59:38] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,103)
19%|█▉ | 26/136 [2:34:15<9:50:38, 322.17s/it]
{'loss': -0.0108, 'grad_norm': 0.043192270355031, 'learning_rate': 2.847672186137282e-06, 'num_tokens': 30924112.0, 'completions/mean_length': 772.5709838867188, 'completions/min_length': 123.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.05859375, 'completions/mean_terminated_length': 629.4530029296875, 'completions/min_terminated_length': 123.0, 'completions/max_terminated_length': 2558.0, 'rewards/reward_correctness/mean': 0.5305989384651184, 'rewards/reward_correctness/std': 0.499225378036499, 'reward': 0.5305989980697632, 'reward_std': 0.49922531843185425, 'frac_reward_zero_std': 0.5390625, 'sampling/sampling_logp_difference/mean': 0.0235705878585577, 'sampling/sampling_logp_difference/max': 0.9405348300933838, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5015259385108948, 'sampling/importance_sampling_ratio/max': 2.97066068649292, 'entropy': 0.5492652888642624, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 293.17900502029806, 'epoch': 0.38}
19%|█▉ | 26/136 [2:34:15<9:50:38, 322.17s/it][2026-08-13 01:04:31] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,4)
[2026-08-13 01:04:32] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,4)
[2026-08-13 01:04:32] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,315)
[2026-08-13 01:04:32] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\)' @(1,437)
[2026-08-13 01:04:32] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,91)
[2026-08-13 01:04:32] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,92)
[2026-08-13 01:04:32] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,93)
20%|█▉ | 27/136 [2:39:07<9:28:45, 313.08s/it]
{'loss': -0.024, 'grad_norm': 0.0600763917962412, 'learning_rate': 2.8323895991589866e-06, 'num_tokens': 32203420.0, 'completions/mean_length': 741.4765625, 'completions/min_length': 133.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.0611979179084301, 'completions/mean_terminated_length': 589.5562133789062, 'completions/min_terminated_length': 133.0, 'completions/max_terminated_length': 1762.0, 'rewards/reward_correctness/mean': 0.5944010615348816, 'rewards/reward_correctness/std': 0.49116748571395874, 'reward': 0.5944010615348816, 'reward_std': 0.49116748571395874, 'frac_reward_zero_std': 0.46875, 'sampling/sampling_logp_difference/mean': 0.025170018896460533, 'sampling/sampling_logp_difference/max': 0.997067928314209, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5418947339057922, 'sampling/importance_sampling_ratio/max': 2.97564697265625, 'entropy': 0.5579174467129633, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 290.5366625688039, 'epoch': 0.4}
20%|█▉ | 27/136 [2:39:07<9:28:45, 313.08s/it][2026-08-13 01:09:26] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,109)
21%|██ | 28/136 [2:44:36<9:32:12, 317.89s/it]
{'loss': -0.0144, 'grad_norm': 0.041126690061418614, 'learning_rate': 2.8164270309259034e-06, 'num_tokens': 33424049.0, 'completions/mean_length': 702.5709838867188, 'completions/min_length': 149.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.0481770858168602, 'completions/mean_terminated_length': 584.0923461914062, 'completions/min_terminated_length': 149.0, 'completions/max_terminated_length': 1965.0, 'rewards/reward_correctness/mean': 0.5846354365348816, 'rewards/reward_correctness/std': 0.49294528365135193, 'reward': 0.5846354365348816, 'reward_std': 0.49294525384902954, 'frac_reward_zero_std': 0.515625, 'sampling/sampling_logp_difference/mean': 0.041104961186647415, 'sampling/sampling_logp_difference/max': 10.12062931060791, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.4718739092350006, 'sampling/importance_sampling_ratio/max': 2.9962897300720215, 'entropy': 0.5106982593424618, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 327.90168132819235, 'epoch': 0.41}
21%|██ | 28/136 [2:44:36<9:32:12, 317.89s/it][2026-08-13 01:14:51] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,24)
[2026-08-13 01:14:52] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "text" @(1,5)
Open LaTeX blocks:
@(1,5) arguments of macro "text"
[2026-08-13 01:14:53] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,22)
21%|██▏ | 29/136 [2:49:22<9:09:54, 308.36s/it]
{'loss': -0.0207, 'grad_norm': 0.05843225178744546, 'learning_rate': 2.79979366136247e-06, 'num_tokens': 34740930.0, 'completions/mean_length': 766.5319213867188, 'completions/min_length': 146.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.0716145858168602, 'completions/mean_terminated_length': 588.6907348632812, 'completions/min_terminated_length': 146.0, 'completions/max_terminated_length': 1842.0, 'rewards/reward_correctness/mean': 0.560546875, 'rewards/reward_correctness/std': 0.4964821934700012, 'reward': 0.560546875, 'reward_std': 0.49648216366767883, 'frac_reward_zero_std': 0.453125, 'sampling/sampling_logp_difference/mean': 0.026250643655657768, 'sampling/sampling_logp_difference/max': 1.1054515838623047, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5357107520103455, 'sampling/importance_sampling_ratio/max': 2.9964332580566406, 'entropy': 0.5984537588665262, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 284.9848792338744, 'epoch': 0.43}
21%|██▏ | 29/136 [2:49:22<9:09:54, 308.36s/it][2026-08-13 01:19:36] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,45)
[2026-08-13 01:19:36] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,46)
22%|██▏ | 30/136 [2:54:19<8:58:46, 304.97s/it]
{'loss': -0.0076, 'grad_norm': 0.04678745046376859, 'learning_rate': 2.7824990561647276e-06, 'num_tokens': 35927967.0, 'completions/mean_length': 686.326171875, 'completions/min_length': 120.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.0494791679084301, 'completions/mean_terminated_length': 562.62744140625, 'completions/min_terminated_length': 120.0, 'completions/max_terminated_length': 1768.0, 'rewards/reward_correctness/mean': 0.6080729365348816, 'rewards/reward_correctness/std': 0.48833951354026794, 'reward': 0.6080729365348816, 'reward_std': 0.48833951354026794, 'frac_reward_zero_std': 0.5, 'sampling/sampling_logp_difference/mean': 0.02340259775519371, 'sampling/sampling_logp_difference/max': 1.1287574768066406, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5746848583221436, 'sampling/importance_sampling_ratio/max': 2.9987664222717285, 'entropy': 0.49350566777866334, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 295.85874052438885, 'epoch': 0.44}
22%|██▏ | 30/136 [2:54:19<8:58:46, 304.97s/it]
0%| | 0/84 [00:00<?, ?it/s]
2%|▏ | 2/84 [00:06<04:20, 3.18s/it]
4%|▎ | 3/84 [00:10<04:57, 3.67s/it]
5%|▍ | 4/84 [00:16<06:00, 4.51s/it]
6%|▌ | 5/84 [00:23<06:56, 5.27s/it]
7%|▋ | 6/84 [00:31<08:09, 6.28s/it]
8%|▊ | 7/84 [00:39<08:42, 6.78s/it]
10%|▉ | 8/84 [00:42<07:07, 5.63s/it]
11%|█ | 9/84 [00:51<08:25, 6.73s/it]
12%|█▏ | 10/84 [00:58<08:10, 6.63s/it]
13%|█▎ | 11/84 [01:06<08:46, 7.22s/it]
14%|█▍ | 12/84 [01:12<08:05, 6.75s/it]
15%|█▌ | 13/84 [01:36<14:01, 11.85s/it]
17%|█▋ | 14/84 [01:42<11:58, 10.27s/it]
18%|█▊ | 15/84 [01:48<10:09, 8.83s/it]
19%|█▉ | 16/84 [02:02<11:55, 10.52s/it]
20%|██ | 17/84 [02:08<10:17, 9.22s/it]
21%|██▏ | 18/84 [02:13<08:36, 7.82s/it]
23%|██▎ | 19/84 [02:20<08:22, 7.73s/it]
24%|██▍ | 20/84 [02:26<07:41, 7.21s/it]
25%|██▌ | 21/84 [02:32<06:59, 6.66s/it]
26%|██▌ | 22/84 [02:37<06:23, 6.18s/it]
27%|██▋ | 23/84 [02:42<05:56, 5.84s/it]
29%|██▊ | 24/84 [02:48<05:54, 5.90s/it]
30%|██▉ | 25/84 [02:52<05:19, 5.41s/it]
31%|███ | 26/84 [02:57<05:09, 5.33s/it]
32%|███▏ | 27/84 [03:04<05:29, 5.78s/it]
33%|███▎ | 28/84 [03:10<05:22, 5.77s/it]
35%|███▍ | 29/84 [03:17<05:35, 6.10s/it]
36%|███▌ | 30/84 [03:24<05:46, 6.42s/it]
37%|███▋ | 31/84 [03:29<05:26, 6.16s/it]
38%|███▊ | 32/84 [03:35<05:03, 5.84s/it]
39%|███▉ | 33/84 [03:45<06:15, 7.35s/it]
40%|████ | 34/84 [03:54<06:32, 7.85s/it]
42%|████▏ | 35/84 [04:04<06:44, 8.26s/it]
43%|████▎ | 36/84 [04:10<06:07, 7.66s/it]
44%|████▍ | 37/84 [04:15<05:22, 6.86s/it]
45%|████▌ | 38/84 [04:21<04:58, 6.50s/it]
46%|████▋ | 39/84 [04:27<04:57, 6.62s/it]
48%|████▊ | 40/84 [04:49<08:12, 11.19s/it]
49%|████▉ | 41/84 [04:54<06:41, 9.34s/it]
50%|█████ | 42/84 [05:03<06:20, 9.06s/it]
51%|█████ | 43/84 [05:10<05:53, 8.62s/it]
52%|█████▏ | 44/84 [05:15<04:54, 7.36s/it]
54%|█████▎ | 45/84 [05:25<05:24, 8.32s/it]
55%|█████▍ | 46/84 [05:34<05:20, 8.43s/it]
56%|█████▌ | 47/84 [05:44<05:28, 8.87s/it]
57%|█████▋ | 48/84 [05:52<05:13, 8.70s/it]
58%|█████▊ | 49/84 [05:59<04:43, 8.11s/it]
60%|█████▉ | 50/84 [06:08<04:42, 8.31s/it]
61%|██████ | 51/84 [06:13<04:03, 7.37s/it]
62%|██████▏ | 52/84 [06:20<03:56, 7.38s/it]
63%|██████▎ | 53/84 [06:25<03:24, 6.60s/it]
64%|██████▍ | 54/84 [06:31<03:07, 6.25s/it]
65%|██████▌ | 55/84 [06:42<03:44, 7.74s/it]
67%|██████▋ | 56/84 [06:46<03:04, 6.57s/it]
68%|██████▊ | 57/84 [06:52<02:59, 6.65s/it]
69%|██████▉ | 58/84 [06:58<02:43, 6.27s/it]
70%|███████ | 59/84 [07:04<02:39, 6.37s/it]
71%|███████▏ | 60/84 [07:10<02:25, 6.04s/it]
73%|███████▎ | 61/84 [07:16<02:19, 6.07s/it]
74%|███████▍ | 62/84 [07:23<02:23, 6.51s/it]
75%|███████▌ | 63/84 [07:33<02:33, 7.30s/it]
76%|███████▌ | 64/84 [07:42<02:39, 7.97s/it]
77%|███████▋ | 65/84 [07:54<02:56, 9.31s/it]
79%|███████▊ | 66/84 [08:01<02:33, 8.51s/it]
80%|███████▉ | 67/84 [08:08<02:17, 8.07s/it]
81%|████████ | 68/84 [08:20<02:24, 9.06s/it]
82%|████████▏ | 69/84 [08:27<02:07, 8.53s/it]
83%|████████▎ | 70/84 [08:32<01:47, 7.66s/it]
85%|████████▍ | 71/84 [08:36<01:22, 6.38s/it]
86%|████████▌ | 72/84 [08:43<01:18, 6.52s/it]
87%|████████▋ | 73/84 [08:51<01:17, 7.07s/it]
88%|████████▊ | 74/84 [08:58<01:11, 7.16s/it]
89%|████████▉ | 75/84 [09:05<01:02, 6.93s/it]
90%|█████████ | 76/84 [09:12<00:56, 7.03s/it]
92%|█████████▏| 77/84 [09:18<00:46, 6.61s/it]
93%|█████████▎| 78/84 [09:24<00:38, 6.48s/it]
94%|█████████▍| 79/84 [09:31<00:33, 6.72s/it]
95%|█████████▌| 80/84 [09:40<00:29, 7.25s/it]
96%|█████████▋| 81/84 [09:46<00:21, 7.01s/it]
98%|█████████▊| 82/84 [09:55<00:15, 7.56s/it]
99%|█████████▉| 83/84 [10:01<00:07, 7.09s/it]
100%|██████████| 84/84 [10:08<00:00, 7.03s/it]
{'eval_loss': 0.0, 'eval_runtime': 614.0577, 'eval_samples_per_second': 0.814, 'eval_steps_per_second': 0.137, 'eval_num_tokens': 35927967.0, 'eval_completions/mean_length': 562.2936713809058, 'eval_completions/min_length': 265.85714285714283, 'eval_completions/max_length': 1007.0119047619048, 'eval_completions/clipped_ratio': 0.001984127043258576, 'eval_completions/mean_terminated_length': 557.4218459356399, 'eval_completions/min_terminated_length': 265.85714285714283, 'eval_completions/max_terminated_length': 979.1071428571429, 'eval_rewards/reward_correctness/mean': 0.7162698402646042, 'eval_rewards/reward_correctness/std': 0.43203149097306387, 'eval_reward': 0.7162698643902937, 'eval_reward_std': 0.4320314891991161, 'eval_frac_reward_zero_std': 1.0, 'eval_sampling/sampling_logp_difference/mean': 0.05504328708740927, 'eval_sampling/sampling_logp_difference/max': 2.629038010324751, 'eval_sampling/importance_sampling_ratio/min': 2.4404202079057176e-06, 'eval_sampling/importance_sampling_ratio/mean': 0.04295138852800543, 'eval_sampling/importance_sampling_ratio/max': 0.22348843947717492, 'eval_entropy': 0.22903412624838806, 'eval_clip_ratio/low_mean': 0.0, 'eval_clip_ratio/low_min': 0.0, 'eval_clip_ratio/high_mean': 0.0, 'eval_clip_ratio/high_max': 0.0, 'eval_clip_ratio/region_mean': 0.0, 'epoch': 0.44}
22%|██▏ | 30/136 [3:04:33<8:58:46, 304.97s/it]
100%|██████████| 84/84 [10:08<00:00, 7.03s/it]
[2026-08-13 01:34:50] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,79)
[2026-08-13 01:34:51] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,116)
23%|██▎ | 31/136 [3:09:32<14:12:56, 487.39s/it]
{'loss': -0.0266, 'grad_norm': 0.048461134100390504, 'learning_rate': 2.7645531612991763e-06, 'num_tokens': 37204179.0, 'completions/mean_length': 737.421875, 'completions/min_length': 147.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.064453125, 'completions/mean_terminated_length': 576.7390747070312, 'completions/min_terminated_length': 147.0, 'completions/max_terminated_length': 1561.0, 'rewards/reward_correctness/mean': 0.58203125, 'rewards/reward_correctness/std': 0.4933856129646301, 'reward': 0.58203125, 'reward_std': 0.49338558316230774, 'frac_reward_zero_std': 0.53125, 'sampling/sampling_logp_difference/mean': 0.023642152547836304, 'sampling/sampling_logp_difference/max': 1.9513862133026123, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5539844036102295, 'sampling/importance_sampling_ratio/max': 2.982936382293701, 'entropy': 0.48714099975768477, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 292.74457041174173, 'epoch': 0.46}
23%|██▎ | 31/136 [3:09:32<14:12:56, 487.39s/it]
24%|██▎ | 32/136 [3:14:13<12:17:42, 425.61s/it]
{'loss': -0.0023, 'grad_norm': 0.048193616973374806, 'learning_rate': 2.745966297282944e-06, 'num_tokens': 38554066.0, 'completions/mean_length': 799.5123901367188, 'completions/min_length': 128.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.0690104216337204, 'completions/mean_terminated_length': 631.062255859375, 'completions/min_terminated_length': 128.0, 'completions/max_terminated_length': 1547.0, 'rewards/reward_correctness/mean': 0.564453125, 'rewards/reward_correctness/std': 0.4959898591041565, 'reward': 0.564453125, 'reward_std': 0.4959898591041565, 'frac_reward_zero_std': 0.515625, 'sampling/sampling_logp_difference/mean': 0.02345161698758602, 'sampling/sampling_logp_difference/max': 6.243207931518555, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5355834364891052, 'sampling/importance_sampling_ratio/max': 2.99735164642334, 'entropy': 0.5004677245160565, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 280.2582347355783, 'epoch': 0.47}
24%|██▎ | 32/136 [3:14:13<12:17:42, 425.61s/it][2026-08-13 01:44:34] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,122)
[2026-08-13 01:44:34] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,11)
[2026-08-13 01:44:35] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "frac" @(1,1)
Open LaTeX blocks:
@(1,1) arguments of macro "frac"
[2026-08-13 01:44:35] WARNING __init__.py:1389: WARNING: Error in configuration: macro '\frac' failed its substitution!
24%|██▍ | 33/136 [3:19:17<11:07:36, 388.90s/it]
{'loss': -0.0388, 'grad_norm': 0.05577718738565114, 'learning_rate': 2.726749153248549e-06, 'num_tokens': 39900384.0, 'completions/mean_length': 778.2904052734375, 'completions/min_length': 150.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.076171875, 'completions/mean_terminated_length': 589.1683959960938, 'completions/min_terminated_length': 150.0, 'completions/max_terminated_length': 2621.0, 'rewards/reward_correctness/mean': 0.5826823115348816, 'rewards/reward_correctness/std': 0.49327683448791504, 'reward': 0.5826823115348816, 'reward_std': 0.49327683448791504, 'frac_reward_zero_std': 0.4921875, 'sampling/sampling_logp_difference/mean': 0.026426147669553757, 'sampling/sampling_logp_difference/max': 0.9872760772705078, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5347757935523987, 'sampling/importance_sampling_ratio/max': 2.9583816528320312, 'entropy': 0.5261909196851775, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 302.06614747131243, 'epoch': 0.49}
24%|██▍ | 33/136 [3:19:17<11:07:36, 388.90s/it][2026-08-13 01:49:40] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,4)
[2026-08-13 01:49:40] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,6)
[2026-08-13 01:49:40] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,7)
25%|██▌ | 34/136 [3:24:23<10:19:13, 364.25s/it]
{'loss': -0.0477, 'grad_norm': 0.06494124688870825, 'learning_rate': 2.706912780796687e-06, 'num_tokens': 41176124.0, 'completions/mean_length': 722.3021240234375, 'completions/min_length': 186.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.0611979179084301, 'completions/mean_terminated_length': 569.1824340820312, 'completions/min_terminated_length': 186.0, 'completions/max_terminated_length': 1492.0, 'rewards/reward_correctness/mean': 0.6178385615348816, 'rewards/reward_correctness/std': 0.48607397079467773, 'reward': 0.6178385615348816, 'reward_std': 0.48607394099235535, 'frac_reward_zero_std': 0.5078125, 'sampling/sampling_logp_difference/mean': 0.024992134422063828, 'sampling/sampling_logp_difference/max': 14.96677017211914, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5623457431793213, 'sampling/importance_sampling_ratio/max': 2.944589614868164, 'entropy': 0.507065630867146, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 305.60695807402954, 'epoch': 0.5}
25%|██▌ | 34/136 [3:24:23<10:19:13, 364.25s/it][2026-08-13 01:54:41] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,31)
26%|██▌ | 35/136 [3:29:22<9:39:54, 344.50s/it]
{'loss': -0.0135, 'grad_norm': 0.053901537826546005, 'learning_rate': 2.686468587640551e-06, 'num_tokens': 42518328.0, 'completions/mean_length': 775.7916870117188, 'completions/min_length': 148.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.0677083358168602, 'completions/mean_terminated_length': 609.6054077148438, 'completions/min_terminated_length': 148.0, 'completions/max_terminated_length': 3068.0, 'rewards/reward_correctness/mean': 0.60546875, 'rewards/reward_correctness/std': 0.48890894651412964, 'reward': 0.60546875, 'reward_std': 0.48890891671180725, 'frac_reward_zero_std': 0.46875, 'sampling/sampling_logp_difference/mean': 0.023517360910773277, 'sampling/sampling_logp_difference/max': 1.987520694732666, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5776769518852234, 'sampling/importance_sampling_ratio/max': 2.9897232055664062, 'entropy': 0.4794918481493369, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 297.2315682750195, 'epoch': 0.51}
26%|██▌ | 35/136 [3:29:22<9:39:54, 344.50s/it][2026-08-13 01:59:48] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,262)
[2026-08-13 01:59:49] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "frac" @(1,299)
Open LaTeX blocks:
@(1,299) arguments of macro "frac"
[2026-08-13 01:59:49] WARNING __init__.py:1389: WARNING: Error in configuration: macro '\frac' failed its substitution!
26%|██▋ | 36/136 [3:34:33<9:17:44, 334.64s/it]
{'loss': -0.0357, 'grad_norm': 0.05513570710996907, 'learning_rate': 2.6654283310453644e-06, 'num_tokens': 43934372.0, 'completions/mean_length': 819.9349365234375, 'completions/min_length': 151.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.091796875, 'completions/mean_terminated_length': 593.167724609375, 'completions/min_terminated_length': 151.0, 'completions/max_terminated_length': 1907.0, 'rewards/reward_correctness/mean': 0.5709635615348816, 'rewards/reward_correctness/std': 0.4950997531414032, 'reward': 0.5709635615348816, 'reward_std': 0.4950997233390808, 'frac_reward_zero_std': 0.5234375, 'sampling/sampling_logp_difference/mean': 0.029513418674468994, 'sampling/sampling_logp_difference/max': 2.370652437210083, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5423656105995178, 'sampling/importance_sampling_ratio/max': 2.9943625926971436, 'entropy': 0.5791506089735776, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 310.507658964023, 'epoch': 0.53}
26%|██▋ | 36/136 [3:34:33<9:17:44, 334.64s/it][2026-08-13 02:04:58] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,43)
[2026-08-13 02:04:58] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,45)
27%|██▋ | 37/136 [3:39:46<9:01:15, 328.03s/it]
{'loss': -0.0157, 'grad_norm': 0.046276621163544786, 'learning_rate': 2.643804111066888e-06, 'num_tokens': 45276405.0, 'completions/mean_length': 779.7272338867188, 'completions/min_length': 157.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.0813802108168602, 'completions/mean_terminated_length': 576.839111328125, 'completions/min_terminated_length': 157.0, 'completions/max_terminated_length': 2379.0, 'rewards/reward_correctness/mean': 0.6263020634651184, 'rewards/reward_correctness/std': 0.4839424192905426, 'reward': 0.6263021230697632, 'reward_std': 0.4839423894882202, 'frac_reward_zero_std': 0.5234375, 'sampling/sampling_logp_difference/mean': 0.026051774621009827, 'sampling/sampling_logp_difference/max': 0.8799822330474854, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5235354900360107, 'sampling/importance_sampling_ratio/max': 2.992263078689575, 'entropy': 0.49720528721809387, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 311.45988411689177, 'epoch': 0.54}
27%|██▋ | 37/136 [3:39:46<9:01:15, 328.03s/it][2026-08-13 02:10:11] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,6)
[2026-08-13 02:10:11] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,237)
[2026-08-13 02:10:11] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,498)
[2026-08-13 02:10:11] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,1184)
28%|██▊ | 38/136 [3:44:46<8:42:02, 319.61s/it]
{'loss': -0.0109, 'grad_norm': 0.0379364986328415, 'learning_rate': 2.6216083635927896e-06, 'num_tokens': 46624845.0, 'completions/mean_length': 790.8515625, 'completions/min_length': 112.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.0872395858168602, 'completions/mean_terminated_length': 572.8245239257812, 'completions/min_terminated_length': 112.0, 'completions/max_terminated_length': 1678.0, 'rewards/reward_correctness/mean': 0.5638020634651184, 'rewards/reward_correctness/std': 0.4960740804672241, 'reward': 0.5638021230697632, 'reward_std': 0.4960740804672241, 'frac_reward_zero_std': 0.53125, 'sampling/sampling_logp_difference/mean': 0.028229426592588425, 'sampling/sampling_logp_difference/max': 1.3098070621490479, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5088026523590088, 'sampling/importance_sampling_ratio/max': 2.985919713973999, 'entropy': 0.5321638238383457, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 298.7274723085575, 'epoch': 0.56}
28%|██▊ | 38/136 [3:44:46<8:42:02, 319.61s/it][2026-08-13 02:15:10] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,758)
29%|██▊ | 39/136 [3:49:50<8:29:18, 315.04s/it]
{'loss': -0.0118, 'grad_norm': 0.05913528982233647, 'learning_rate': 2.598853853190882e-06, 'num_tokens': 47975881.0, 'completions/mean_length': 789.1823120117188, 'completions/min_length': 120.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.083984375, 'completions/mean_terminated_length': 579.9559326171875, 'completions/min_terminated_length': 120.0, 'completions/max_terminated_length': 2886.0, 'rewards/reward_correctness/mean': 0.6067708134651184, 'rewards/reward_correctness/std': 0.48862606287002563, 'reward': 0.6067708730697632, 'reward_std': 0.48862606287002563, 'frac_reward_zero_std': 0.46875, 'sampling/sampling_logp_difference/mean': 0.026590874418616295, 'sampling/sampling_logp_difference/max': 1.0663793087005615, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.540636420249939, 'sampling/importance_sampling_ratio/max': 2.999361991882324, 'entropy': 0.48998611723072827, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 303.23120561568066, 'epoch': 0.57}
29%|██▊ | 39/136 [3:49:50<8:29:18, 315.04s/it][2026-08-13 02:20:14] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,1)
[2026-08-13 02:20:15] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,0)
[2026-08-13 02:20:15] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,17)
[2026-08-13 02:20:16] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,4)
[2026-08-13 02:20:16] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "frac" @(1,6)
Open LaTeX blocks:
@(1,6) arguments of macro "frac"
[2026-08-13 02:20:16] WARNING __init__.py:1389: WARNING: Error in configuration: macro '\frac' failed its substitution!
29%|██▉ | 40/136 [3:54:54<8:18:44, 311.71s/it]
{'loss': -0.0209, 'grad_norm': 0.0500404243931231, 'learning_rate': 2.5755536657683354e-06, 'num_tokens': 49384103.0, 'completions/mean_length': 825.0299682617188, 'completions/min_length': 158.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1028645858168602, 'completions/mean_terminated_length': 567.4934692382812, 'completions/min_terminated_length': 158.0, 'completions/max_terminated_length': 2553.0, 'rewards/reward_correctness/mean': 0.6022135615348816, 'rewards/reward_correctness/std': 0.4896003007888794, 'reward': 0.6022135615348816, 'reward_std': 0.489600270986557, 'frac_reward_zero_std': 0.5390625, 'sampling/sampling_logp_difference/mean': 0.02729925513267517, 'sampling/sampling_logp_difference/max': 0.9798126220703125, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5711022019386292, 'sampling/importance_sampling_ratio/max': 2.989989757537842, 'entropy': 0.5337136797606945, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 302.83849718095735, 'epoch': 0.59}
29%|██▉ | 40/136 [3:54:54<8:18:44, 311.71s/it]
0%| | 0/84 [00:00<?, ?it/s]
2%|▏ | 2/84 [00:05<03:41, 2.70s/it]
4%|▎ | 3/84 [00:09<04:19, 3.20s/it]
5%|▍ | 4/84 [00:32<14:23, 10.79s/it]
6%|▌ | 5/84 [00:39<12:25, 9.43s/it]
7%|▋ | 6/84 [00:46<11:13, 8.64s/it]
8%|▊ | 7/84 [00:54<10:36, 8.26s/it]
10%|▉ | 8/84 [00:57<08:33, 6.76s/it]
11%|█ | 9/84 [01:05<08:55, 7.14s/it]
12%|█▏ | 10/84 [01:11<08:13, 6.67s/it]
13%|█▎ | 11/84 [01:18<08:16, 6.80s/it]
14%|█▍ | 12/84 [01:24<08:03, 6.71s/it]
15%|█▌ | 13/84 [01:31<07:54, 6.69s/it]
17%|█▋ | 14/84 [01:37<07:32, 6.47s/it]
18%|█▊ | 15/84 [01:43<07:25, 6.46s/it]
19%|█▉ | 16/84 [01:49<07:01, 6.20s/it]
20%|██ | 17/84 [01:55<06:48, 6.10s/it]
21%|██▏ | 18/84 [02:18<12:25, 11.29s/it]
23%|██▎ | 19/84 [02:28<11:39, 10.77s/it]
24%|██▍ | 20/84 [02:36<10:37, 9.96s/it]
25%|██▌ | 21/84 [02:49<11:25, 10.89s/it]
26%|██▌ | 22/84 [02:53<09:00, 8.72s/it]
27%|██▋ | 23/84 [02:59<08:06, 7.98s/it]
29%|██▊ | 24/84 [03:05<07:22, 7.38s/it]
30%|██▉ | 25/84 [03:12<07:14, 7.36s/it]
31%|███ | 26/84 [03:17<06:27, 6.68s/it]
32%|███▏ | 27/84 [03:24<06:22, 6.71s/it]
33%|███▎ | 28/84 [03:30<06:04, 6.51s/it]
35%|███▍ | 29/84 [03:37<06:12, 6.78s/it]
36%|███▌ | 30/84 [03:44<05:56, 6.59s/it]
37%|███▋ | 31/84 [04:07<10:18, 11.67s/it]
38%|███▊ | 32/84 [04:12<08:18, 9.58s/it]
39%|███▉ | 33/84 [04:20<07:41, 9.04s/it]
40%|████ | 34/84 [04:27<07:13, 8.67s/it]
42%|████▏ | 35/84 [04:51<10:43, 13.13s/it]
43%|████▎ | 36/84 [04:57<08:40, 10.85s/it]
44%|████▍ | 37/84 [05:01<06:58, 8.90s/it]
45%|████▌ | 38/84 [05:08<06:24, 8.36s/it]
46%|████▋ | 39/84 [05:14<05:40, 7.57s/it]
48%|████▊ | 40/84 [05:24<06:13, 8.49s/it]
49%|████▉ | 41/84 [05:29<05:20, 7.45s/it]
50%|█████ | 42/84 [05:37<05:15, 7.51s/it]
51%|█████ | 43/84 [05:45<05:19, 7.78s/it]
52%|█████▏ | 44/84 [05:51<04:40, 7.01s/it]
54%|█████▎ | 45/84 [06:15<07:52, 12.11s/it]
55%|█████▍ | 46/84 [06:21<06:35, 10.40s/it]
56%|█████▌ | 47/84 [06:26<05:28, 8.87s/it]
57%|█████▋ | 48/84 [06:35<05:11, 8.66s/it]
58%|█████▊ | 49/84 [06:41<04:41, 8.05s/it]
60%|█████▉ | 50/84 [06:50<04:45, 8.41s/it]
61%|██████ | 51/84 [06:58<04:31, 8.24s/it]
62%|██████▏ | 52/84 [07:04<04:00, 7.50s/it]
63%|██████▎ | 53/84 [07:08<03:21, 6.49s/it]
64%|██████▍ | 54/84 [07:12<02:53, 5.77s/it]
65%|██████▌ | 55/84 [07:18<02:45, 5.72s/it]
67%|██████▋ | 56/84 [07:22<02:23, 5.12s/it]
68%|██████▊ | 57/84 [07:28<02:31, 5.62s/it]
69%|██████▉ | 58/84 [07:34<02:26, 5.65s/it]
70%|███████ | 59/84 [07:42<02:35, 6.21s/it]
71%|███████▏ | 60/84 [07:47<02:20, 5.86s/it]
73%|███████▎ | 61/84 [07:52<02:14, 5.87s/it]
74%|███████▍ | 62/84 [08:00<02:20, 6.39s/it]
75%|███████▌ | 63/84 [08:09<02:29, 7.10s/it]
76%|███████▌ | 64/84 [08:17<02:26, 7.30s/it]
77%|███████▋ | 65/84 [08:29<02:49, 8.95s/it]
79%|███████▊ | 66/84 [08:34<02:16, 7.57s/it]
80%|███████▉ | 67/84 [08:41<02:04, 7.33s/it]
81%|████████ | 68/84 [08:50<02:06, 7.94s/it]
82%|████████▏ | 69/84 [08:57<01:53, 7.58s/it]
83%|████████▎ | 70/84 [09:04<01:46, 7.62s/it]
85%|████████▍ | 71/84 [09:09<01:26, 6.63s/it]
86%|████████▌ | 72/84 [09:15<01:19, 6.62s/it]
87%|████████▋ | 73/84 [09:21<01:10, 6.43s/it]
88%|████████▊ | 74/84 [09:27<01:02, 6.24s/it]
89%|████████▉ | 75/84 [09:34<00:59, 6.57s/it]
90%|█████████ | 76/84 [09:44<00:58, 7.37s/it]
92%|█████████▏| 77/84 [09:49<00:46, 6.71s/it]
93%|█████████▎| 78/84 [09:55<00:40, 6.69s/it]
94%|█████████▍| 79/84 [10:02<00:33, 6.65s/it]
95%|█████████▌| 80/84 [10:10<00:28, 7.09s/it]
96%|█████████▋| 81/84 [10:18<00:21, 7.20s/it]
98%|█████████▊| 82/84 [10:41<00:24, 12.13s/it]
99%|█████████▉| 83/84 [10:47<00:10, 10.22s/it]
100%|██████████| 84/84 [10:53<00:00, 8.90s/it]
{'eval_loss': 0.0, 'eval_runtime': 658.6168, 'eval_samples_per_second': 0.759, 'eval_steps_per_second': 0.128, 'eval_num_tokens': 49384103.0, 'eval_completions/mean_length': 575.2956495739165, 'eval_completions/min_length': 274.1904761904762, 'eval_completions/max_length': 1071.4404761904761, 'eval_completions/clipped_ratio': 0.011904762259551458, 'eval_completions/mean_terminated_length': 546.4504107520694, 'eval_completions/min_terminated_length': 274.1904761904762, 'eval_completions/max_terminated_length': 927.4404761904761, 'eval_rewards/reward_correctness/mean': 0.7103174589574337, 'eval_rewards/reward_correctness/std': 0.40739349524180096, 'eval_reward': 0.7103174795352277, 'eval_reward_std': 0.40739349524180096, 'eval_frac_reward_zero_std': 1.0, 'eval_sampling/sampling_logp_difference/mean': 0.05068082945598733, 'eval_sampling/sampling_logp_difference/max': 2.551057347229549, 'eval_sampling/importance_sampling_ratio/min': 6.423417091128806e-05, 'eval_sampling/importance_sampling_ratio/mean': 0.03784746756954995, 'eval_sampling/importance_sampling_ratio/max': 0.19148892058893535, 'eval_entropy': 0.20750312196711698, 'eval_clip_ratio/low_mean': 0.0, 'eval_clip_ratio/low_min': 0.0, 'eval_clip_ratio/high_mean': 0.0, 'eval_clip_ratio/high_max': 0.0, 'eval_clip_ratio/region_mean': 0.0, 'epoch': 0.59}
29%|██▉ | 40/136 [4:05:53<8:18:44, 311.71s/it]
100%|██████████| 84/84 [10:53<00:00, 8.90s/it]

30%|███ | 41/136 [4:10:57<13:22:56, 507.12s/it]
{'loss': -0.0233, 'grad_norm': 0.05273673949955187, 'learning_rate': 2.551721201046098e-06, 'num_tokens': 50811980.0, 'completions/mean_length': 836.333984375, 'completions/min_length': 164.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1022135466337204, 'completions/mean_terminated_length': 581.802001953125, 'completions/min_terminated_length': 164.0, 'completions/max_terminated_length': 2139.0, 'rewards/reward_correctness/mean': 0.5833333134651184, 'rewards/reward_correctness/std': 0.4931671917438507, 'reward': 0.5833333730697632, 'reward_std': 0.4931671619415283, 'frac_reward_zero_std': 0.46875, 'sampling/sampling_logp_difference/mean': 0.02731555700302124, 'sampling/sampling_logp_difference/max': 1.419299840927124, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.528105616569519, 'sampling/importance_sampling_ratio/max': 2.9984641075134277, 'entropy': 0.515403852565214, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 298.77585212513804, 'epoch': 0.6}
30%|███ | 41/136 [4:10:57<13:22:56, 507.12s/it][2026-08-13 02:41:25] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "frac" @(1,2)
Open LaTeX blocks:
@(1,2) arguments of macro "frac"
[2026-08-13 02:41:25] WARNING __init__.py:1389: WARNING: Error in configuration: macro '\frac' failed its substitution!
31%|███ | 42/136 [4:16:00<11:38:21, 445.76s/it]
{'loss': -0.0332, 'grad_norm': 0.04893596319312678, 'learning_rate': 2.5273701648528393e-06, 'num_tokens': 52270286.0, 'completions/mean_length': 858.44140625, 'completions/min_length': 135.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1015625, 'completions/mean_terminated_length': 608.2130737304688, 'completions/min_terminated_length': 135.0, 'completions/max_terminated_length': 1971.0, 'rewards/reward_correctness/mean': 0.615234375, 'rewards/reward_correctness/std': 0.48669832944869995, 'reward': 0.615234375, 'reward_std': 0.48669832944869995, 'frac_reward_zero_std': 0.5078125, 'sampling/sampling_logp_difference/mean': 0.02453477308154106, 'sampling/sampling_logp_difference/max': 1.9809277057647705, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5512368083000183, 'sampling/importance_sampling_ratio/max': 2.9960122108459473, 'entropy': 0.47979187313467264, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 301.4260444557294, 'epoch': 0.62}
31%|███ | 42/136 [4:16:00<11:38:21, 445.76s/it][2026-08-13 02:46:27] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,0)
[2026-08-13 02:46:27] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Bad \end macro: expected {environmentname} @(1,22)
[2026-08-13 02:46:29] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,25)
[2026-08-13 02:46:30] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\)' @(1,1)
32%|███▏ | 43/136 [4:21:12<10:28:34, 405.53s/it]
{'loss': -0.0288, 'grad_norm': 0.06291468578251484, 'learning_rate': 2.5025145612428566e-06, 'num_tokens': 53690056.0, 'completions/mean_length': 823.6107177734375, 'completions/min_length': 92.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1002604216337204, 'completions/mean_terminated_length': 573.2756958007812, 'completions/min_terminated_length': 92.0, 'completions/max_terminated_length': 2852.0, 'rewards/reward_correctness/mean': 0.6158854365348816, 'rewards/reward_correctness/std': 0.48654359579086304, 'reward': 0.6158854365348816, 'reward_std': 0.48654359579086304, 'frac_reward_zero_std': 0.4765625, 'sampling/sampling_logp_difference/mean': 0.02832927368581295, 'sampling/sampling_logp_difference/max': 0.9814343452453613, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5603159070014954, 'sampling/importance_sampling_ratio/max': 2.9952244758605957, 'entropy': 0.5395284099504352, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 310.514519975055, 'epoch': 0.63}
32%|███▏ | 43/136 [4:21:12<10:28:34, 405.53s/it]
32%|███▏ | 44/136 [4:26:33<9:43:07, 380.30s/it]
{'loss': -0.0273, 'grad_norm': 0.043823548874597246, 'learning_rate': 2.47716868444247e-06, 'num_tokens': 55219440.0, 'completions/mean_length': 891.1302490234375, 'completions/min_length': 145.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.12109375, 'completions/mean_terminated_length': 591.8214721679688, 'completions/min_terminated_length': 145.0, 'completions/max_terminated_length': 1776.0, 'rewards/reward_correctness/mean': 0.5872395634651184, 'rewards/reward_correctness/std': 0.4924907684326172, 'reward': 0.5872396230697632, 'reward_std': 0.4924907386302948, 'frac_reward_zero_std': 0.4609375, 'sampling/sampling_logp_difference/mean': 0.02752448432147503, 'sampling/sampling_logp_difference/max': 0.9179315567016602, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5122290253639221, 'sampling/importance_sampling_ratio/max': 2.9666354656219482, 'entropy': 0.5378130340250209, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 320.2574398214929, 'epoch': 0.65}
32%|███▏ | 44/136 [4:26:33<9:43:07, 380.30s/it][2026-08-13 02:57:06] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,489)
33%|███▎ | 45/136 [4:31:45<9:05:42, 359.81s/it]
{'loss': -0.0239, 'grad_norm': 0.04884731721807342, 'learning_rate': 2.4513471106295523e-06, 'num_tokens': 56670071.0, 'completions/mean_length': 843.5462646484375, 'completions/min_length': 130.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1041666716337204, 'completions/mean_terminated_length': 584.4927368164062, 'completions/min_terminated_length': 130.0, 'completions/max_terminated_length': 2330.0, 'rewards/reward_correctness/mean': 0.6197916865348816, 'rewards/reward_correctness/std': 0.4855960011482239, 'reward': 0.6197916865348816, 'reward_std': 0.4855959713459015, 'frac_reward_zero_std': 0.5078125, 'sampling/sampling_logp_difference/mean': 0.02560652606189251, 'sampling/sampling_logp_difference/max': 1.006209373474121, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.552994430065155, 'sampling/importance_sampling_ratio/max': 2.9755513668060303, 'entropy': 0.48280362505465746, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 310.76242945715785, 'epoch': 0.66}
33%|███▎ | 45/136 [4:31:45<9:05:42, 359.81s/it][2026-08-13 03:02:27] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,18)
[2026-08-13 03:02:28] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,1294)
34%|███▍ | 46/136 [4:37:06<8:42:11, 348.12s/it]
{'loss': -0.0243, 'grad_norm': 0.045178072823470854, 'learning_rate': 2.4250646895508992e-06, 'num_tokens': 58269528.0, 'completions/mean_length': 938.2428588867188, 'completions/min_length': 151.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1380208432674408, 'completions/mean_terminated_length': 596.5838623046875, 'completions/min_terminated_length': 151.0, 'completions/max_terminated_length': 2546.0, 'rewards/reward_correctness/mean': 0.578125, 'rewards/reward_correctness/std': 0.49401959776878357, 'reward': 0.578125, 'reward_std': 0.49401959776878357, 'frac_reward_zero_std': 0.5078125, 'sampling/sampling_logp_difference/mean': 0.02904244139790535, 'sampling/sampling_logp_difference/max': 1.1360305547714233, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5216982960700989, 'sampling/importance_sampling_ratio/max': 2.972738027572632, 'entropy': 0.575145949027501, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 319.69419858045876, 'epoch': 0.68}
34%|███▍ | 46/136 [4:37:06<8:42:11, 348.12s/it]
35%|███▍ | 47/136 [4:42:44<8:31:49, 345.05s/it]
{'loss': -0.0358, 'grad_norm': 0.05207771800211288, 'learning_rate': 2.3983365359822804e-06, 'num_tokens': 59714840.0, 'completions/mean_length': 843.1536865234375, 'completions/min_length': 131.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.103515625, 'completions/mean_terminated_length': 588.5592041015625, 'completions/min_terminated_length': 131.0, 'completions/max_terminated_length': 2219.0, 'rewards/reward_correctness/mean': 0.6100260615348816, 'rewards/reward_correctness/std': 0.4879029095172882, 'reward': 0.6100260615348816, 'reward_std': 0.4879028797149658, 'frac_reward_zero_std': 0.4140625, 'sampling/sampling_logp_difference/mean': 0.04360029101371765, 'sampling/sampling_logp_difference/max': 19.950422286987305, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.48770689964294434, 'sampling/importance_sampling_ratio/max': 2.9499571323394775, 'entropy': 0.47189394827000797, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 336.6966767683625, 'epoch': 0.69}
35%|███▍ | 47/136 [4:42:44<8:31:49, 345.05s/it][2026-08-13 03:13:41] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\)' @(1,4)
35%|███▌ | 48/136 [4:48:17<8:20:50, 341.48s/it]
{'loss': -0.0401, 'grad_norm': 0.05296078839838602, 'learning_rate': 2.3711780210360726e-06, 'num_tokens': 61317331.0, 'completions/mean_length': 949.4993896484375, 'completions/min_length': 144.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1432291716337204, 'completions/mean_terminated_length': 594.6740112304688, 'completions/min_terminated_length': 144.0, 'completions/max_terminated_length': 2586.0, 'rewards/reward_correctness/mean': 0.5950520634651184, 'rewards/reward_correctness/std': 0.4910418391227722, 'reward': 0.5950521230697632, 'reward_std': 0.49104180932044983, 'frac_reward_zero_std': 0.46875, 'sampling/sampling_logp_difference/mean': 0.027828708291053772, 'sampling/sampling_logp_difference/max': 2.5000057220458984, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5409334897994995, 'sampling/importance_sampling_ratio/max': 2.9128077030181885, 'entropy': 0.5676321291830391, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 331.9593873512931, 'epoch': 0.71}
35%|███▌ | 48/136 [4:48:17<8:20:50, 341.48s/it][2026-08-13 03:19:00] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,1519)
[2026-08-13 03:19:01] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,89)
[2026-08-13 03:19:04] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,14)
36%|███▌ | 49/136 [4:53:41<8:07:31, 336.22s/it]
{'loss': -0.0243, 'grad_norm': 0.052468022399943345, 'learning_rate': 2.343604763321476e-06, 'num_tokens': 62849270.0, 'completions/mean_length': 907.8952026367188, 'completions/min_length': 170.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1315104216337204, 'completions/mean_terminated_length': 580.3995361328125, 'completions/min_terminated_length': 170.0, 'completions/max_terminated_length': 1644.0, 'rewards/reward_correctness/mean': 0.5716145634651184, 'rewards/reward_correctness/std': 0.4950059652328491, 'reward': 0.5716146230697632, 'reward_std': 0.49500593543052673, 'frac_reward_zero_std': 0.5234375, 'sampling/sampling_logp_difference/mean': 0.027292978018522263, 'sampling/sampling_logp_difference/max': 1.333378791809082, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5400270223617554, 'sampling/importance_sampling_ratio/max': 2.9808614253997803, 'entropy': 0.5398647881811485, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 322.7655723937787, 'epoch': 0.72}
36%|███▌ | 49/136 [4:53:41<8:07:31, 336.22s/it][2026-08-13 03:24:37] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,477)
37%|███▋ | 50/136 [4:59:14<8:00:46, 335.42s/it]
{'loss': -0.0262, 'grad_norm': 0.03900787430902907, 'learning_rate': 2.3156326199623965e-06, 'num_tokens': 64360474.0, 'completions/mean_length': 896.2552490234375, 'completions/min_length': 142.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.134765625, 'completions/mean_terminated_length': 557.3694458007812, 'completions/min_terminated_length': 142.0, 'completions/max_terminated_length': 2113.0, 'rewards/reward_correctness/mean': 0.6158854365348816, 'rewards/reward_correctness/std': 0.48654359579086304, 'reward': 0.6158854365348816, 'reward_std': 0.4865436255931854, 'frac_reward_zero_std': 0.515625, 'sampling/sampling_logp_difference/mean': 0.028033524751663208, 'sampling/sampling_logp_difference/max': 0.9845733642578125, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5666263699531555, 'sampling/importance_sampling_ratio/max': 2.997588872909546, 'entropy': 0.5093316873535514, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 332.36779574258253, 'epoch': 0.74}
37%|███▋ | 50/136 [4:59:14<8:00:46, 335.42s/it]
0%| | 0/84 [00:00<?, ?it/s]
2%|▏ | 2/84 [00:07<04:55, 3.61s/it]
4%|▎ | 3/84 [00:11<05:26, 4.03s/it]
5%|▍ | 4/84 [00:35<15:07, 11.34s/it]
6%|▌ | 5/84 [00:42<13:11, 10.01s/it]
7%|▋ | 6/84 [00:49<11:37, 8.94s/it]
8%|▊ | 7/84 [00:58<11:32, 8.99s/it]
10%|▉ | 8/84 [01:02<09:10, 7.24s/it]
11%|█ | 9/84 [01:11<09:37, 7.70s/it]
12%|█▏ | 10/84 [01:16<08:50, 7.17s/it]
13%|█▎ | 11/84 [01:23<08:29, 6.97s/it]
14%|█▍ | 12/84 [01:29<08:01, 6.69s/it]
15%|█▌ | 13/84 [01:43<10:41, 9.03s/it]
17%|█▋ | 14/84 [01:50<09:32, 8.18s/it]
18%|█▊ | 15/84 [01:56<08:45, 7.61s/it]
19%|█▉ | 16/84 [02:03<08:24, 7.42s/it]
20%|██ | 17/84 [02:10<08:03, 7.22s/it]
21%|██▏ | 18/84 [02:15<07:18, 6.64s/it]
23%|██▎ | 19/84 [02:39<12:45, 11.78s/it]
24%|██▍ | 20/84 [02:47<11:22, 10.66s/it]
25%|██▌ | 21/84 [02:53<09:49, 9.35s/it]
26%|██▌ | 22/84 [02:58<08:11, 7.92s/it]
27%|██▋ | 23/84 [03:05<07:44, 7.62s/it]
29%|██▊ | 24/84 [03:13<07:51, 7.86s/it]
30%|██▉ | 25/84 [03:17<06:37, 6.74s/it]
31%|███ | 26/84 [03:22<06:05, 6.30s/it]
32%|███▏ | 27/84 [03:30<06:25, 6.77s/it]
33%|███▎ | 28/84 [03:35<05:44, 6.14s/it]
35%|███▍ | 29/84 [03:43<06:00, 6.56s/it]
36%|███▌ | 30/84 [03:50<06:05, 6.78s/it]
37%|███▋ | 31/84 [03:56<05:56, 6.73s/it]
38%|███▊ | 32/84 [04:01<05:10, 5.97s/it]
39%|███▉ | 33/84 [04:07<05:08, 6.05s/it]
40%|████ | 34/84 [04:15<05:29, 6.59s/it]
42%|████▏ | 35/84 [04:25<06:17, 7.71s/it]
43%|████▎ | 36/84 [04:31<05:44, 7.18s/it]
44%|████▍ | 37/84 [04:34<04:40, 5.97s/it]
45%|████▌ | 38/84 [04:40<04:30, 5.87s/it]
46%|████▋ | 39/84 [04:47<04:40, 6.24s/it]
48%|████▊ | 40/84 [05:11<08:26, 11.51s/it]
49%|████▉ | 41/84 [05:16<07:01, 9.79s/it]
50%|█████ | 42/84 [05:40<09:45, 13.94s/it]
51%|█████ | 43/84 [05:46<07:54, 11.58s/it]
52%|█████▏ | 44/84 [05:51<06:16, 9.42s/it]
54%|█████▎ | 45/84 [05:58<05:43, 8.80s/it]
55%|█████▍ | 46/84 [06:08<05:49, 9.19s/it]
56%|█████▌ | 47/84 [06:14<05:01, 8.14s/it]
57%|█████▋ | 48/84 [06:19<04:21, 7.26s/it]
58%|█████▊ | 49/84 [06:25<04:04, 6.99s/it]
60%|█████▉ | 50/84 [06:40<05:14, 9.24s/it]
61%|██████ | 51/84 [06:51<05:21, 9.74s/it]
62%|██████▏ | 52/84 [06:58<04:51, 9.10s/it]
63%|██████▎ | 53/84 [07:03<03:57, 7.65s/it]
64%|██████▍ | 54/84 [07:06<03:15, 6.51s/it]
65%|██████▌ | 55/84 [07:18<03:55, 8.12s/it]
67%|██████▋ | 56/84 [07:22<03:09, 6.76s/it]
68%|██████▊ | 57/84 [07:29<03:04, 6.84s/it]
69%|██████▉ | 58/84 [07:35<02:52, 6.63s/it]
70%|███████ | 59/84 [07:44<03:00, 7.22s/it]
71%|███████▏ | 60/84 [07:49<02:41, 6.74s/it]
73%|███████▎ | 61/84 [07:58<02:48, 7.32s/it]
74%|███████▍ | 62/84 [08:04<02:33, 6.99s/it]
75%|███████▌ | 63/84 [08:15<02:48, 8.04s/it]
76%|███████▌ | 64/84 [08:22<02:38, 7.95s/it]
77%|███████▋ | 65/84 [08:29<02:20, 7.41s/it]
79%|███████▊ | 66/84 [08:34<02:05, 6.95s/it]
80%|███████▉ | 67/84 [08:50<02:39, 9.40s/it]
81%|████████ | 68/84 [08:58<02:26, 9.18s/it]
82%|████████▏ | 69/84 [09:05<02:06, 8.43s/it]
83%|████████▎ | 70/84 [09:12<01:51, 7.96s/it]
85%|████████▍ | 71/84 [09:16<01:29, 6.87s/it]
86%|████████▌ | 72/84 [09:22<01:20, 6.71s/it]
87%|████████▋ | 73/84 [09:29<01:13, 6.64s/it]
88%|████████▊ | 74/84 [09:37<01:12, 7.24s/it]
89%|████████▉ | 75/84 [09:45<01:06, 7.35s/it]
90%|█████████ | 76/84 [09:54<01:02, 7.86s/it]
92%|█████████▏| 77/84 [09:59<00:49, 7.11s/it]
93%|█████████▎| 78/84 [10:06<00:42, 7.01s/it]
94%|█████████▍| 79/84 [10:12<00:32, 6.49s/it]
95%|█████████▌| 80/84 [10:19<00:27, 6.76s/it]
96%|█████████▋| 81/84 [10:26<00:20, 6.76s/it]
98%|█████████▊| 82/84 [10:49<00:23, 11.85s/it]
99%|█████████▉| 83/84 [10:56<00:10, 10.40s/it]
100%|██████████| 84/84 [11:01<00:00, 8.70s/it]
{'eval_loss': 0.0, 'eval_runtime': 669.8868, 'eval_samples_per_second': 0.746, 'eval_steps_per_second': 0.125, 'eval_num_tokens': 64360474.0, 'eval_completions/mean_length': 577.4623176938012, 'eval_completions/min_length': 264.57142857142856, 'eval_completions/max_length': 1090.095238095238, 'eval_completions/clipped_ratio': 0.00992063521629288, 'eval_completions/mean_terminated_length': 552.9960454305013, 'eval_completions/min_terminated_length': 264.57142857142856, 'eval_completions/max_terminated_length': 980.0952380952381, 'eval_rewards/reward_correctness/mean': 0.6746031773232278, 'eval_rewards/reward_correctness/std': 0.45623940797079177, 'eval_reward': 0.6746031936435473, 'eval_reward_std': 0.45623941009952906, 'eval_frac_reward_zero_std': 1.0, 'eval_sampling/sampling_logp_difference/mean': 0.05015929979050443, 'eval_sampling/sampling_logp_difference/max': 2.6629502290771123, 'eval_sampling/importance_sampling_ratio/min': 1.1791101893997088e-05, 'eval_sampling/importance_sampling_ratio/mean': 0.041787284431111835, 'eval_sampling/importance_sampling_ratio/max': 0.2080012608648325, 'eval_entropy': 0.20710848183149383, 'eval_clip_ratio/low_mean': 0.0, 'eval_clip_ratio/low_min': 0.0, 'eval_clip_ratio/high_mean': 0.0, 'eval_clip_ratio/high_max': 0.0, 'eval_clip_ratio/region_mean': 0.0, 'epoch': 0.74}
37%|███▋ | 50/136 [5:10:24<8:00:46, 335.42s/it]
100%|██████████| 84/84 [11:01<00:00, 8.70s/it]
[2026-08-13 03:40:57] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,14786)
38%|███▊ | 51/136 [5:15:31<12:27:40, 527.77s/it]
{'loss': -0.024, 'grad_norm': 0.0426987144482271, 'learning_rate': 2.2872776774781627e-06, 'num_tokens': 65872596.0, 'completions/mean_length': 901.7982177734375, 'completions/min_length': 141.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1354166716337204, 'completions/mean_terminated_length': 561.8870239257812, 'completions/min_terminated_length': 141.0, 'completions/max_terminated_length': 2179.0, 'rewards/reward_correctness/mean': 0.6282551884651184, 'rewards/reward_correctness/std': 0.48342812061309814, 'reward': 0.6282552480697632, 'reward_std': 0.48342809081077576, 'frac_reward_zero_std': 0.4296875, 'sampling/sampling_logp_difference/mean': 0.028055667877197266, 'sampling/sampling_logp_difference/max': 1.6584514379501343, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5335763096809387, 'sampling/importance_sampling_ratio/max': 2.981839895248413, 'entropy': 0.5385016986401752, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 300.4838323034346, 'epoch': 0.75}
38%|███▊ | 51/136 [5:15:31<12:27:40, 527.77s/it][2026-08-13 03:46:13] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,127)
[2026-08-13 03:46:14] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "frac" @(1,1)
Open LaTeX blocks:
@(1,1) arguments of macro "frac"
[2026-08-13 03:46:14] WARNING __init__.py:1389: WARNING: Error in configuration: macro '\frac' failed its substitution!
38%|███▊ | 52/136 [5:21:16<11:01:53, 472.78s/it]
{'loss': -0.0177, 'grad_norm': 0.04766427212983689, 'learning_rate': 2.258556242532317e-06, 'num_tokens': 67414141.0, 'completions/mean_length': 904.3053588867188, 'completions/min_length': 144.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1302083432674408, 'completions/mean_terminated_length': 580.3997192382812, 'completions/min_terminated_length': 144.0, 'completions/max_terminated_length': 2621.0, 'rewards/reward_correctness/mean': 0.57421875, 'rewards/reward_correctness/std': 0.4946219325065613, 'reward': 0.57421875, 'reward_std': 0.4946219325065613, 'frac_reward_zero_std': 0.5078125, 'sampling/sampling_logp_difference/mean': 0.04324951022863388, 'sampling/sampling_logp_difference/max': 32.999961853027344, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.48218855261802673, 'sampling/importance_sampling_ratio/max': 2.9827163219451904, 'entropy': 0.5088172309333459, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 343.32239248557016, 'epoch': 0.76}
38%|███▊ | 52/136 [5:21:16<11:01:53, 472.78s/it][2026-08-13 03:52:02] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,1)
[2026-08-13 03:52:02] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Bad \end macro: expected {environmentname} @(1,22)
[2026-08-13 03:52:02] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,87)
[2026-08-13 03:52:08] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "frac" @(1,1)
Open LaTeX blocks:
@(1,1) arguments of macro "frac"
[2026-08-13 03:52:08] WARNING __init__.py:1389: WARNING: Error in configuration: macro '\frac' failed its substitution!
39%|███▉ | 53/136 [5:26:38<9:51:28, 427.58s/it]
{'loss': -0.0349, 'grad_norm': 0.03832772147605136, 'learning_rate': 2.2294848325548066e-06, 'num_tokens': 69140778.0, 'completions/mean_length': 1033.1204833984375, 'completions/min_length': 122.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1686197966337204, 'completions/mean_terminated_length': 619.5967407226562, 'completions/min_terminated_length': 122.0, 'completions/max_terminated_length': 2094.0, 'rewards/reward_correctness/mean': 0.5305989384651184, 'rewards/reward_correctness/std': 0.499225378036499, 'reward': 0.5305989980697632, 'reward_std': 0.49922534823417664, 'frac_reward_zero_std': 0.4296875, 'sampling/sampling_logp_difference/mean': 0.02942274883389473, 'sampling/sampling_logp_difference/max': 1.1630563735961914, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.4914858043193817, 'sampling/importance_sampling_ratio/max': 2.947634696960449, 'entropy': 0.5954927550628781, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 320.89451397629455, 'epoch': 0.78}
39%|███▉ | 53/136 [5:26:38<9:51:28, 427.58s/it][2026-08-13 03:57:18] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,23)
[2026-08-13 03:57:18] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "frac" @(1,1)
Open LaTeX blocks:
@(1,1) arguments of macro "frac"
[2026-08-13 03:57:18] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,3044)
[2026-08-13 03:57:18] WARNING __init__.py:1389: WARNING: Error in configuration: macro '\frac' failed its substitution!
40%|███▉ | 54/136 [5:31:53<8:58:16, 393.86s/it]
{'loss': -0.0314, 'grad_norm': 0.05567566279138782, 'learning_rate': 2.200080166242961e-06, 'num_tokens': 70557443.0, 'completions/mean_length': 832.9329833984375, 'completions/min_length': 142.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1139322966337204, 'completions/mean_terminated_length': 545.0293579101562, 'completions/min_terminated_length': 142.0, 'completions/max_terminated_length': 2290.0, 'rewards/reward_correctness/mean': 0.6595051884651184, 'rewards/reward_correctness/std': 0.4740299582481384, 'reward': 0.6595052480697632, 'reward_std': 0.47402989864349365, 'frac_reward_zero_std': 0.453125, 'sampling/sampling_logp_difference/mean': 0.02493918500840664, 'sampling/sampling_logp_difference/max': 4.315248012542725, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.576991856098175, 'sampling/importance_sampling_ratio/max': 2.9649579524993896, 'entropy': 0.45392935909330845, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 313.9590461105108, 'epoch': 0.79}
40%|███▉ | 54/136 [5:31:53<8:58:16, 393.86s/it]
40%|████ | 55/136 [5:37:36<8:31:02, 378.55s/it]
{'loss': -0.0246, 'grad_norm': 0.04392399703168934, 'learning_rate': 2.1703591539467283e-06, 'num_tokens': 72337356.0, 'completions/mean_length': 1055.672607421875, 'completions/min_length': 149.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.18359375, 'completions/mean_terminated_length': 602.5557861328125, 'completions/min_terminated_length': 149.0, 'completions/max_terminated_length': 2932.0, 'rewards/reward_correctness/mean': 0.5475260615348816, 'rewards/reward_correctness/std': 0.49789825081825256, 'reward': 0.5475260615348816, 'reward_std': 0.49789825081825256, 'frac_reward_zero_std': 0.5, 'sampling/sampling_logp_difference/mean': 0.029556401073932648, 'sampling/sampling_logp_difference/max': 0.9878766536712646, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.512884795665741, 'sampling/importance_sampling_ratio/max': 2.961296796798706, 'entropy': 0.6210943238111213, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 341.59292052686214, 'epoch': 0.81}
40%|████ | 55/136 [5:37:36<8:31:02, 378.55s/it][2026-08-13 04:08:22] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,6)
[2026-08-13 04:08:23] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,2)
[2026-08-13 04:08:23] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,100)
[2026-08-13 04:08:28] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\)' @(1,1)
41%|████ | 56/136 [5:43:07<8:05:58, 364.48s/it]
{'loss': -0.0141, 'grad_norm': 0.04530116895001296, 'learning_rate': 2.140338887943686e-06, 'num_tokens': 73948192.0, 'completions/mean_length': 957.0729370117188, 'completions/min_length': 130.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.154296875, 'completions/mean_terminated_length': 571.8860473632812, 'completions/min_terminated_length': 130.0, 'completions/max_terminated_length': 2412.0, 'rewards/reward_correctness/mean': 0.6041666865348816, 'rewards/reward_correctness/std': 0.48918819427490234, 'reward': 0.6041666865348816, 'reward_std': 0.48918819427490234, 'frac_reward_zero_std': 0.5, 'sampling/sampling_logp_difference/mean': 0.029566694051027298, 'sampling/sampling_logp_difference/max': 1.0761712789535522, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5680562853813171, 'sampling/importance_sampling_ratio/max': 2.985471487045288, 'entropy': 0.5751702012494206, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 330.4778544139117, 'epoch': 0.82}
41%|████ | 56/136 [5:43:07<8:05:58, 364.48s/it]
42%|████▏ | 57/136 [5:48:21<7:39:52, 349.27s/it]
{'loss': -0.0408, 'grad_norm': 0.04280765723311196, 'learning_rate': 2.110036632609435e-06, 'num_tokens': 75445696.0, 'completions/mean_length': 892.671875, 'completions/min_length': 104.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1302083432674408, 'completions/mean_terminated_length': 566.4251708984375, 'completions/min_terminated_length': 104.0, 'completions/max_terminated_length': 2915.0, 'rewards/reward_correctness/mean': 0.6490885615348816, 'rewards/reward_correctness/std': 0.4774107038974762, 'reward': 0.6490885615348816, 'reward_std': 0.4774106740951538, 'frac_reward_zero_std': 0.4453125, 'sampling/sampling_logp_difference/mean': 0.02581387758255005, 'sampling/sampling_logp_difference/max': 2.199824333190918, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5869173407554626, 'sampling/importance_sampling_ratio/max': 2.9966816902160645, 'entropy': 0.49189388286322355, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 312.6213083774783, 'epoch': 0.84}
42%|████▏ | 57/136 [5:48:21<7:39:52, 349.27s/it][2026-08-13 04:19:05] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,6)
[2026-08-13 04:19:10] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "frac" @(1,41)
Open LaTeX blocks:
@(1,41) arguments of macro "frac"
[2026-08-13 04:19:10] WARNING __init__.py:1389: WARNING: Error in configuration: macro '\frac' failed its substitution!
43%|████▎ | 58/136 [5:54:10<7:34:02, 349.27s/it]
{'loss': -0.045, 'grad_norm': 0.04534803213910246, 'learning_rate': 2.0794698144890156e-06, 'num_tokens': 77166719.0, 'completions/mean_length': 1012.7077026367188, 'completions/min_length': 134.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1666666716337204, 'completions/mean_terminated_length': 606.9664306640625, 'completions/min_terminated_length': 134.0, 'completions/max_terminated_length': 3065.0, 'rewards/reward_correctness/mean': 0.5846354365348816, 'rewards/reward_correctness/std': 0.49294528365135193, 'reward': 0.5846354365348816, 'reward_std': 0.49294525384902954, 'frac_reward_zero_std': 0.4921875, 'sampling/sampling_logp_difference/mean': 0.04774170368909836, 'sampling/sampling_logp_difference/max': 21.9999942779541, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.44743868708610535, 'sampling/importance_sampling_ratio/max': 2.991367816925049, 'entropy': 0.609487400855869, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 348.07748995441943, 'epoch': 0.85}
43%|████▎ | 58/136 [5:54:10<7:34:02, 349.27s/it][2026-08-13 04:25:16] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,0)
[2026-08-13 04:25:17] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "frac" @(1,2)
Open LaTeX blocks:
@(1,2) arguments of macro "frac"
[2026-08-13 04:25:17] WARNING __init__.py:1389: WARNING: Error in configuration: macro '\frac' failed its substitution!
[2026-08-13 04:25:18] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,2)
[2026-08-13 04:25:18] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\)' @(1,1)
[2026-08-13 04:25:19] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,0)
43%|████▎ | 59/136 [6:00:04<7:30:05, 350.73s/it]
{'loss': -0.0201, 'grad_norm': 0.04810753389214618, 'learning_rate': 2.048656012275064e-06, 'num_tokens': 78909004.0, 'completions/mean_length': 1030.2064208984375, 'completions/min_length': 157.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1751302182674408, 'completions/mean_terminated_length': 597.9454956054688, 'completions/min_terminated_length': 157.0, 'completions/max_terminated_length': 2943.0, 'rewards/reward_correctness/mean': 0.5247395634651184, 'rewards/reward_correctness/std': 0.4995502233505249, 'reward': 0.5247396230697632, 'reward_std': 0.4995501935482025, 'frac_reward_zero_std': 0.4609375, 'sampling/sampling_logp_difference/mean': 0.030788302421569824, 'sampling/sampling_logp_difference/max': 1.0024757385253906, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5370829105377197, 'sampling/importance_sampling_ratio/max': 2.9893922805786133, 'entropy': 0.6143204279942438, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 352.94321115035564, 'epoch': 0.87}
43%|████▎ | 59/136 [6:00:04<7:30:05, 350.73s/it]
44%|████▍ | 60/136 [6:06:03<7:27:16, 353.11s/it]
{'loss': -0.0359, 'grad_norm': 0.058038002186570306, 'learning_rate': 2.017612946698471e-06, 'num_tokens': 80753634.0, 'completions/mean_length': 1106.5794677734375, 'completions/min_length': 111.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1979166716337204, 'completions/mean_terminated_length': 621.6055297851562, 'completions/min_terminated_length': 111.0, 'completions/max_terminated_length': 2871.0, 'rewards/reward_correctness/mean': 0.5325520634651184, 'rewards/reward_correctness/std': 0.4991017282009125, 'reward': 0.5325521230697632, 'reward_std': 0.4991016983985901, 'frac_reward_zero_std': 0.5, 'sampling/sampling_logp_difference/mean': 0.030230939388275146, 'sampling/sampling_logp_difference/max': 1.1060563325881958, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.525791347026825, 'sampling/importance_sampling_ratio/max': 2.9939475059509277, 'entropy': 0.6316488899756223, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 357.5172311188653, 'epoch': 0.88}
44%|████▍ | 60/136 [6:06:03<7:27:16, 353.11s/it]
0%| | 0/84 [00:00<?, ?it/s]
2%|▏ | 2/84 [00:05<03:39, 2.68s/it]
4%|▎ | 3/84 [00:10<04:54, 3.64s/it]
5%|▍ | 4/84 [00:16<06:10, 4.63s/it]
6%|▌ | 5/84 [00:25<07:59, 6.07s/it]
7%|▋ | 6/84 [00:32<08:21, 6.43s/it]
8%|▊ | 7/84 [00:56<15:27, 12.05s/it]
10%|▉ | 8/84 [01:04<13:38, 10.76s/it]
11%|█ | 9/84 [01:11<12:08, 9.71s/it]
12%|█▏ | 10/84 [01:18<10:46, 8.73s/it]
13%|█▎ | 11/84 [01:25<09:52, 8.12s/it]
14%|█▍ | 12/84 [01:31<09:13, 7.69s/it]
15%|█▌ | 13/84 [01:39<09:14, 7.81s/it]
17%|█▋ | 14/84 [01:46<08:44, 7.49s/it]
18%|█▊ | 15/84 [01:57<09:45, 8.49s/it]
19%|█▉ | 16/84 [02:04<09:10, 8.10s/it]
20%|██ | 17/84 [02:11<08:35, 7.69s/it]
21%|██▏ | 18/84 [02:20<08:47, 7.99s/it]
23%|██▎ | 19/84 [02:26<08:05, 7.47s/it]
24%|██▍ | 20/84 [02:35<08:28, 7.94s/it]
25%|██▌ | 21/84 [02:41<07:48, 7.44s/it]
26%|██▌ | 22/84 [02:46<06:57, 6.74s/it]
27%|██▋ | 23/84 [02:53<06:59, 6.88s/it]
29%|██▊ | 24/84 [03:01<07:02, 7.03s/it]
30%|██▉ | 25/84 [03:05<05:57, 6.07s/it]
31%|███ | 26/84 [03:28<10:57, 11.33s/it]
32%|███▏ | 27/84 [03:52<14:20, 15.10s/it]
33%|███▎ | 28/84 [03:57<11:18, 12.11s/it]
35%|███▍ | 29/84 [04:04<09:34, 10.44s/it]
36%|███▌ | 30/84 [04:11<08:34, 9.53s/it]
37%|███▋ | 31/84 [04:17<07:20, 8.30s/it]
38%|███▊ | 32/84 [04:21<06:09, 7.11s/it]
39%|███▉ | 33/84 [04:45<10:14, 12.05s/it]
40%|████ | 34/84 [05:08<12:57, 15.54s/it]
42%|████▏ | 35/84 [05:32<14:45, 18.08s/it]
43%|████▎ | 36/84 [05:40<11:57, 14.95s/it]
44%|████▍ | 37/84 [05:46<09:38, 12.32s/it]
45%|████▌ | 38/84 [05:53<08:07, 10.59s/it]
46%|████▋ | 39/84 [06:16<10:52, 14.51s/it]
48%|████▊ | 40/84 [06:40<12:41, 17.31s/it]
49%|████▉ | 41/84 [06:46<09:58, 13.92s/it]
50%|█████ | 42/84 [06:54<08:33, 12.23s/it]
51%|█████ | 43/84 [07:02<07:20, 10.74s/it]
52%|█████▏ | 44/84 [07:06<05:50, 8.76s/it]
54%|█████▎ | 45/84 [07:13<05:21, 8.23s/it]
55%|█████▍ | 46/84 [07:36<08:07, 12.84s/it]
56%|█████▌ | 47/84 [07:43<06:44, 10.94s/it]
57%|█████▋ | 48/84 [07:50<05:54, 9.84s/it]
58%|█████▊ | 49/84 [07:59<05:35, 9.60s/it]
60%|█████▉ | 50/84 [08:23<07:50, 13.85s/it]
61%|██████ | 51/84 [08:29<06:18, 11.48s/it]
62%|██████▏ | 52/84 [08:37<05:30, 10.34s/it]
63%|██████▎ | 53/84 [08:42<04:36, 8.93s/it]
64%|██████▍ | 54/84 [08:46<03:37, 7.25s/it]
65%|██████▌ | 55/84 [08:52<03:21, 6.95s/it]
67%|██████▋ | 56/84 [08:55<02:46, 5.95s/it]
68%|██████▊ | 57/84 [09:08<03:31, 7.83s/it]
69%|██████▉ | 58/84 [09:12<02:58, 6.85s/it]
70%|███████ | 59/84 [09:22<03:13, 7.74s/it]
71%|███████▏ | 60/84 [09:27<02:43, 6.83s/it]
73%|███████▎ | 61/84 [09:32<02:25, 6.34s/it]
74%|███████▍ | 62/84 [09:40<02:31, 6.89s/it]
75%|███████▌ | 63/84 [09:47<02:25, 6.91s/it]
76%|███████▌ | 64/84 [09:56<02:29, 7.46s/it]
77%|███████▋ | 65/84 [10:20<03:56, 12.46s/it]
79%|███████▊ | 66/84 [10:27<03:12, 10.70s/it]
80%|███████▉ | 67/84 [10:50<04:07, 14.55s/it]
81%|████████ | 68/84 [11:03<03:43, 13.95s/it]
82%|████████▏ | 69/84 [11:09<02:54, 11.60s/it]
83%|████████▎ | 70/84 [11:18<02:33, 10.97s/it]
85%|████████▍ | 71/84 [11:23<01:59, 9.18s/it]
86%|████████▌ | 72/84 [11:30<01:42, 8.57s/it]
87%|████████▋ | 73/84 [11:36<01:24, 7.70s/it]
88%|████████▊ | 74/84 [12:00<02:05, 12.51s/it]
89%|████████▉ | 75/84 [12:07<01:38, 10.96s/it]
90%|█████████ | 76/84 [12:16<01:23, 10.47s/it]
92%|█████████▏| 77/84 [12:24<01:07, 9.67s/it]
93%|█████████▎| 78/84 [12:34<00:57, 9.55s/it]
94%|█████████▍| 79/84 [12:57<01:08, 13.79s/it]
95%|█████████▌| 80/84 [13:07<00:50, 12.72s/it]
96%|█████████▋| 81/84 [13:14<00:32, 10.77s/it]
98%|█████████▊| 82/84 [13:25<00:21, 10.81s/it]
99%|█████████▉| 83/84 [13:48<00:14, 14.67s/it]
100%|██████████| 84/84 [14:12<00:00, 17.35s/it]
{'eval_loss': 0.0, 'eval_runtime': 858.7508, 'eval_samples_per_second': 0.582, 'eval_steps_per_second': 0.098, 'eval_num_tokens': 80753634.0, 'eval_completions/mean_length': 636.5297782534644, 'eval_completions/min_length': 268.7857142857143, 'eval_completions/max_length': 1374.0119047619048, 'eval_completions/clipped_ratio': 0.033730159735395795, 'eval_completions/mean_terminated_length': 549.8853305634998, 'eval_completions/min_terminated_length': 268.7857142857143, 'eval_completions/max_terminated_length': 942.7857142857143, 'eval_rewards/reward_correctness/mean': 0.700396828353405, 'eval_rewards/reward_correctness/std': 0.4480584930805933, 'eval_reward': 0.7003968453833035, 'eval_reward_std': 0.4480584969832784, 'eval_frac_reward_zero_std': 1.0, 'eval_sampling/sampling_logp_difference/mean': 0.048185367846772784, 'eval_sampling/sampling_logp_difference/max': 2.5949947152818953, 'eval_sampling/importance_sampling_ratio/min': 6.596829363872345e-05, 'eval_sampling/importance_sampling_ratio/mean': 0.042331399595457786, 'eval_sampling/importance_sampling_ratio/max': 0.20012787568356588, 'eval_entropy': 0.20073904212386834, 'eval_clip_ratio/low_mean': 0.0, 'eval_clip_ratio/low_min': 0.0, 'eval_clip_ratio/high_mean': 0.0, 'eval_clip_ratio/high_max': 0.0, 'eval_clip_ratio/region_mean': 0.0, 'epoch': 0.88}
44%|████▍ | 60/136 [6:20:22<7:27:16, 353.11s/it]
100%|██████████| 84/84 [14:12<00:00, 17.35s/it]

45%|████▍ | 61/136 [6:26:00<12:37:48, 606.24s/it]
{'loss': -0.0424, 'grad_norm': 0.054576992161028744, 'learning_rate': 1.9863584703373534e-06, 'num_tokens': 82474254.0, 'completions/mean_length': 1015.671875, 'completions/min_length': 150.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1731770932674408, 'completions/mean_terminated_length': 585.7787475585938, 'completions/min_terminated_length': 150.0, 'completions/max_terminated_length': 2855.0, 'rewards/reward_correctness/mean': 0.5514323115348816, 'rewards/reward_correctness/std': 0.497509628534317, 'reward': 0.5514323115348816, 'reward_std': 0.497509628534317, 'frac_reward_zero_std': 0.4375, 'sampling/sampling_logp_difference/mean': 0.030388731509447098, 'sampling/sampling_logp_difference/max': 5.065879821777344, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5062848925590515, 'sampling/importance_sampling_ratio/max': 2.995573043823242, 'entropy': 0.6050877087982371, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 331.8993032807484, 'epoch': 0.9}
45%|████▍ | 61/136 [6:26:00<12:37:48, 606.24s/it][2026-08-13 04:56:58] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,147)
46%|████▌ | 62/136 [6:31:42<10:50:02, 527.07s/it]
{'loss': -0.0448, 'grad_norm': 0.05019150172655544, 'learning_rate': 1.954910557350202e-06, 'num_tokens': 84280817.0, 'completions/mean_length': 1073.9447021484375, 'completions/min_length': 131.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1725260466337204, 'completions/mean_terminated_length': 658.0574340820312, 'completions/min_terminated_length': 131.0, 'completions/max_terminated_length': 2556.0, 'rewards/reward_correctness/mean': 0.6087239384651184, 'rewards/reward_correctness/std': 0.48819491267204285, 'reward': 0.6087239980697632, 'reward_std': 0.48819488286972046, 'frac_reward_zero_std': 0.390625, 'sampling/sampling_logp_difference/mean': 0.028018062934279442, 'sampling/sampling_logp_difference/max': 1.4889273643493652, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5425732731819153, 'sampling/importance_sampling_ratio/max': 2.9991607666015625, 'entropy': 0.5703155723167583, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 341.14307256788015, 'epoch': 0.91}
46%|████▌ | 62/136 [6:31:42<10:50:02, 527.07s/it][2026-08-13 05:02:26] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "frac" @(1,1)
Open LaTeX blocks:
@(1,1) arguments of macro "frac"
[2026-08-13 05:02:26] WARNING __init__.py:1389: WARNING: Error in configuration: macro '\frac' failed its substitution!
46%|████▋ | 63/136 [6:37:07<9:27:27, 466.41s/it]
{'loss': -0.028, 'grad_norm': 0.0804020292899743, 'learning_rate': 1.9232872931391114e-06, 'num_tokens': 86030823.0, 'completions/mean_length': 1036.990966796875, 'completions/min_length': 161.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1653645932674408, 'completions/mean_terminated_length': 633.828369140625, 'completions/min_terminated_length': 161.0, 'completions/max_terminated_length': 2430.0, 'rewards/reward_correctness/mean': 0.4895833432674408, 'rewards/reward_correctness/std': 0.5000542402267456, 'reward': 0.4895833432674408, 'reward_std': 0.5000542402267456, 'frac_reward_zero_std': 0.3828125, 'sampling/sampling_logp_difference/mean': 0.030840862542390823, 'sampling/sampling_logp_difference/max': 1.5172169208526611, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.527662456035614, 'sampling/importance_sampling_ratio/max': 2.9987845420837402, 'entropy': 0.6102959946729243, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 323.75026592426, 'epoch': 0.93}
46%|████▋ | 63/136 [6:37:07<9:27:27, 466.41s/it][2026-08-13 05:07:52] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "frac" @(1,2)
Open LaTeX blocks:
@(1,2) arguments of macro "frac"
[2026-08-13 05:07:52] WARNING __init__.py:1389: WARNING: Error in configuration: macro '\frac' failed its substitution!
[2026-08-13 05:07:54] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,362)
[2026-08-13 05:07:55] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,0)
47%|████▋ | 64/136 [6:42:48<8:34:30, 428.75s/it]
{'loss': -0.0213, 'grad_norm': 0.0563704880498028, 'learning_rate': 1.8915068639490344e-06, 'num_tokens': 87680029.0, 'completions/mean_length': 976.2330932617188, 'completions/min_length': 100.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.154296875, 'completions/mean_terminated_length': 595.9468383789062, 'completions/min_terminated_length': 100.0, 'completions/max_terminated_length': 2693.0, 'rewards/reward_correctness/mean': 0.5852864384651184, 'rewards/reward_correctness/std': 0.4928329586982727, 'reward': 0.5852864980697632, 'reward_std': 0.4928329586982727, 'frac_reward_zero_std': 0.5, 'sampling/sampling_logp_difference/mean': 0.04849446564912796, 'sampling/sampling_logp_difference/max': 18.054588317871094, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.4369260370731354, 'sampling/importance_sampling_ratio/max': 2.9954569339752197, 'entropy': 0.5973206992493942, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 339.74727647705004, 'epoch': 0.94}
47%|████▋ | 64/136 [6:42:48<8:34:30, 428.75s/it][2026-08-13 05:13:37] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,663)
48%|████▊ | 65/136 [6:48:20<7:52:48, 399.56s/it]
{'loss': -0.0417, 'grad_norm': 0.04591077537993929, 'learning_rate': 1.8595875464090389e-06, 'num_tokens': 89316793.0, 'completions/mean_length': 966.34375, 'completions/min_length': 155.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.15234375, 'completions/mean_terminated_length': 588.47314453125, 'completions/min_terminated_length': 155.0, 'completions/max_terminated_length': 2495.0, 'rewards/reward_correctness/mean': 0.6009114384651184, 'rewards/reward_correctness/std': 0.489870548248291, 'reward': 0.6009114980697632, 'reward_std': 0.48987048864364624, 'frac_reward_zero_std': 0.4140625, 'sampling/sampling_logp_difference/mean': 0.03166817128658295, 'sampling/sampling_logp_difference/max': 1.4514122009277344, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.4967881739139557, 'sampling/importance_sampling_ratio/max': 2.9966180324554443, 'entropy': 0.6048457206925377, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 330.3027874575928, 'epoch': 0.96}
48%|████▊ | 65/136 [6:48:20<7:52:48, 399.56s/it][2026-08-13 05:18:52] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,345)
[2026-08-13 05:18:52] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,416)
[2026-08-13 05:18:52] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,494)
[2026-08-13 05:18:54] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Bad \begin macro: expected {environmentname} @(1,0)
49%|████▊ | 66/136 [6:53:33<7:16:02, 373.75s/it]
{'loss': -0.0406, 'grad_norm': 0.06999161224416137, 'learning_rate': 1.8275476970215906e-06, 'num_tokens': 90877761.0, 'completions/mean_length': 915.1068115234375, 'completions/min_length': 125.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.126953125, 'completions/mean_terminated_length': 601.72705078125, 'completions/min_terminated_length': 125.0, 'completions/max_terminated_length': 2902.0, 'rewards/reward_correctness/mean': 0.5625, 'rewards/reward_correctness/std': 0.49623993039131165, 'reward': 0.5625, 'reward_std': 0.49623993039131165, 'frac_reward_zero_std': 0.46875, 'sampling/sampling_logp_difference/mean': 0.02776806429028511, 'sampling/sampling_logp_difference/max': 1.640268325805664, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5559518337249756, 'sampling/importance_sampling_ratio/max': 2.9245617389678955, 'entropy': 0.5254314645426348, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 312.4176756278612, 'epoch': 0.97}
49%|████▊ | 66/136 [6:53:33<7:16:02, 373.75s/it][2026-08-13 05:24:05] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,1)
[2026-08-13 05:24:05] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Bad \end macro: expected {environmentname} @(1,29)
49%|████▉ | 67/136 [6:58:40<6:46:53, 353.82s/it]
{'loss': -0.0278, 'grad_norm': 0.05063560268720398, 'learning_rate': 1.7954057416059002e-06, 'num_tokens': 92443733.0, 'completions/mean_length': 934.9114990234375, 'completions/min_length': 137.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1256510466337204, 'completions/mean_terminated_length': 627.7944946289062, 'completions/min_terminated_length': 137.0, 'completions/max_terminated_length': 2985.0, 'rewards/reward_correctness/mean': 0.5774739384651184, 'rewards/reward_correctness/std': 0.4941222071647644, 'reward': 0.5774739980697632, 'reward_std': 0.49412214756011963, 'frac_reward_zero_std': 0.484375, 'sampling/sampling_logp_difference/mean': 0.027582859620451927, 'sampling/sampling_logp_difference/max': 1.3922834396362305, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.522100031375885, 'sampling/importance_sampling_ratio/max': 2.982328176498413, 'entropy': 0.5442105032270774, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 306.1620469666086, 'epoch': 0.99}
49%|████▉ | 67/136 [6:58:40<6:46:53, 353.82s/it][2026-08-13 05:29:05] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,8131)
50%|█████ | 68/136 [7:03:37<6:21:37, 336.72s/it]
{'loss': -0.0278, 'grad_norm': 0.05421093036584026, 'learning_rate': 1.7631801647014034e-06, 'num_tokens': 93931514.0, 'completions/mean_length': 878.169921875, 'completions/min_length': 166.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1178385466337204, 'completions/mean_terminated_length': 585.1195678710938, 'completions/min_terminated_length': 166.0, 'completions/max_terminated_length': 2569.0, 'rewards/reward_correctness/mean': 0.6119791865348816, 'rewards/reward_correctness/std': 0.4874580502510071, 'reward': 0.6119791865348816, 'reward_std': 0.4874580502510071, 'frac_reward_zero_std': 0.4453125, 'sampling/sampling_logp_difference/mean': 0.02798476442694664, 'sampling/sampling_logp_difference/max': 1.140676498413086, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5653523802757263, 'sampling/importance_sampling_ratio/max': 2.9934051036834717, 'entropy': 0.5163884535431862, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 295.66030500503257, 'epoch': 1.0}
50%|█████ | 68/136 [7:03:37<6:21:37, 336.72s/it][2026-08-13 05:34:28] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "frac" @(1,3)
Open LaTeX blocks:
@(1,3) arguments of macro "frac"
[2026-08-13 05:34:28] WARNING __init__.py:1389: WARNING: Error in configuration: macro '\frac' failed its substitution!
[2026-08-13 05:34:30] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "frac" @(1,2)
Open LaTeX blocks:
@(1,2) arguments of macro "frac"
[2026-08-13 05:34:30] WARNING __init__.py:1389: WARNING: Error in configuration: macro '\frac' failed its substitution!
[2026-08-13 05:34:30] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,13)
51%|█████ | 69/136 [7:09:12<6:15:29, 336.26s/it]
{'loss': -0.0366, 'grad_norm': 0.050663113215641147, 'learning_rate': 1.7308894989374766e-06, 'num_tokens': 95572970.0, 'completions/mean_length': 971.5078125, 'completions/min_length': 152.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1484375, 'completions/mean_terminated_length': 605.4052124023438, 'completions/min_terminated_length': 152.0, 'completions/max_terminated_length': 2770.0, 'rewards/reward_correctness/mean': 0.5735676884651184, 'rewards/reward_correctness/std': 0.4947192370891571, 'reward': 0.5735677480697632, 'reward_std': 0.4947192072868347, 'frac_reward_zero_std': 0.53125, 'sampling/sampling_logp_difference/mean': 0.029270043596625328, 'sampling/sampling_logp_difference/max': 0.9864826202392578, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5567998290061951, 'sampling/importance_sampling_ratio/max': 2.9747977256774902, 'entropy': 0.5773363119224086, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 333.99876602645963, 'epoch': 1.01}
51%|█████ | 69/136 [7:09:12<6:15:29, 336.26s/it]
51%|█████▏ | 70/136 [7:14:13<5:57:58, 325.43s/it]
{'loss': -0.0316, 'grad_norm': 0.057038639174071336, 'learning_rate': 1.6985523143754952e-06, 'num_tokens': 97092444.0, 'completions/mean_length': 905.9362182617188, 'completions/min_length': 121.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1217447966337204, 'completions/mean_terminated_length': 605.673828125, 'completions/min_terminated_length': 121.0, 'completions/max_terminated_length': 2845.0, 'rewards/reward_correctness/mean': 0.6282551884651184, 'rewards/reward_correctness/std': 0.48342812061309814, 'reward': 0.6282552480697632, 'reward_std': 0.48342809081077576, 'frac_reward_zero_std': 0.46875, 'sampling/sampling_logp_difference/mean': 0.027136720716953278, 'sampling/sampling_logp_difference/max': 1.991262435913086, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5966787934303284, 'sampling/importance_sampling_ratio/max': 2.9838263988494873, 'entropy': 0.5115884636761621, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 298.9868591595441, 'epoch': 1.03}
51%|█████▏ | 70/136 [7:14:13<5:57:58, 325.43s/it]
0%| | 0/84 [00:00<?, ?it/s]
2%|▏ | 2/84 [00:07<05:09, 3.78s/it]
4%|▎ | 3/84 [00:11<05:22, 3.98s/it]
5%|▍ | 4/84 [00:18<06:24, 4.81s/it]
6%|▌ | 5/84 [00:24<07:18, 5.55s/it]
7%|▋ | 6/84 [00:34<08:43, 6.71s/it]
8%|▊ | 7/84 [00:42<09:16, 7.23s/it]
10%|▉ | 8/84 [00:47<08:20, 6.59s/it]
11%|█ | 9/84 [00:55<08:47, 7.04s/it]
12%|█▏ | 10/84 [01:01<08:16, 6.71s/it]
13%|█▎ | 11/84 [01:07<07:47, 6.40s/it]
14%|█▍ | 12/84 [01:13<07:30, 6.26s/it]
15%|█▌ | 13/84 [01:29<11:00, 9.31s/it]
17%|█▋ | 14/84 [01:37<10:18, 8.84s/it]
18%|█▊ | 15/84 [01:44<09:34, 8.32s/it]
19%|█▉ | 16/84 [01:54<09:53, 8.73s/it]
20%|██ | 17/84 [02:00<09:06, 8.16s/it]
21%|██▏ | 18/84 [02:06<08:08, 7.40s/it]
23%|██▎ | 19/84 [02:12<07:40, 7.09s/it]
24%|██▍ | 20/84 [02:20<07:39, 7.18s/it]
25%|██▌ | 21/84 [02:29<08:11, 7.80s/it]
26%|██▌ | 22/84 [02:34<07:15, 7.03s/it]
27%|██▋ | 23/84 [02:39<06:30, 6.39s/it]
29%|██▊ | 24/84 [02:47<06:56, 6.95s/it]
30%|██▉ | 25/84 [02:51<05:57, 6.07s/it]
31%|███ | 26/84 [02:57<05:45, 5.95s/it]
32%|███▏ | 27/84 [03:21<10:46, 11.33s/it]
33%|███▎ | 28/84 [03:26<08:44, 9.37s/it]
35%|███▍ | 29/84 [03:32<07:41, 8.40s/it]
36%|███▌ | 30/84 [03:39<07:07, 7.91s/it]
37%|███▋ | 31/84 [03:44<06:25, 7.27s/it]
38%|███▊ | 32/84 [03:49<05:32, 6.40s/it]
39%|███▉ | 33/84 [04:00<06:40, 7.86s/it]
40%|████ | 34/84 [04:11<07:14, 8.70s/it]
42%|████▏ | 35/84 [04:35<10:48, 13.24s/it]
43%|████▎ | 36/84 [04:42<09:12, 11.51s/it]
44%|████▍ | 37/84 [04:47<07:24, 9.47s/it]
45%|████▌ | 38/84 [04:53<06:34, 8.59s/it]
46%|████▋ | 39/84 [05:04<06:50, 9.11s/it]
48%|████▊ | 40/84 [05:27<09:55, 13.54s/it]
49%|████▉ | 41/84 [05:35<08:18, 11.60s/it]
50%|█████ | 42/84 [05:42<07:20, 10.50s/it]
51%|█████ | 43/84 [05:49<06:26, 9.42s/it]
52%|█████▏ | 44/84 [05:54<05:23, 8.09s/it]
54%|█████▎ | 45/84 [06:02<05:09, 7.94s/it]
55%|█████▍ | 46/84 [06:09<04:47, 7.56s/it]
56%|█████▌ | 47/84 [06:32<07:39, 12.43s/it]
57%|█████▋ | 48/84 [06:42<06:54, 11.51s/it]
58%|█████▊ | 49/84 [06:48<05:47, 9.92s/it]
60%|█████▉ | 50/84 [06:56<05:14, 9.26s/it]
61%|██████ | 51/84 [07:05<05:08, 9.35s/it]
62%|██████▏ | 52/84 [07:13<04:42, 8.82s/it]
63%|██████▎ | 53/84 [07:16<03:45, 7.26s/it]
64%|██████▍ | 54/84 [07:20<03:01, 6.05s/it]
65%|██████▌ | 55/84 [07:26<03:00, 6.21s/it]
67%|██████▋ | 56/84 [07:30<02:33, 5.49s/it]
68%|██████▊ | 57/84 [07:54<04:57, 11.03s/it]
69%|██████▉ | 58/84 [07:59<03:55, 9.05s/it]
70%|███████ | 59/84 [08:06<03:33, 8.52s/it]
71%|███████▏ | 60/84 [08:11<03:03, 7.63s/it]
73%|███████▎ | 61/84 [08:20<03:02, 7.92s/it]
74%|███████▍ | 62/84 [08:28<02:55, 7.97s/it][2026-08-13 05:51:54] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "frac" @(1,1)
Open LaTeX blocks:
@(1,1) arguments of macro "frac"
[2026-08-13 05:51:54] WARNING __init__.py:1389: WARNING: Error in configuration: macro '\frac' failed its substitution!
75%|███████▌ | 63/84 [08:52<04:26, 12.68s/it]
76%|███████▌ | 64/84 [09:01<03:56, 11.81s/it]
77%|███████▋ | 65/84 [09:26<04:54, 15.50s/it]
79%|███████▊ | 66/84 [09:30<03:41, 12.31s/it]
80%|███████▉ | 67/84 [09:43<03:29, 12.33s/it]
81%|████████ | 68/84 [09:53<03:06, 11.65s/it]
82%|████████▏ | 69/84 [09:59<02:31, 10.08s/it]
83%|████████▎ | 70/84 [10:05<02:01, 8.71s/it]
85%|████████▍ | 71/84 [10:11<01:41, 7.82s/it]
86%|████████▌ | 72/84 [10:17<01:28, 7.37s/it]
87%|████████▋ | 73/84 [10:26<01:28, 8.04s/it]
88%|████████▊ | 74/84 [10:34<01:17, 7.77s/it]
89%|████████▉ | 75/84 [10:40<01:05, 7.24s/it]
90%|█████████ | 76/84 [10:50<01:05, 8.19s/it]
92%|█████████▏| 77/84 [10:57<00:54, 7.73s/it]
93%|█████████▎| 78/84 [11:03<00:43, 7.27s/it]
94%|█████████▍| 79/84 [11:09<00:35, 7.07s/it]
95%|█████████▌| 80/84 [11:19<00:31, 7.91s/it]
96%|█████████▋| 81/84 [11:27<00:23, 7.72s/it]
98%|█████████▊| 82/84 [11:33<00:14, 7.36s/it]
99%|█████████▉| 83/84 [11:57<00:12, 12.27s/it]
100%|██████████| 84/84 [12:01<00:00, 9.95s/it]
{'eval_loss': 0.0, 'eval_runtime': 745.652, 'eval_samples_per_second': 0.671, 'eval_steps_per_second': 0.113, 'eval_num_tokens': 97092444.0, 'eval_completions/mean_length': 602.8393056960333, 'eval_completions/min_length': 266.26190476190476, 'eval_completions/max_length': 1204.2619047619048, 'eval_completions/clipped_ratio': 0.017857143389327184, 'eval_completions/mean_terminated_length': 558.6147010439918, 'eval_completions/min_terminated_length': 266.26190476190476, 'eval_completions/max_terminated_length': 987.0714285714286, 'eval_rewards/reward_correctness/mean': 0.7003968279986155, 'eval_rewards/reward_correctness/std': 0.4369296765043622, 'eval_reward': 0.700396845738093, 'eval_reward_std': 0.4369296765043622, 'eval_frac_reward_zero_std': 1.0, 'eval_sampling/sampling_logp_difference/mean': 0.045404179309982626, 'eval_sampling/sampling_logp_difference/max': 2.500789415268671, 'eval_sampling/importance_sampling_ratio/min': 3.2069904805975946e-05, 'eval_sampling/importance_sampling_ratio/mean': 0.0594283837005302, 'eval_sampling/importance_sampling_ratio/max': 0.31871562657137736, 'eval_entropy': 0.18525113555647077, 'eval_clip_ratio/low_mean': 0.0, 'eval_clip_ratio/low_min': 0.0, 'eval_clip_ratio/high_mean': 0.0, 'eval_clip_ratio/high_max': 0.0, 'eval_clip_ratio/region_mean': 0.0, 'epoch': 1.03}
51%|█████▏ | 70/136 [7:26:38<5:57:58, 325.43s/it]
100%|██████████| 84/84 [12:01<00:00, 9.95s/it]
[2026-08-13 05:57:26] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,732)
[2026-08-13 05:57:26] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "." @(1,4)
Open LaTeX blocks:
@(1,4) arguments of macro "."
[2026-08-13 05:57:30] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\)' @(1,10)
52%|█████▏ | 71/136 [7:32:03<9:54:39, 548.92s/it]
{'loss': -0.0331, 'grad_norm': 0.05220541352576947, 'learning_rate': 1.6661872078293582e-06, 'num_tokens': 98757763.0, 'completions/mean_length': 993.2702026367188, 'completions/min_length': 101.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.150390625, 'completions/mean_terminated_length': 625.3110961914062, 'completions/min_terminated_length': 101.0, 'completions/max_terminated_length': 2564.0, 'rewards/reward_correctness/mean': 0.5774739384651184, 'rewards/reward_correctness/std': 0.4941222071647644, 'reward': 0.5774739980697632, 'reward_std': 0.49412214756011963, 'frac_reward_zero_std': 0.484375, 'sampling/sampling_logp_difference/mean': 0.029383253306150436, 'sampling/sampling_logp_difference/max': 2.067214012145996, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5434272289276123, 'sampling/importance_sampling_ratio/max': 2.986877202987671, 'entropy': 0.5595204918645322, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 318.425050730817, 'epoch': 1.04}
52%|█████▏ | 71/136 [7:32:03<9:54:39, 548.92s/it][2026-08-13 06:02:55] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "frac" @(1,54)
Open LaTeX blocks:
@(1,54) arguments of macro "frac"
[2026-08-13 06:02:55] WARNING __init__.py:1389: WARNING: Error in configuration: macro '\frac' failed its substitution!
[2026-08-13 06:02:55] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,49)
53%|█████▎ | 72/136 [7:37:36<8:36:29, 484.21s/it]
{'loss': -0.0358, 'grad_norm': 0.05531152948843457, 'learning_rate': 1.6338127921706424e-06, 'num_tokens': 100533745.0, 'completions/mean_length': 1053.35546875, 'completions/min_length': 157.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1705729216337204, 'completions/mean_terminated_length': 638.43017578125, 'completions/min_terminated_length': 157.0, 'completions/max_terminated_length': 3068.0, 'rewards/reward_correctness/mean': 0.5221354365348816, 'rewards/reward_correctness/std': 0.49967244267463684, 'reward': 0.5221354365348816, 'reward_std': 0.49967244267463684, 'frac_reward_zero_std': 0.4609375, 'sampling/sampling_logp_difference/mean': 0.030927255749702454, 'sampling/sampling_logp_difference/max': 1.4990592002868652, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5030667781829834, 'sampling/importance_sampling_ratio/max': 2.9968209266662598, 'entropy': 0.6264856760390103, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 331.996779977344, 'epoch': 1.06}
53%|█████▎ | 72/136 [7:37:36<8:36:29, 484.21s/it][2026-08-13 06:08:14] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,0)
[2026-08-13 06:08:15] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,9184)
[2026-08-13 06:08:16] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,102)
54%|█████▎ | 73/136 [7:43:00<7:37:57, 436.16s/it]
{'loss': -0.0186, 'grad_norm': 0.05601516375217364, 'learning_rate': 1.6014476856245056e-06, 'num_tokens': 102047766.0, 'completions/mean_length': 894.5814208984375, 'completions/min_length': 166.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1256510466337204, 'completions/mean_terminated_length': 582.1735229492188, 'completions/min_terminated_length': 166.0, 'completions/max_terminated_length': 3043.0, 'rewards/reward_correctness/mean': 0.6028645634651184, 'rewards/reward_correctness/std': 0.48946383595466614, 'reward': 0.6028646230697632, 'reward_std': 0.48946380615234375, 'frac_reward_zero_std': 0.515625, 'sampling/sampling_logp_difference/mean': 0.028643345460295677, 'sampling/sampling_logp_difference/max': 1.042614459991455, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5782585740089417, 'sampling/importance_sampling_ratio/max': 2.9608514308929443, 'entropy': 0.5137150924419984, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 322.8676867983304, 'epoch': 1.07}
54%|█████▎ | 73/136 [7:43:00<7:37:57, 436.16s/it][2026-08-13 06:13:39] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,1)
[2026-08-13 06:13:39] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Bad \end macro: expected {environmentname} @(1,35)
[2026-08-13 06:13:39] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Bad \begin macro: expected {environmentname} @(1,0)
[2026-08-13 06:13:39] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing environment: 'pmatrix' @(1,24)
54%|█████▍ | 74/136 [7:48:15<6:53:08, 399.82s/it]
{'loss': -0.0311, 'grad_norm': 0.04147663365501501, 'learning_rate': 1.5691105010625233e-06, 'num_tokens': 103724162.0, 'completions/mean_length': 1002.8411865234375, 'completions/min_length': 130.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1555989682674408, 'completions/mean_terminated_length': 621.5543212890625, 'completions/min_terminated_length': 130.0, 'completions/max_terminated_length': 3022.0, 'rewards/reward_correctness/mean': 0.5787760615348816, 'rewards/reward_correctness/std': 0.49391618371009827, 'reward': 0.5787760615348816, 'reward_std': 0.4939161241054535, 'frac_reward_zero_std': 0.59375, 'sampling/sampling_logp_difference/mean': 0.029481105506420135, 'sampling/sampling_logp_difference/max': 1.729773759841919, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5173272490501404, 'sampling/importance_sampling_ratio/max': 2.9887375831604004, 'entropy': 0.5605857335031033, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 313.93872283026576, 'epoch': 1.09}
54%|█████▍ | 74/136 [7:48:15<6:53:08, 399.82s/it][2026-08-13 06:18:51] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "frac" @(1,1)
Open LaTeX blocks:
@(1,1) arguments of macro "frac"
[2026-08-13 06:18:51] WARNING __init__.py:1389: WARNING: Error in configuration: macro '\frac' failed its substitution!
[2026-08-13 06:18:51] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,2)
[2026-08-13 06:18:53] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "frac" @(1,2)
Open LaTeX blocks:
@(1,2) arguments of macro "frac"
[2026-08-13 06:18:53] WARNING __init__.py:1389: WARNING: Error in configuration: macro '\frac' failed its substitution!
55%|█████▌ | 75/136 [7:53:41<6:24:00, 377.71s/it]
{'loss': -0.036, 'grad_norm': 0.04361207970660597, 'learning_rate': 1.536819835298597e-06, 'num_tokens': 105242741.0, 'completions/mean_length': 887.205078125, 'completions/min_length': 167.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.12890625, 'completions/mean_terminated_length': 564.169677734375, 'completions/min_terminated_length': 167.0, 'completions/max_terminated_length': 2407.0, 'rewards/reward_correctness/mean': 0.6302083134651184, 'rewards/reward_correctness/std': 0.48290541768074036, 'reward': 0.6302083730697632, 'reward_std': 0.4829053580760956, 'frac_reward_zero_std': 0.546875, 'sampling/sampling_logp_difference/mean': 0.028588254004716873, 'sampling/sampling_logp_difference/max': 1.2616276741027832, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5378239750862122, 'sampling/importance_sampling_ratio/max': 2.95489764213562, 'entropy': 0.5045698995236307, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 324.96197805320844, 'epoch': 1.1}
55%|█████▌ | 75/136 [7:53:41<6:24:00, 377.71s/it][2026-08-13 06:24:26] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,1)
[2026-08-13 06:24:27] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,4)
56%|█████▌ | 76/136 [7:59:11<6:03:21, 363.35s/it]
{'loss': -0.021, 'grad_norm': 0.04873235501593393, 'learning_rate': 1.5045942583941002e-06, 'num_tokens': 106833713.0, 'completions/mean_length': 942.65625, 'completions/min_length': 146.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1393229216337204, 'completions/mean_terminated_length': 599.3380737304688, 'completions/min_terminated_length': 146.0, 'completions/max_terminated_length': 2865.0, 'rewards/reward_correctness/mean': 0.5553385615348816, 'rewards/reward_correctness/std': 0.49709007143974304, 'reward': 0.5553385615348816, 'reward_std': 0.49709001183509827, 'frac_reward_zero_std': 0.484375, 'sampling/sampling_logp_difference/mean': 0.02943594567477703, 'sampling/sampling_logp_difference/max': 3.937028408050537, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5509781241416931, 'sampling/importance_sampling_ratio/max': 2.947646141052246, 'entropy': 0.5472859075525776, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 328.5258017824963, 'epoch': 1.12}
56%|█████▌ | 76/136 [7:59:11<6:03:21, 363.35s/it][2026-08-13 06:30:14] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "frac" @(1,2)
Open LaTeX blocks:
@(1,2) arguments of macro "frac"
[2026-08-13 06:30:14] WARNING __init__.py:1389: WARNING: Error in configuration: macro '\frac' failed its substitution!
[2026-08-13 06:30:19] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,64)
57%|█████▋ | 77/136 [8:04:57<5:52:00, 357.97s/it]
{'loss': -0.0379, 'grad_norm': 0.058636649821457894, 'learning_rate': 1.4724523029784097e-06, 'num_tokens': 108318123.0, 'completions/mean_length': 885.1627807617188, 'completions/min_length': 132.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.12109375, 'completions/mean_terminated_length': 583.9644165039062, 'completions/min_terminated_length': 132.0, 'completions/max_terminated_length': 2860.0, 'rewards/reward_correctness/mean': 0.677734375, 'rewards/reward_correctness/std': 0.4674963057041168, 'reward': 0.677734375, 'reward_std': 0.46749624609947205, 'frac_reward_zero_std': 0.5078125, 'sampling/sampling_logp_difference/mean': 0.024872370064258575, 'sampling/sampling_logp_difference/max': 0.9959211349487305, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5988159775733948, 'sampling/importance_sampling_ratio/max': 2.9267632961273193, 'entropy': 0.45201024040579796, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 344.25549465743825, 'epoch': 1.13}
57%|█████▋ | 77/136 [8:04:57<5:52:00, 357.97s/it][2026-08-13 06:35:28] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,2161)
[2026-08-13 06:35:29] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "frac" @(1,1)
Open LaTeX blocks:
@(1,1) arguments of macro "frac"
[2026-08-13 06:35:29] WARNING __init__.py:1389: WARNING: Error in configuration: macro '\frac' failed its substitution!
57%|█████▋ | 78/136 [8:10:04<5:31:17, 342.72s/it]
{'loss': -0.0319, 'grad_norm': 0.056322046975264516, 'learning_rate': 1.4404124535909613e-06, 'num_tokens': 109893544.0, 'completions/mean_length': 937.7116088867188, 'completions/min_length': 134.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1354166716337204, 'completions/mean_terminated_length': 603.4254150390625, 'completions/min_terminated_length': 134.0, 'completions/max_terminated_length': 2615.0, 'rewards/reward_correctness/mean': 0.6536458134651184, 'rewards/reward_correctness/std': 0.47596263885498047, 'reward': 0.6536458730697632, 'reward_std': 0.4759626090526581, 'frac_reward_zero_std': 0.46875, 'sampling/sampling_logp_difference/mean': 0.02814130112528801, 'sampling/sampling_logp_difference/max': 1.4308843612670898, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.573235273361206, 'sampling/importance_sampling_ratio/max': 2.9960668087005615, 'entropy': 0.5290156111586839, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 305.98608493292704, 'epoch': 1.15}
57%|█████▋ | 78/136 [8:10:04<5:31:17, 342.72s/it][2026-08-13 06:40:49] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "frac" @(1,1)
Open LaTeX blocks:
@(1,1) arguments of macro "frac"
[2026-08-13 06:40:49] WARNING __init__.py:1389: WARNING: Error in configuration: macro '\frac' failed its substitution!
[2026-08-13 06:40:50] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "frac" @(1,1)
Open LaTeX blocks:
@(1,1) arguments of macro "frac"
[2026-08-13 06:40:50] WARNING __init__.py:1389: WARNING: Error in configuration: macro '\frac' failed its substitution!
[2026-08-13 06:40:52] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,4)
[2026-08-13 06:40:53] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,3518)
[2026-08-13 06:40:53] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,4303)
58%|█████▊ | 79/136 [8:15:30<5:20:54, 337.79s/it]
{'loss': -0.0304, 'grad_norm': 0.04613175848104597, 'learning_rate': 1.4084931360509656e-06, 'num_tokens': 111509776.0, 'completions/mean_length': 959.6171875, 'completions/min_length': 118.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1393229216337204, 'completions/mean_terminated_length': 617.6732177734375, 'completions/min_terminated_length': 118.0, 'completions/max_terminated_length': 2859.0, 'rewards/reward_correctness/mean': 0.6471354365348816, 'rewards/reward_correctness/std': 0.4780166745185852, 'reward': 0.6471354365348816, 'reward_std': 0.4780166447162628, 'frac_reward_zero_std': 0.4609375, 'sampling/sampling_logp_difference/mean': 0.027207151055336, 'sampling/sampling_logp_difference/max': 2.120018482208252, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5594146251678467, 'sampling/importance_sampling_ratio/max': 2.997708320617676, 'entropy': 0.5062569079454988, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 325.19561221823096, 'epoch': 1.16}
58%|█████▊ | 79/136 [8:15:30<5:20:54, 337.79s/it][2026-08-13 06:46:17] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,46)
[2026-08-13 06:46:18] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,113)
[2026-08-13 06:46:18] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "sqrt" @(1,13)
Open LaTeX blocks:
@(1,13) arguments of macro "sqrt"
[2026-08-13 06:46:21] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,601)
59%|█████▉ | 80/136 [8:20:57<5:12:20, 334.66s/it]
{'loss': -0.0295, 'grad_norm': 0.04659602550832931, 'learning_rate': 1.376712706860888e-06, 'num_tokens': 113157622.0, 'completions/mean_length': 977.91015625, 'completions/min_length': 174.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1471354216337204, 'completions/mean_terminated_length': 616.6397094726562, 'completions/min_terminated_length': 174.0, 'completions/max_terminated_length': 2118.0, 'rewards/reward_correctness/mean': 0.5631510615348816, 'rewards/reward_correctness/std': 0.4961574673652649, 'reward': 0.5631510615348816, 'reward_std': 0.4961574077606201, 'frac_reward_zero_std': 0.5, 'sampling/sampling_logp_difference/mean': 0.028690865263342857, 'sampling/sampling_logp_difference/max': 1.0071921348571777, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5413715243339539, 'sampling/importance_sampling_ratio/max': 2.9972426891326904, 'entropy': 0.5284926120657474, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 326.1782124824822, 'epoch': 1.18}
59%|█████▉ | 80/136 [8:20:57<5:12:20, 334.66s/it]
0%| | 0/84 [00:00<?, ?it/s]
2%|▏ | 2/84 [00:05<03:52, 2.83s/it]
4%|▎ | 3/84 [00:10<04:49, 3.58s/it]
5%|▍ | 4/84 [00:17<06:21, 4.76s/it]
6%|▌ | 5/84 [00:40<14:56, 11.34s/it]
7%|▋ | 6/84 [00:48<13:18, 10.24s/it]
8%|▊ | 7/84 [00:56<11:57, 9.32s/it]
10%|▉ | 8/84 [01:06<12:14, 9.67s/it]
11%|█ | 9/84 [01:15<11:53, 9.51s/it]
12%|█▏ | 10/84 [01:21<10:25, 8.45s/it]
13%|█▎ | 11/84 [01:29<09:52, 8.11s/it]
14%|█▍ | 12/84 [01:35<08:56, 7.46s/it]
15%|█▌ | 13/84 [01:40<08:12, 6.94s/it]
17%|█▋ | 14/84 [01:46<07:38, 6.56s/it]
18%|█▊ | 15/84 [01:54<08:01, 6.98s/it]
19%|█▉ | 16/84 [02:02<08:08, 7.19s/it]
20%|██ | 17/84 [02:07<07:25, 6.65s/it]
21%|██▏ | 18/84 [02:12<06:43, 6.12s/it]
23%|██▎ | 19/84 [02:22<08:01, 7.40s/it]
24%|██▍ | 20/84 [02:30<07:54, 7.41s/it]
25%|██▌ | 21/84 [02:44<09:48, 9.34s/it]
26%|██▌ | 22/84 [02:48<08:16, 8.01s/it]
27%|██▋ | 23/84 [02:56<07:59, 7.86s/it]
29%|██▊ | 24/84 [03:05<08:21, 8.36s/it]
30%|██▉ | 25/84 [03:10<07:03, 7.18s/it]
31%|███ | 26/84 [03:15<06:19, 6.54s/it]
32%|███▏ | 27/84 [03:24<06:51, 7.21s/it]
33%|███▎ | 28/84 [03:29<06:06, 6.54s/it]
35%|███▍ | 29/84 [03:37<06:24, 6.99s/it]
36%|███▌ | 30/84 [04:00<10:45, 11.95s/it]
37%|███▋ | 31/84 [04:09<09:46, 11.06s/it]
38%|███▊ | 32/84 [04:14<07:52, 9.09s/it]
39%|███▉ | 33/84 [04:27<08:41, 10.22s/it]
40%|████ | 34/84 [04:35<08:05, 9.71s/it]
42%|████▏ | 35/84 [04:59<11:19, 13.86s/it]
43%|████▎ | 36/84 [05:06<09:35, 11.99s/it]
44%|████▍ | 37/84 [05:11<07:42, 9.84s/it]
45%|████▌ | 38/84 [05:19<07:00, 9.15s/it]
46%|████▋ | 39/84 [05:42<10:06, 13.48s/it]
48%|████▊ | 40/84 [06:06<12:09, 16.59s/it]
49%|████▉ | 41/84 [06:12<09:35, 13.39s/it]
50%|█████ | 42/84 [06:21<08:24, 12.01s/it]
51%|█████ | 43/84 [06:26<06:48, 9.97s/it]
52%|█████▏ | 44/84 [06:31<05:37, 8.44s/it]
54%|█████▎ | 45/84 [06:40<05:37, 8.64s/it]
55%|█████▍ | 46/84 [06:47<05:11, 8.19s/it]
56%|█████▌ | 47/84 [07:11<07:53, 12.81s/it]
57%|█████▋ | 48/84 [07:17<06:34, 10.95s/it]
58%|█████▊ | 49/84 [07:28<06:16, 10.75s/it]
60%|█████▉ | 50/84 [07:43<06:50, 12.07s/it]
61%|██████ | 51/84 [07:52<06:10, 11.24s/it]
62%|██████▏ | 52/84 [08:00<05:31, 10.36s/it]
63%|██████▎ | 53/84 [08:07<04:49, 9.34s/it]
64%|██████▍ | 54/84 [08:11<03:46, 7.56s/it]
65%|██████▌ | 55/84 [08:24<04:26, 9.17s/it]
67%|██████▋ | 56/84 [08:27<03:30, 7.52s/it]
68%|██████▊ | 57/84 [08:35<03:20, 7.44s/it]
69%|██████▉ | 58/84 [08:40<02:59, 6.92s/it]
70%|███████ | 59/84 [08:49<03:08, 7.54s/it]
71%|███████▏ | 60/84 [08:55<02:47, 6.98s/it]
73%|███████▎ | 61/84 [09:03<02:45, 7.19s/it]
74%|███████▍ | 62/84 [09:10<02:36, 7.13s/it]
75%|███████▌ | 63/84 [09:19<02:46, 7.91s/it]
76%|███████▌ | 64/84 [09:27<02:37, 7.89s/it]
77%|███████▋ | 65/84 [09:51<04:01, 12.69s/it]
79%|███████▊ | 66/84 [09:56<03:03, 10.21s/it]
80%|███████▉ | 67/84 [10:02<02:35, 9.16s/it]
81%|████████ | 68/84 [10:16<02:50, 10.65s/it]
82%|████████▏ | 69/84 [10:23<02:22, 9.49s/it]
83%|████████▎ | 70/84 [10:30<01:59, 8.57s/it]
85%|████████▍ | 71/84 [10:33<01:31, 7.04s/it]
86%|████████▌ | 72/84 [10:39<01:20, 6.72s/it]
87%|████████▋ | 73/84 [10:45<01:12, 6.58s/it]
88%|████████▊ | 74/84 [10:51<01:03, 6.39s/it]
89%|████████▉ | 75/84 [10:58<00:58, 6.51s/it]
90%|█████████ | 76/84 [11:07<00:57, 7.25s/it]
92%|█████████▏| 77/84 [11:14<00:49, 7.06s/it]
93%|█████████▎| 78/84 [11:21<00:43, 7.18s/it]
94%|█████████▍| 79/84 [11:26<00:32, 6.45s/it]
95%|█████████▌| 80/84 [11:49<00:46, 11.56s/it]
96%|█████████▋| 81/84 [11:59<00:33, 11.07s/it]
98%|█████████▊| 82/84 [12:07<00:20, 10.15s/it]
99%|█████████▉| 83/84 [12:14<00:09, 9.21s/it]
100%|██████████| 84/84 [12:20<00:00, 8.06s/it]
{'eval_loss': 0.0, 'eval_runtime': 748.4449, 'eval_samples_per_second': 0.668, 'eval_steps_per_second': 0.112, 'eval_num_tokens': 113157622.0, 'eval_completions/mean_length': 619.6428756713867, 'eval_completions/min_length': 269.5833333333333, 'eval_completions/max_length': 1213.75, 'eval_completions/clipped_ratio': 0.015873016346068608, 'eval_completions/mean_terminated_length': 581.0579517909458, 'eval_completions/min_terminated_length': 269.5833333333333, 'eval_completions/max_terminated_length': 1036.107142857143, 'eval_rewards/reward_correctness/mean': 0.7182539705009687, 'eval_rewards/reward_correctness/std': 0.4072274182524, 'eval_reward': 0.7182539861117091, 'eval_reward_std': 0.4072274203811373, 'eval_frac_reward_zero_std': 1.0, 'eval_sampling/sampling_logp_difference/mean': 0.042117656696410405, 'eval_sampling/sampling_logp_difference/max': 2.371966535136813, 'eval_sampling/importance_sampling_ratio/min': 4.0693722128255106e-05, 'eval_sampling/importance_sampling_ratio/mean': 0.05727769766408961, 'eval_sampling/importance_sampling_ratio/max': 0.2644888570031894, 'eval_entropy': 0.17011943459510803, 'eval_clip_ratio/low_mean': 0.0, 'eval_clip_ratio/low_min': 0.0, 'eval_clip_ratio/high_mean': 0.0, 'eval_clip_ratio/high_max': 0.0, 'eval_clip_ratio/region_mean': 0.0, 'epoch': 1.18}
59%|█████▉ | 80/136 [8:33:26<5:12:20, 334.66s/it]
100%|██████████| 84/84 [12:20<00:00, 8.06s/it]
[2026-08-13 07:04:32] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,21)
60%|█████▉ | 81/136 [8:39:09<8:34:53, 561.70s/it]
{'loss': -0.0294, 'grad_norm': 0.05378785526390247, 'learning_rate': 1.3450894426497986e-06, 'num_tokens': 114818414.0, 'completions/mean_length': 995.4479370117188, 'completions/min_length': 131.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1529947966337204, 'completions/mean_terminated_length': 620.3597412109375, 'completions/min_terminated_length': 131.0, 'completions/max_terminated_length': 2791.0, 'rewards/reward_correctness/mean': 0.5611979365348816, 'rewards/reward_correctness/std': 0.4964022934436798, 'reward': 0.5611979365348816, 'reward_std': 0.4964022934436798, 'frac_reward_zero_std': 0.4375, 'sampling/sampling_logp_difference/mean': 0.028549566864967346, 'sampling/sampling_logp_difference/max': 1.353340744972229, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5383284091949463, 'sampling/importance_sampling_ratio/max': 2.973362684249878, 'entropy': 0.5432424978353083, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 336.11909542698413, 'epoch': 1.19}
60%|█████▉ | 81/136 [8:39:09<8:34:53, 561.70s/it]
60%|██████ | 82/136 [8:44:38<7:22:48, 492.01s/it]
{'loss': -0.043, 'grad_norm': 0.056012673442915696, 'learning_rate': 1.313641529662647e-06, 'num_tokens': 116402598.0, 'completions/mean_length': 943.6744995117188, 'completions/min_length': 157.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1432291716337204, 'completions/mean_terminated_length': 587.8753662109375, 'completions/min_terminated_length': 157.0, 'completions/max_terminated_length': 2812.0, 'rewards/reward_correctness/mean': 0.64453125, 'rewards/reward_correctness/std': 0.47881099581718445, 'reward': 0.64453125, 'reward_std': 0.47881096601486206, 'frac_reward_zero_std': 0.4609375, 'sampling/sampling_logp_difference/mean': 0.029155386611819267, 'sampling/sampling_logp_difference/max': 1.1875901222229004, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5390397906303406, 'sampling/importance_sampling_ratio/max': 2.9883127212524414, 'entropy': 0.5193857927224599, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 328.2109436998144, 'epoch': 1.21}
60%|██████ | 82/136 [8:44:38<7:22:48, 492.01s/it][2026-08-13 07:15:16] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,119)
61%|██████ | 83/136 [8:50:07<6:31:23, 443.08s/it]
{'loss': -0.0327, 'grad_norm': 0.04006469425367404, 'learning_rate': 1.2823870533015295e-06, 'num_tokens': 117905218.0, 'completions/mean_length': 880.8619995117188, 'completions/min_length': 166.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1145833358168602, 'completions/mean_terminated_length': 597.6353149414062, 'completions/min_terminated_length': 166.0, 'completions/max_terminated_length': 2260.0, 'rewards/reward_correctness/mean': 0.6380208134651184, 'rewards/reward_correctness/std': 0.480729341506958, 'reward': 0.6380208730697632, 'reward_std': 0.4807293117046356, 'frac_reward_zero_std': 0.5703125, 'sampling/sampling_logp_difference/mean': 0.02623382955789566, 'sampling/sampling_logp_difference/max': 1.315820574760437, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5929659008979797, 'sampling/importance_sampling_ratio/max': 2.9866812229156494, 'entropy': 0.4548957828665152, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 327.8097211890854, 'epoch': 1.22}
61%|██████ | 83/136 [8:50:07<6:31:23, 443.08s/it]
62%|██████▏ | 84/136 [8:55:34<5:53:48, 408.25s/it]
{'loss': -0.0493, 'grad_norm': 0.04990192862275496, 'learning_rate': 1.2513439877249363e-06, 'num_tokens': 119565737.0, 'completions/mean_length': 979.0045776367188, 'completions/min_length': 113.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.15234375, 'completions/mean_terminated_length': 603.8709716796875, 'completions/min_terminated_length': 113.0, 'completions/max_terminated_length': 2345.0, 'rewards/reward_correctness/mean': 0.6236979365348816, 'rewards/reward_correctness/std': 0.4846150279045105, 'reward': 0.6236979365348816, 'reward_std': 0.4846149981021881, 'frac_reward_zero_std': 0.5625, 'sampling/sampling_logp_difference/mean': 0.030388643965125084, 'sampling/sampling_logp_difference/max': 1.264847993850708, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5159881114959717, 'sampling/importance_sampling_ratio/max': 2.9869635105133057, 'entropy': 0.5572732011787593, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 325.7631025062874, 'epoch': 1.24}
62%|██████▏ | 84/136 [8:55:34<5:53:48, 408.25s/it]
62%|██████▎ | 85/136 [9:01:07<5:27:40, 385.51s/it]
{'loss': -0.0292, 'grad_norm': 0.05425624294733418, 'learning_rate': 1.220530185510985e-06, 'num_tokens': 121217736.0, 'completions/mean_length': 979.5514526367188, 'completions/min_length': 136.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.162109375, 'completions/mean_terminated_length': 574.71875, 'completions/min_terminated_length': 136.0, 'completions/max_terminated_length': 2351.0, 'rewards/reward_correctness/mean': 0.6087239384651184, 'rewards/reward_correctness/std': 0.48819488286972046, 'reward': 0.6087239980697632, 'reward_std': 0.48819488286972046, 'frac_reward_zero_std': 0.53125, 'sampling/sampling_logp_difference/mean': 0.029596565291285515, 'sampling/sampling_logp_difference/max': 2.14804744720459, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5493928790092468, 'sampling/importance_sampling_ratio/max': 2.8903768062591553, 'entropy': 0.5549764286261052, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 331.2769885966554, 'epoch': 1.25}
62%|██████▎ | 85/136 [9:01:07<5:27:40, 385.51s/it][2026-08-13 07:31:43] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,238)
63%|██████▎ | 86/136 [9:06:15<5:02:06, 362.52s/it]
{'loss': -0.0236, 'grad_norm': 0.049832296894574454, 'learning_rate': 1.189963367390565e-06, 'num_tokens': 122787444.0, 'completions/mean_length': 938.671875, 'completions/min_length': 152.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1197916716337204, 'completions/mean_terminated_length': 648.3372802734375, 'completions/min_terminated_length': 152.0, 'completions/max_terminated_length': 2747.0, 'rewards/reward_correctness/mean': 0.6236979365348816, 'rewards/reward_correctness/std': 0.4846150279045105, 'reward': 0.6236979365348816, 'reward_std': 0.4846149981021881, 'frac_reward_zero_std': 0.515625, 'sampling/sampling_logp_difference/mean': 0.02557907998561859, 'sampling/sampling_logp_difference/max': 1.1519529819488525, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5417976379394531, 'sampling/importance_sampling_ratio/max': 2.9678292274475098, 'entropy': 0.46246786625124514, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 307.65380067098886, 'epoch': 1.26}
63%|██████▎ | 86/136 [9:06:15<5:02:06, 362.52s/it][2026-08-13 07:37:06] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,441)
64%|██████▍ | 87/136 [9:12:00<4:51:44, 357.24s/it]
{'loss': -0.0092, 'grad_norm': 0.05934246184930205, 'learning_rate': 1.159661112056314e-06, 'num_tokens': 124479045.0, 'completions/mean_length': 986.380859375, 'completions/min_length': 165.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.13671875, 'completions/mean_terminated_length': 656.8952026367188, 'completions/min_terminated_length': 165.0, 'completions/max_terminated_length': 2030.0, 'rewards/reward_correctness/mean': 0.5924479365348816, 'rewards/reward_correctness/std': 0.4915390610694885, 'reward': 0.5924479365348816, 'reward_std': 0.4915390610694885, 'frac_reward_zero_std': 0.46875, 'sampling/sampling_logp_difference/mean': 0.043718449771404266, 'sampling/sampling_logp_difference/max': 27.75, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.46693944931030273, 'sampling/importance_sampling_ratio/max': 2.974931240081787, 'entropy': 0.49913884582929313, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 343.7939372379333, 'epoch': 1.28}
64%|██████▍ | 87/136 [9:12:00<4:51:44, 357.24s/it]
65%|██████▍ | 88/136 [9:17:06<4:33:27, 341.82s/it]
{'loss': -0.0334, 'grad_norm': 0.057870776481504994, 'learning_rate': 1.1296408460532715e-06, 'num_tokens': 125892913.0, 'completions/mean_length': 822.8307495117188, 'completions/min_length': 182.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.0891927108168602, 'completions/mean_terminated_length': 602.7734375, 'completions/min_terminated_length': 182.0, 'completions/max_terminated_length': 1973.0, 'rewards/reward_correctness/mean': 0.6725260615348816, 'rewards/reward_correctness/std': 0.4694446325302124, 'reward': 0.6725260615348816, 'reward_std': 0.4694445729255676, 'frac_reward_zero_std': 0.5078125, 'sampling/sampling_logp_difference/mean': 0.023823007941246033, 'sampling/sampling_logp_difference/max': 1.2202494144439697, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.6000463366508484, 'sampling/importance_sampling_ratio/max': 2.993168830871582, 'entropy': 0.40294282842660323, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 304.67693279404193, 'epoch': 1.29}
65%|██████▍ | 88/136 [9:17:06<4:33:27, 341.82s/it][2026-08-13 07:47:34] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,1)
[2026-08-13 07:47:35] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,2)
[2026-08-13 07:47:36] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,3)
[2026-08-13 07:47:36] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,1)
[2026-08-13 07:47:36] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,2)
[2026-08-13 07:47:36] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing environment: 'aligned' @(1,10665)
65%|██████▌ | 89/136 [9:22:06<4:17:57, 329.32s/it]
{'loss': -0.0112, 'grad_norm': 0.05036942737826866, 'learning_rate': 1.0999198337570392e-06, 'num_tokens': 127410409.0, 'completions/mean_length': 899.1484375, 'completions/min_length': 145.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.107421875, 'completions/mean_terminated_length': 637.6455078125, 'completions/min_terminated_length': 145.0, 'completions/max_terminated_length': 2732.0, 'rewards/reward_correctness/mean': 0.6126301884651184, 'rewards/reward_correctness/std': 0.48730793595314026, 'reward': 0.6126302480697632, 'reward_std': 0.48730790615081787, 'frac_reward_zero_std': 0.53125, 'sampling/sampling_logp_difference/mean': 0.02451782487332821, 'sampling/sampling_logp_difference/max': 0.9818124771118164, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5925201773643494, 'sampling/importance_sampling_ratio/max': 2.9690043926239014, 'entropy': 0.4273973684757948, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 298.9374023540877, 'epoch': 1.31}
65%|██████▌ | 89/136 [9:22:06<4:17:57, 329.32s/it][2026-08-13 07:52:49] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,76)
[2026-08-13 07:52:49] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,215)
[2026-08-13 07:52:49] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,342)
66%|██████▌ | 90/136 [9:27:30<4:11:08, 327.57s/it]
{'loss': -0.0373, 'grad_norm': 0.043707152943409704, 'learning_rate': 1.0705151674451938e-06, 'num_tokens': 129020677.0, 'completions/mean_length': 954.78125, 'completions/min_length': 175.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1223958358168602, 'completions/mean_terminated_length': 659.6676635742188, 'completions/min_terminated_length': 175.0, 'completions/max_terminated_length': 2859.0, 'rewards/reward_correctness/mean': 0.595703125, 'rewards/reward_correctness/std': 0.49091529846191406, 'reward': 0.595703125, 'reward_std': 0.4909152686595917, 'frac_reward_zero_std': 0.5234375, 'sampling/sampling_logp_difference/mean': 0.02434036135673523, 'sampling/sampling_logp_difference/max': 1.6809594631195068, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5464567542076111, 'sampling/importance_sampling_ratio/max': 2.9962658882141113, 'entropy': 0.43672167870681733, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 322.3661844804883, 'epoch': 1.32}
66%|██████▌ | 90/136 [9:27:30<4:11:08, 327.57s/it]
0%| | 0/84 [00:00<?, ?it/s]
2%|▏ | 2/84 [00:06<04:14, 3.10s/it]
4%|▎ | 3/84 [00:11<05:40, 4.20s/it]
5%|▍ | 4/84 [00:18<06:57, 5.22s/it]
6%|▌ | 5/84 [00:34<11:42, 8.89s/it]
7%|▋ | 6/84 [00:41<10:29, 8.07s/it]
8%|▊ | 7/84 [00:49<10:23, 8.09s/it]
10%|▉ | 8/84 [00:53<08:48, 6.96s/it]
11%|█ | 9/84 [01:06<11:04, 8.85s/it]
12%|█▏ | 10/84 [01:12<09:52, 8.01s/it]
13%|█▎ | 11/84 [01:18<09:00, 7.40s/it]
14%|█▍ | 12/84 [01:25<08:25, 7.02s/it]
15%|█▌ | 13/84 [01:31<08:10, 6.91s/it]
17%|█▋ | 14/84 [01:55<13:56, 11.95s/it]
18%|█▊ | 15/84 [02:02<12:01, 10.46s/it]
19%|█▉ | 16/84 [02:07<10:04, 8.90s/it]
20%|██ | 17/84 [02:14<09:09, 8.20s/it]
21%|██▏ | 18/84 [02:19<08:12, 7.47s/it]
23%|██▎ | 19/84 [02:32<09:48, 9.05s/it]
24%|██▍ | 20/84 [02:41<09:36, 9.01s/it]
25%|██▌ | 21/84 [02:50<09:30, 9.06s/it]
26%|██▌ | 22/84 [02:54<07:51, 7.61s/it]
27%|██▋ | 23/84 [03:01<07:26, 7.32s/it]
29%|██▊ | 24/84 [03:10<07:44, 7.74s/it]
30%|██▉ | 25/84 [03:14<06:36, 6.72s/it]
31%|███ | 26/84 [03:19<05:51, 6.05s/it]
32%|███▏ | 27/84 [03:43<10:49, 11.40s/it]
33%|███▎ | 28/84 [03:46<08:28, 9.08s/it]
35%|███▍ | 29/84 [03:53<07:40, 8.37s/it]
36%|███▌ | 30/84 [04:17<11:39, 12.95s/it]
37%|███▋ | 31/84 [04:23<09:40, 10.95s/it]
38%|███▊ | 32/84 [04:28<07:56, 9.17s/it]
39%|███▉ | 33/84 [04:35<07:21, 8.65s/it]
40%|████ | 34/84 [04:59<10:55, 13.11s/it]
42%|████▏ | 35/84 [05:07<09:31, 11.67s/it]
43%|████▎ | 36/84 [05:13<07:56, 9.93s/it]
44%|████▍ | 37/84 [05:18<06:38, 8.48s/it]
45%|████▌ | 38/84 [05:24<05:57, 7.77s/it]
46%|████▋ | 39/84 [05:33<05:58, 7.97s/it]
48%|████▊ | 40/84 [05:45<06:48, 9.27s/it]
49%|████▉ | 41/84 [05:51<06:03, 8.45s/it]
50%|█████ | 42/84 [06:00<05:50, 8.34s/it]
51%|█████ | 43/84 [06:07<05:34, 8.15s/it]
52%|█████▏ | 44/84 [06:12<04:44, 7.10s/it]
54%|█████▎ | 45/84 [06:19<04:34, 7.03s/it]
55%|█████▍ | 46/84 [06:30<05:17, 8.36s/it]
56%|█████▌ | 47/84 [06:38<04:57, 8.04s/it]
57%|█████▋ | 48/84 [06:45<04:45, 7.92s/it]
58%|█████▊ | 49/84 [06:54<04:45, 8.15s/it]
60%|█████▉ | 50/84 [07:02<04:34, 8.08s/it]
61%|██████ | 51/84 [07:26<07:02, 12.81s/it]
62%|██████▏ | 52/84 [07:34<06:05, 11.43s/it]
63%|██████▎ | 53/84 [07:39<04:53, 9.47s/it]
64%|██████▍ | 54/84 [07:42<03:46, 7.56s/it]
65%|██████▌ | 55/84 [08:05<05:57, 12.33s/it]
67%|██████▋ | 56/84 [08:09<04:28, 9.60s/it]
68%|██████▊ | 57/84 [08:15<03:56, 8.76s/it]
69%|██████▉ | 58/84 [08:19<03:07, 7.20s/it]
70%|███████ | 59/84 [08:28<03:14, 7.80s/it]
71%|███████▏ | 60/84 [08:34<02:55, 7.31s/it]
73%|███████▎ | 61/84 [08:41<02:46, 7.23s/it]
74%|███████▍ | 62/84 [08:51<02:56, 8.04s/it]
75%|███████▌ | 63/84 [09:03<03:10, 9.08s/it]
76%|███████▌ | 64/84 [09:15<03:20, 10.04s/it]
77%|███████▋ | 65/84 [09:27<03:24, 10.74s/it]
79%|███████▊ | 66/84 [09:32<02:42, 9.02s/it]
80%|███████▉ | 67/84 [09:38<02:15, 7.98s/it]
81%|████████ | 68/84 [09:48<02:16, 8.55s/it]
82%|████████▏ | 69/84 [09:55<02:00, 8.02s/it]
83%|████████▎ | 70/84 [10:02<01:47, 7.70s/it]
85%|████████▍ | 71/84 [10:06<01:25, 6.61s/it]
86%|████████▌ | 72/84 [10:12<01:19, 6.64s/it]
87%|████████▋ | 73/84 [10:19<01:13, 6.64s/it]
88%|████████▊ | 74/84 [10:27<01:10, 7.01s/it]
89%|████████▉ | 75/84 [10:33<01:01, 6.79s/it]
90%|█████████ | 76/84 [10:41<00:57, 7.16s/it]
92%|█████████▏| 77/84 [11:05<01:24, 12.10s/it]
93%|█████████▎| 78/84 [11:12<01:03, 10.53s/it]
94%|█████████▍| 79/84 [11:19<00:48, 9.66s/it]
95%|█████████▌| 80/84 [11:34<00:44, 11.09s/it]
96%|█████████▋| 81/84 [11:40<00:28, 9.53s/it]
98%|█████████▊| 82/84 [11:47<00:18, 9.02s/it]
99%|█████████▉| 83/84 [11:53<00:08, 8.07s/it]
100%|██████████| 84/84 [12:00<00:00, 7.59s/it]
{'eval_loss': 0.0, 'eval_runtime': 727.5572, 'eval_samples_per_second': 0.687, 'eval_steps_per_second': 0.115, 'eval_num_tokens': 129020677.0, 'eval_completions/mean_length': 611.1190646943592, 'eval_completions/min_length': 273.5952380952381, 'eval_completions/max_length': 1180.452380952381, 'eval_completions/clipped_ratio': 0.013888889302810034, 'eval_completions/mean_terminated_length': 576.907953171503, 'eval_completions/min_terminated_length': 273.5952380952381, 'eval_completions/max_terminated_length': 1016.547619047619, 'eval_rewards/reward_correctness/mean': 0.7242063500341915, 'eval_rewards/reward_correctness/std': 0.405771173891567, 'eval_reward': 0.72420636706409, 'eval_reward_std': 0.4057711760203044, 'eval_frac_reward_zero_std': 1.0, 'eval_sampling/sampling_logp_difference/mean': 0.03981469148060396, 'eval_sampling/sampling_logp_difference/max': 2.489054736636934, 'eval_sampling/importance_sampling_ratio/min': 3.644251580033999e-05, 'eval_sampling/importance_sampling_ratio/mean': 0.05651876312922035, 'eval_sampling/importance_sampling_ratio/max': 0.2706919278598867, 'eval_entropy': 0.1594409356337218, 'eval_clip_ratio/low_mean': 0.0, 'eval_clip_ratio/low_min': 0.0, 'eval_clip_ratio/high_mean': 0.0, 'eval_clip_ratio/high_max': 0.0, 'eval_clip_ratio/region_mean': 0.0, 'epoch': 1.32}
66%|██████▌ | 90/136 [9:39:37<4:11:08, 327.57s/it]
100%|██████████| 84/84 [12:00<00:00, 7.59s/it]

67%|██████▋ | 91/136 [9:45:42<6:57:42, 556.94s/it]
{'loss': -0.0285, 'grad_norm': 0.049403019001355145, 'learning_rate': 1.0414437574676832e-06, 'num_tokens': 130553533.0, 'completions/mean_length': 886.9453125, 'completions/min_length': 162.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1184895858168602, 'completions/mean_terminated_length': 594.0214233398438, 'completions/min_terminated_length': 162.0, 'completions/max_terminated_length': 2849.0, 'rewards/reward_correctness/mean': 0.6666666865348816, 'rewards/reward_correctness/std': 0.4715580344200134, 'reward': 0.6666666865348816, 'reward_std': 0.4715580344200134, 'frac_reward_zero_std': 0.53125, 'sampling/sampling_logp_difference/mean': 0.0383615642786026, 'sampling/sampling_logp_difference/max': 23.5, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.4857485294342041, 'sampling/importance_sampling_ratio/max': 2.987356662750244, 'entropy': 0.4434245915035717, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 357.55924094608054, 'epoch': 1.34}
67%|██████▋ | 91/136 [9:45:42<6:57:42, 556.94s/it]
68%|██████▊ | 92/136 [9:51:06<5:57:12, 487.11s/it]
{'loss': -0.0261, 'grad_norm': 0.047660426114125534, 'learning_rate': 1.0127223225218379e-06, 'num_tokens': 132141518.0, 'completions/mean_length': 936.6178588867188, 'completions/min_length': 146.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1276041716337204, 'completions/mean_terminated_length': 624.3798828125, 'completions/min_terminated_length': 146.0, 'completions/max_terminated_length': 2910.0, 'rewards/reward_correctness/mean': 0.6321614384651184, 'rewards/reward_correctness/std': 0.4823741614818573, 'reward': 0.6321614980697632, 'reward_std': 0.4823741316795349, 'frac_reward_zero_std': 0.5078125, 'sampling/sampling_logp_difference/mean': 0.026067327708005905, 'sampling/sampling_logp_difference/max': 0.9881277084350586, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5875497460365295, 'sampling/importance_sampling_ratio/max': 2.9987378120422363, 'entropy': 0.4767748566227965, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 323.0196398496628, 'epoch': 1.35}
68%|██████▊ | 92/136 [9:51:06<5:57:12, 487.11s/it][2026-08-13 08:21:39] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,379)
68%|██████▊ | 93/136 [9:56:28<5:13:39, 437.65s/it]
{'loss': -0.0073, 'grad_norm': 0.047670989020446516, 'learning_rate': 9.843673800376037e-07, 'num_tokens': 133603158.0, 'completions/mean_length': 850.3541870117188, 'completions/min_length': 113.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.103515625, 'completions/mean_terminated_length': 594.8068237304688, 'completions/min_terminated_length': 113.0, 'completions/max_terminated_length': 2182.0, 'rewards/reward_correctness/mean': 0.6373698115348816, 'rewards/reward_correctness/std': 0.480915904045105, 'reward': 0.6373698115348816, 'reward_std': 0.480915904045105, 'frac_reward_zero_std': 0.5390625, 'sampling/sampling_logp_difference/mean': 0.02443034015595913, 'sampling/sampling_logp_difference/max': 1.733330488204956, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5859780311584473, 'sampling/importance_sampling_ratio/max': 2.926886558532715, 'entropy': 0.41502448660321534, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 321.0396940037608, 'epoch': 1.37}
68%|██████▊ | 93/136 [9:56:28<5:13:39, 437.65s/it]
69%|██████▉ | 94/136 [10:01:45<4:40:55, 401.33s/it]
{'loss': -0.0288, 'grad_norm': 0.05508775780210909, 'learning_rate': 9.563952366785246e-07, 'num_tokens': 135027286.0, 'completions/mean_length': 832.0885620117188, 'completions/min_length': 120.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.095703125, 'completions/mean_terminated_length': 595.0352783203125, 'completions/min_terminated_length': 120.0, 'completions/max_terminated_length': 2675.0, 'rewards/reward_correctness/mean': 0.6360676884651184, 'rewards/reward_correctness/std': 0.4812861680984497, 'reward': 0.6360677480697632, 'reward_std': 0.4812861680984497, 'frac_reward_zero_std': 0.5625, 'sampling/sampling_logp_difference/mean': 0.024181541055440903, 'sampling/sampling_logp_difference/max': 1.4866766929626465, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5902597904205322, 'sampling/importance_sampling_ratio/max': 2.9579274654388428, 'entropy': 0.4050459030549973, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 315.43966129794717, 'epoch': 1.38}
69%|██████▉ | 94/136 [10:01:45<4:40:55, 401.33s/it][2026-08-13 08:32:27] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,53)
[2026-08-13 08:32:28] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,322)
[2026-08-13 08:32:28] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,542)
70%|██████▉ | 95/136 [10:07:07<4:17:52, 377.39s/it]
{'loss': -0.0079, 'grad_norm': 0.04858255336422534, 'learning_rate': 9.288219789639276e-07, 'num_tokens': 136612824.0, 'completions/mean_length': 941.1107177734375, 'completions/min_length': 129.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1243489608168602, 'completions/mean_terminated_length': 638.5300903320312, 'completions/min_terminated_length': 129.0, 'completions/max_terminated_length': 2420.0, 'rewards/reward_correctness/mean': 0.6061198115348816, 'rewards/reward_correctness/std': 0.48876795172691345, 'reward': 0.6061198115348816, 'reward_std': 0.48876795172691345, 'frac_reward_zero_std': 0.46875, 'sampling/sampling_logp_difference/mean': 0.026240212842822075, 'sampling/sampling_logp_difference/max': 3.132884979248047, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5338165163993835, 'sampling/importance_sampling_ratio/max': 2.995532989501953, 'entropy': 0.46013247850351036, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 320.3804359170608, 'epoch': 1.4}
70%|██████▉ | 95/136 [10:07:07<4:17:52, 377.39s/it][2026-08-13 08:37:48] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,1360)
71%|███████ | 96/136 [10:12:23<3:59:29, 359.23s/it]
{'loss': -0.0346, 'grad_norm': 0.048290932446018155, 'learning_rate': 9.016634640177203e-07, 'num_tokens': 138131150.0, 'completions/mean_length': 899.4466552734375, 'completions/min_length': 149.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1126302108168602, 'completions/mean_terminated_length': 623.693359375, 'completions/min_terminated_length': 149.0, 'completions/max_terminated_length': 2845.0, 'rewards/reward_correctness/mean': 0.6145833134651184, 'rewards/reward_correctness/std': 0.48685210943222046, 'reward': 0.6145833730697632, 'reward_std': 0.48685210943222046, 'frac_reward_zero_std': 0.515625, 'sampling/sampling_logp_difference/mean': 0.025148890912532806, 'sampling/sampling_logp_difference/max': 1.4211063385009766, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.583075225353241, 'sampling/importance_sampling_ratio/max': 2.995443820953369, 'entropy': 0.4312205637106672, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 315.7374519892037, 'epoch': 1.41}
71%|███████ | 96/136 [10:12:23<3:59:29, 359.23s/it][2026-08-13 08:43:05] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "frac" @(1,1768)
Open LaTeX blocks:
@(1,1768) arguments of macro "frac"
[2026-08-13 08:43:05] WARNING __init__.py:1389: WARNING: Error in configuration: macro '\frac' failed its substitution!
[2026-08-13 08:43:06] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\)' @(1,202)
[2026-08-13 08:43:08] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,194)
[2026-08-13 08:43:08] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,476)
71%|███████▏ | 97/136 [10:17:54<3:47:52, 350.57s/it]
{'loss': -0.019, 'grad_norm': 0.04319162219933892, 'learning_rate': 8.74935310449101e-07, 'num_tokens': 139587937.0, 'completions/mean_length': 847.4447021484375, 'completions/min_length': 120.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.09765625, 'completions/mean_terminated_length': 606.69189453125, 'completions/min_terminated_length': 120.0, 'completions/max_terminated_length': 2736.0, 'rewards/reward_correctness/mean': 0.6692708134651184, 'rewards/reward_correctness/std': 0.47062891721725464, 'reward': 0.6692708730697632, 'reward_std': 0.47062891721725464, 'frac_reward_zero_std': 0.5625, 'sampling/sampling_logp_difference/mean': 0.023664306849241257, 'sampling/sampling_logp_difference/max': 0.9985542297363281, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5773332715034485, 'sampling/importance_sampling_ratio/max': 2.9770994186401367, 'entropy': 0.395701271481812, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 329.263949975837, 'epoch': 1.43}
71%|███████▏ | 97/136 [10:17:54<3:47:52, 350.57s/it][2026-08-13 08:48:30] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,4)
[2026-08-13 08:48:31] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,86)
72%|███████▏ | 98/136 [10:23:13<3:36:06, 341.23s/it]
{'loss': -0.0084, 'grad_norm': 0.05443612081724022, 'learning_rate': 8.486528893704481e-07, 'num_tokens': 141091804.0, 'completions/mean_length': 890.892578125, 'completions/min_length': 181.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1100260466337204, 'completions/mean_terminated_length': 621.432373046875, 'completions/min_terminated_length': 181.0, 'completions/max_terminated_length': 2920.0, 'rewards/reward_correctness/mean': 0.6516926884651184, 'rewards/reward_correctness/std': 0.4765891432762146, 'reward': 0.6516927480697632, 'reward_std': 0.4765890836715698, 'frac_reward_zero_std': 0.5703125, 'sampling/sampling_logp_difference/mean': 0.02437897026538849, 'sampling/sampling_logp_difference/max': 3.037632465362549, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5741823315620422, 'sampling/importance_sampling_ratio/max': 2.99269700050354, 'entropy': 0.41403548134258017, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 318.34623634815216, 'epoch': 1.44}
72%|███████▏ | 98/136 [10:23:13<3:36:06, 341.23s/it][2026-08-13 08:53:37] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "frac" @(1,1)
Open LaTeX blocks:
@(1,1) arguments of macro "frac"
[2026-08-13 08:53:37] WARNING __init__.py:1389: WARNING: Error in configuration: macro '\frac' failed its substitution!
[2026-08-13 08:53:37] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,10)
73%|███████▎ | 99/136 [10:28:08<3:21:48, 327.24s/it]
{'loss': -0.0001, 'grad_norm': 0.05116238263191225, 'learning_rate': 8.228313155575304e-07, 'num_tokens': 142478098.0, 'completions/mean_length': 816.14453125, 'completions/min_length': 107.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.0924479216337204, 'completions/mean_terminated_length': 586.3515014648438, 'completions/min_terminated_length': 107.0, 'completions/max_terminated_length': 2389.0, 'rewards/reward_correctness/mean': 0.6341145634651184, 'rewards/reward_correctness/std': 0.48183444142341614, 'reward': 0.6341146230697632, 'reward_std': 0.48183441162109375, 'frac_reward_zero_std': 0.546875, 'sampling/sampling_logp_difference/mean': 0.023311734199523926, 'sampling/sampling_logp_difference/max': 1.2119526863098145, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.623569905757904, 'sampling/importance_sampling_ratio/max': 2.990213632583618, 'entropy': 0.3746476907399483, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 293.37062060274184, 'epoch': 1.46}
73%|███████▎ | 99/136 [10:28:08<3:21:48, 327.24s/it]
74%|███████▎ | 100/136 [10:33:31<3:15:33, 325.92s/it]
{'loss': -0.0102, 'grad_norm': 0.06215867891136184, 'learning_rate': 7.97485438757144e-07, 'num_tokens': 144058587.0, 'completions/mean_length': 935.8157958984375, 'completions/min_length': 163.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1276041716337204, 'completions/mean_terminated_length': 623.3589477539062, 'completions/min_terminated_length': 163.0, 'completions/max_terminated_length': 3069.0, 'rewards/reward_correctness/mean': 0.607421875, 'rewards/reward_correctness/std': 0.4884832501411438, 'reward': 0.607421875, 'reward_std': 0.4884832203388214, 'frac_reward_zero_std': 0.4375, 'sampling/sampling_logp_difference/mean': 0.026227515190839767, 'sampling/sampling_logp_difference/max': 2.526287078857422, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5880123972892761, 'sampling/importance_sampling_ratio/max': 2.9991674423217773, 'entropy': 0.46815127378795296, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 321.7099755220115, 'epoch': 1.47}
74%|███████▎ | 100/136 [10:33:31<3:15:33, 325.92s/it]
0%| | 0/84 [00:00<?, ?it/s]
2%|▏ | 2/84 [00:05<03:41, 2.70s/it]
4%|▎ | 3/84 [00:09<04:33, 3.37s/it]
5%|▍ | 4/84 [00:15<05:48, 4.36s/it]
6%|▌ | 5/84 [00:39<14:35, 11.08s/it]
7%|▋ | 6/84 [00:46<12:33, 9.66s/it]
8%|▊ | 7/84 [00:53<11:30, 8.97s/it]
10%|▉ | 8/84 [00:58<09:50, 7.77s/it]
11%|█ | 9/84 [01:13<12:23, 9.92s/it]
12%|█▏ | 10/84 [01:20<10:56, 8.88s/it]
13%|█▎ | 11/84 [01:26<09:56, 8.18s/it]
14%|█▍ | 12/84 [01:32<09:03, 7.55s/it]
15%|█▌ | 13/84 [01:41<09:18, 7.86s/it]
17%|█▋ | 14/84 [01:47<08:29, 7.27s/it]
18%|█▊ | 15/84 [01:54<08:25, 7.33s/it]
19%|█▉ | 16/84 [02:00<07:41, 6.79s/it]
20%|██ | 17/84 [02:05<07:12, 6.46s/it]
21%|██▏ | 18/84 [02:12<07:00, 6.38s/it]
23%|██▎ | 19/84 [02:20<07:40, 7.09s/it]
24%|██▍ | 20/84 [02:28<07:50, 7.35s/it]
25%|██▌ | 21/84 [02:38<08:21, 7.96s/it]
26%|██▌ | 22/84 [02:42<06:56, 6.72s/it]
27%|██▋ | 23/84 [02:48<06:46, 6.67s/it]
29%|██▊ | 24/84 [02:57<07:25, 7.43s/it]
30%|██▉ | 25/84 [03:03<06:44, 6.85s/it]
31%|███ | 26/84 [03:09<06:19, 6.55s/it]
32%|███▏ | 27/84 [03:18<06:52, 7.24s/it]
33%|███▎ | 28/84 [03:41<11:21, 12.17s/it]
35%|███▍ | 29/84 [03:51<10:24, 11.35s/it]
36%|███▌ | 30/84 [03:59<09:27, 10.50s/it]
37%|███▋ | 31/84 [04:05<08:02, 9.10s/it]
38%|███▊ | 32/84 [04:09<06:37, 7.65s/it]
39%|███▉ | 33/84 [04:33<10:35, 12.46s/it]
40%|████ | 34/84 [04:42<09:29, 11.39s/it]
42%|████▏ | 35/84 [05:06<12:20, 15.12s/it]
43%|████▎ | 36/84 [05:12<10:00, 12.50s/it]
44%|████▍ | 37/84 [05:16<07:47, 9.96s/it]
45%|████▌ | 38/84 [05:23<06:55, 9.04s/it]
46%|████▋ | 39/84 [05:31<06:26, 8.59s/it]
48%|████▊ | 40/84 [05:54<09:39, 13.18s/it]
49%|████▉ | 41/84 [06:03<08:25, 11.74s/it]
50%|█████ | 42/84 [06:27<10:44, 15.34s/it]
51%|█████ | 43/84 [06:33<08:38, 12.66s/it]
52%|█████▏ | 44/84 [06:38<06:53, 10.34s/it]
54%|█████▎ | 45/84 [07:02<09:23, 14.45s/it]
55%|█████▍ | 46/84 [07:08<07:31, 11.89s/it]
56%|█████▌ | 47/84 [07:32<09:31, 15.44s/it]
57%|█████▋ | 48/84 [07:38<07:44, 12.89s/it]
58%|█████▊ | 49/84 [07:50<07:14, 12.40s/it]
60%|█████▉ | 50/84 [07:59<06:28, 11.43s/it]
61%|██████ | 51/84 [08:05<05:25, 9.86s/it]
62%|██████▏ | 52/84 [08:15<05:15, 9.87s/it]
63%|██████▎ | 53/84 [08:20<04:23, 8.49s/it]
64%|██████▍ | 54/84 [08:26<03:46, 7.55s/it]
65%|██████▌ | 55/84 [08:49<05:57, 12.32s/it]
67%|██████▋ | 56/84 [08:53<04:31, 9.71s/it]
68%|██████▊ | 57/84 [09:00<04:00, 8.89s/it]
69%|██████▉ | 58/84 [09:06<03:29, 8.07s/it]
70%|███████ | 59/84 [09:12<03:09, 7.57s/it]
71%|███████▏ | 60/84 [09:18<02:49, 7.05s/it]
73%|███████▎ | 61/84 [09:23<02:27, 6.43s/it]
74%|███████▍ | 62/84 [09:31<02:28, 6.76s/it]
75%|███████▌ | 63/84 [09:39<02:31, 7.20s/it]
76%|███████▌ | 64/84 [09:47<02:30, 7.55s/it]
77%|███████▋ | 65/84 [10:11<03:58, 12.53s/it]
79%|███████▊ | 66/84 [10:16<03:02, 10.13s/it]
80%|███████▉ | 67/84 [10:24<02:39, 9.40s/it]
81%|████████ | 68/84 [10:33<02:29, 9.34s/it]
82%|████████▏ | 69/84 [10:39<02:06, 8.43s/it]
83%|████████▎ | 70/84 [10:47<01:56, 8.29s/it]
85%|████████▍ | 71/84 [10:52<01:34, 7.23s/it]
86%|████████▌ | 72/84 [10:58<01:23, 6.98s/it]
87%|████████▋ | 73/84 [11:07<01:21, 7.45s/it]
88%|████████▊ | 74/84 [11:15<01:16, 7.61s/it]
89%|████████▉ | 75/84 [11:22<01:08, 7.64s/it]
90%|█████████ | 76/84 [11:31<01:03, 8.00s/it]
92%|█████████▏| 77/84 [11:37<00:51, 7.40s/it]
93%|█████████▎| 78/84 [11:45<00:45, 7.55s/it]
94%|█████████▍| 79/84 [11:54<00:39, 7.88s/it]
95%|█████████▌| 80/84 [12:00<00:29, 7.42s/it]
96%|█████████▋| 81/84 [12:08<00:22, 7.61s/it]
98%|█████████▊| 82/84 [12:17<00:16, 8.06s/it]
99%|█████████▉| 83/84 [12:24<00:07, 7.75s/it]
100%|██████████| 84/84 [12:48<00:00, 12.52s/it]
{'eval_loss': 0.0, 'eval_runtime': 775.7291, 'eval_samples_per_second': 0.645, 'eval_steps_per_second': 0.108, 'eval_num_tokens': 144058587.0, 'eval_completions/mean_length': 641.6131166730609, 'eval_completions/min_length': 266.4761904761905, 'eval_completions/max_length': 1247.7380952380952, 'eval_completions/clipped_ratio': 0.027777778605620067, 'eval_completions/mean_terminated_length': 572.130375453404, 'eval_completions/min_terminated_length': 266.4761904761905, 'eval_completions/max_terminated_length': 973.8333333333334, 'eval_rewards/reward_correctness/mean': 0.7301587299222038, 'eval_rewards/reward_correctness/std': 0.4032438120671681, 'eval_reward': 0.7301587504999978, 'eval_reward_std': 0.4032438102932203, 'eval_frac_reward_zero_std': 1.0, 'eval_sampling/sampling_logp_difference/mean': 0.03718282767970647, 'eval_sampling/sampling_logp_difference/max': 2.4407017798650834, 'eval_sampling/importance_sampling_ratio/min': 0.00017360338998916516, 'eval_sampling/importance_sampling_ratio/mean': 0.07854402060807583, 'eval_sampling/importance_sampling_ratio/max': 0.35592690231056795, 'eval_entropy': 0.14853831895050548, 'eval_clip_ratio/low_mean': 0.0, 'eval_clip_ratio/low_min': 0.0, 'eval_clip_ratio/high_mean': 0.0, 'eval_clip_ratio/high_max': 0.0, 'eval_clip_ratio/region_mean': 0.0, 'epoch': 1.47}
74%|███████▎ | 100/136 [10:46:26<3:15:33, 325.92s/it]
100%|██████████| 84/84 [12:48<00:00, 12.52s/it]
[2026-08-13 09:17:06] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,90)
74%|███████▍ | 101/136 [10:51:49<5:25:22, 557.79s/it]
{'loss': -0.0316, 'grad_norm': 0.06088990844615045, 'learning_rate': 7.726298351471607e-07, 'num_tokens': 145506620.0, 'completions/mean_length': 844.6907958984375, 'completions/min_length': 164.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.095703125, 'completions/mean_terminated_length': 609.3268432617188, 'completions/min_terminated_length': 164.0, 'completions/max_terminated_length': 2770.0, 'rewards/reward_correctness/mean': 0.603515625, 'rewards/reward_correctness/std': 0.48932647705078125, 'reward': 0.603515625, 'reward_std': 0.48932644724845886, 'frac_reward_zero_std': 0.5234375, 'sampling/sampling_logp_difference/mean': 0.02406127005815506, 'sampling/sampling_logp_difference/max': 1.7813924551010132, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5787551999092102, 'sampling/importance_sampling_ratio/max': 2.9502859115600586, 'entropy': 0.39180235320236534, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 315.4343173727393, 'epoch': 1.49}
74%|███████▍ | 101/136 [10:51:49<5:25:22, 557.79s/it][2026-08-13 09:22:25] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,288)
[2026-08-13 09:22:27] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\)' @(1,2)
[2026-08-13 09:22:29] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,1836)
75%|███████▌ | 102/136 [10:57:06<4:35:07, 485.51s/it]
{'loss': -0.0336, 'grad_norm': 0.04069827271561709, 'learning_rate': 7.482787989539021e-07, 'num_tokens': 147049660.0, 'completions/mean_length': 902.7318115234375, 'completions/min_length': 159.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1165364608168602, 'completions/mean_terminated_length': 617.4561767578125, 'completions/min_terminated_length': 159.0, 'completions/max_terminated_length': 2674.0, 'rewards/reward_correctness/mean': 0.6106770634651184, 'rewards/reward_correctness/std': 0.48775553703308105, 'reward': 0.6106771230697632, 'reward_std': 0.48775550723075867, 'frac_reward_zero_std': 0.5546875, 'sampling/sampling_logp_difference/mean': 0.026487788185477257, 'sampling/sampling_logp_difference/max': 3.696782350540161, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5489984154701233, 'sampling/importance_sampling_ratio/max': 2.9973056316375732, 'entropy': 0.4538170258747414, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 315.7598335049115, 'epoch': 1.5}
75%|███████▌ | 102/136 [10:57:06<4:35:07, 485.51s/it][2026-08-13 09:27:37] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "frac" @(1,1)
Open LaTeX blocks:
@(1,1) arguments of macro "frac"
[2026-08-13 09:27:37] WARNING __init__.py:1389: WARNING: Error in configuration: macro '\frac' failed its substitution!
[2026-08-13 09:27:38] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "frac" @(1,260)
Open LaTeX blocks:
@(1,260) arguments of macro "frac"
[2026-08-13 09:27:38] WARNING __init__.py:1389: WARNING: Error in configuration: macro '\frac' failed its substitution!
[2026-08-13 09:27:39] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "frac" @(1,2)
Open LaTeX blocks:
@(1,2) arguments of macro "frac"
[2026-08-13 09:27:39] WARNING __init__.py:1389: WARNING: Error in configuration: macro '\frac' failed its substitution!
76%|███████▌ | 103/136 [11:02:11<3:57:09, 431.19s/it]
{'loss': -0.0074, 'grad_norm': 0.0751106349162355, 'learning_rate': 7.244463342316648e-07, 'num_tokens': 148463419.0, 'completions/mean_length': 836.509765625, 'completions/min_length': 129.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.095703125, 'completions/mean_terminated_length': 599.9243774414062, 'completions/min_terminated_length': 129.0, 'completions/max_terminated_length': 2178.0, 'rewards/reward_correctness/mean': 0.6458333134651184, 'rewards/reward_correctness/std': 0.4784157872200012, 'reward': 0.6458333730697632, 'reward_std': 0.47841572761535645, 'frac_reward_zero_std': 0.546875, 'sampling/sampling_logp_difference/mean': 0.02271461859345436, 'sampling/sampling_logp_difference/max': 1.7048001289367676, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.6221364736557007, 'sampling/importance_sampling_ratio/max': 2.998680353164673, 'entropy': 0.38416015822440386, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 303.2456877552904, 'epoch': 1.51}
76%|███████▌ | 103/136 [11:02:11<3:57:09, 431.19s/it][2026-08-13 09:32:36] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,1)
[2026-08-13 09:32:36] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,1)
[2026-08-13 09:32:36] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Bad \end macro: expected {environmentname} @(1,28)
[2026-08-13 09:32:39] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,48)
[2026-08-13 09:32:39] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,250)
76%|███████▋ | 104/136 [11:07:24<3:31:01, 395.67s/it]
{'loss': -0.03, 'grad_norm': 0.04814880690491371, 'learning_rate': 7.011461468091183e-07, 'num_tokens': 149915332.0, 'completions/mean_length': 850.232421875, 'completions/min_length': 158.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.0924479216337204, 'completions/mean_terminated_length': 624.4139404296875, 'completions/min_terminated_length': 158.0, 'completions/max_terminated_length': 1976.0, 'rewards/reward_correctness/mean': 0.603515625, 'rewards/reward_correctness/std': 0.48932647705078125, 'reward': 0.603515625, 'reward_std': 0.48932647705078125, 'frac_reward_zero_std': 0.5234375, 'sampling/sampling_logp_difference/mean': 0.023774094879627228, 'sampling/sampling_logp_difference/max': 3.655331611633301, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5602810978889465, 'sampling/importance_sampling_ratio/max': 2.9855146408081055, 'entropy': 0.3963266082573682, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 311.6580576309934, 'epoch': 1.53}
76%|███████▋ | 104/136 [11:07:24<3:31:01, 395.67s/it]
77%|███████▋ | 105/136 [11:12:18<3:08:45, 365.33s/it]
{'loss': -0.0055, 'grad_norm': 0.05765692003692507, 'learning_rate': 6.783916364072101e-07, 'num_tokens': 151347980.0, 'completions/mean_length': 838.3307495117188, 'completions/min_length': 155.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.0944010466337204, 'completions/mean_terminated_length': 605.4895629882812, 'completions/min_terminated_length': 155.0, 'completions/max_terminated_length': 2853.0, 'rewards/reward_correctness/mean': 0.6145833134651184, 'rewards/reward_correctness/std': 0.48685210943222046, 'reward': 0.6145833730697632, 'reward_std': 0.48685207962989807, 'frac_reward_zero_std': 0.484375, 'sampling/sampling_logp_difference/mean': 0.02304181084036827, 'sampling/sampling_logp_difference/max': 1.5441890954971313, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5794587731361389, 'sampling/importance_sampling_ratio/max': 2.982243061065674, 'entropy': 0.3826165738864802, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 293.3862261297181, 'epoch': 1.54}
77%|███████▋ | 105/136 [11:12:18<3:08:45, 365.33s/it][2026-08-13 09:42:39] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,121)
[2026-08-13 09:42:39] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,416)
78%|███████▊ | 106/136 [11:17:05<2:50:55, 341.85s/it]
{'loss': -0.0204, 'grad_norm': 0.0638999919867717, 'learning_rate': 6.561958889331121e-07, 'num_tokens': 152743628.0, 'completions/mean_length': 823.7578125, 'completions/min_length': 130.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.0787760466337204, 'completions/mean_terminated_length': 631.5053100585938, 'completions/min_terminated_length': 130.0, 'completions/max_terminated_length': 2716.0, 'rewards/reward_correctness/mean': 0.6484375, 'rewards/reward_correctness/std': 0.4776136875152588, 'reward': 0.6484375, 'reward_std': 0.4776136577129364, 'frac_reward_zero_std': 0.53125, 'sampling/sampling_logp_difference/mean': 0.021682944148778915, 'sampling/sampling_logp_difference/max': 1.9093666076660156, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.6033360958099365, 'sampling/importance_sampling_ratio/max': 2.961015462875366, 'entropy': 0.3542632268508896, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 285.9305932777934, 'epoch': 1.56}
78%|███████▊ | 106/136 [11:17:05<2:50:55, 341.85s/it][2026-08-13 09:47:46] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,1067)
79%|███████▊ | 107/136 [11:22:41<2:44:18, 339.95s/it]
{'loss': -0.0165, 'grad_norm': 0.04555968292142207, 'learning_rate': 6.345716689546361e-07, 'num_tokens': 154186080.0, 'completions/mean_length': 837.1588745117188, 'completions/min_length': 132.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.091796875, 'completions/mean_terminated_length': 614.7268676757812, 'completions/min_terminated_length': 132.0, 'completions/max_terminated_length': 2778.0, 'rewards/reward_correctness/mean': 0.6276041865348816, 'rewards/reward_correctness/std': 0.4836004674434662, 'reward': 0.6276041865348816, 'reward_std': 0.4836004674434662, 'frac_reward_zero_std': 0.5390625, 'sampling/sampling_logp_difference/mean': 0.04087618738412857, 'sampling/sampling_logp_difference/max': 22.87156105041504, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.4639313220977783, 'sampling/importance_sampling_ratio/max': 2.9571053981781006, 'entropy': 0.4021087958244607, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 334.3754716110416, 'epoch': 1.57}
79%|███████▊ | 107/136 [11:22:41<2:44:18, 339.95s/it][2026-08-13 09:53:23] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Bad \end macro: expected {environmentname} @(1,0)
79%|███████▉ | 108/136 [11:28:23<2:39:01, 340.75s/it]
{'loss': -0.0359, 'grad_norm': 0.044641183127054146, 'learning_rate': 6.135314123594495e-07, 'num_tokens': 155724383.0, 'completions/mean_length': 895.3118896484375, 'completions/min_length': 154.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1145833358168602, 'completions/mean_terminated_length': 614.1749877929688, 'completions/min_terminated_length': 154.0, 'completions/max_terminated_length': 2424.0, 'rewards/reward_correctness/mean': 0.5911458134651184, 'rewards/reward_correctness/std': 0.49178236722946167, 'reward': 0.5911458730697632, 'reward_std': 0.4917823374271393, 'frac_reward_zero_std': 0.5078125, 'sampling/sampling_logp_difference/mean': 0.03958186134696007, 'sampling/sampling_logp_difference/max': 13.149711608886719, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.47802260518074036, 'sampling/importance_sampling_ratio/max': 2.9951727390289307, 'entropy': 0.41992489667609334, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 341.46006203070283, 'epoch': 1.59}
79%|███████▉ | 108/136 [11:28:23<2:39:01, 340.75s/it][2026-08-13 09:58:55] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\)' @(1,3)
[2026-08-13 09:58:57] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,317)
[2026-08-13 09:58:57] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,440)
[2026-08-13 09:58:57] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,1940)
[2026-08-13 09:58:57] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,2764)
80%|████████ | 109/136 [11:33:34<2:29:14, 331.65s/it]
{'loss': -0.0142, 'grad_norm': 0.044574284241470793, 'learning_rate': 5.93087219203313e-07, 'num_tokens': 157361190.0, 'completions/mean_length': 961.5436401367188, 'completions/min_length': 126.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1197916716337204, 'completions/mean_terminated_length': 674.3380126953125, 'completions/min_terminated_length': 126.0, 'completions/max_terminated_length': 3058.0, 'rewards/reward_correctness/mean': 0.5852864384651184, 'rewards/reward_correctness/std': 0.4928329885005951, 'reward': 0.5852864980697632, 'reward_std': 0.4928329288959503, 'frac_reward_zero_std': 0.484375, 'sampling/sampling_logp_difference/mean': 0.025156300514936447, 'sampling/sampling_logp_difference/max': 1.2640998363494873, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5615074634552002, 'sampling/importance_sampling_ratio/max': 2.992027521133423, 'entropy': 0.44594370829872787, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 309.27131781354547, 'epoch': 1.6}
80%|████████ | 109/136 [11:33:34<2:29:14, 331.65s/it][2026-08-13 10:04:12] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,433)
[2026-08-13 10:04:13] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,365)
[2026-08-13 10:04:13] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,162)
81%|████████ | 110/136 [11:39:01<2:23:07, 330.30s/it]
{'loss': -0.024, 'grad_norm': 0.05142730527423491, 'learning_rate': 5.732508467514508e-07, 'num_tokens': 158932678.0, 'completions/mean_length': 924.3776245117188, 'completions/min_length': 141.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.12109375, 'completions/mean_terminated_length': 628.4829711914062, 'completions/min_terminated_length': 141.0, 'completions/max_terminated_length': 2381.0, 'rewards/reward_correctness/mean': 0.6119791865348816, 'rewards/reward_correctness/std': 0.48745808005332947, 'reward': 0.6119791865348816, 'reward_std': 0.4874580204486847, 'frac_reward_zero_std': 0.484375, 'sampling/sampling_logp_difference/mean': 0.0254482664167881, 'sampling/sampling_logp_difference/max': 3.0439860820770264, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.6046624779701233, 'sampling/importance_sampling_ratio/max': 2.987983465194702, 'entropy': 0.432338533049915, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 326.00678399251774, 'epoch': 1.62}
81%|████████ | 110/136 [11:39:01<2:23:07, 330.30s/it]
0%| | 0/84 [00:00<?, ?it/s]
2%|▏ | 2/84 [00:08<05:54, 4.32s/it]
4%|▎ | 3/84 [00:13<06:22, 4.72s/it]
5%|▍ | 4/84 [00:20<07:19, 5.50s/it]
6%|▌ | 5/84 [00:44<15:35, 11.84s/it]
7%|▋ | 6/84 [00:53<14:07, 10.86s/it]
8%|▊ | 7/84 [01:01<12:55, 10.07s/it]
10%|▉ | 8/84 [01:06<10:35, 8.37s/it]
11%|█ | 9/84 [01:12<09:44, 7.79s/it]
12%|█▏ | 10/84 [01:17<08:33, 6.95s/it]
13%|█▎ | 11/84 [01:26<08:59, 7.39s/it]
14%|█▍ | 12/84 [01:32<08:31, 7.10s/it]
15%|█▌ | 13/84 [01:47<11:11, 9.46s/it]
17%|█▋ | 14/84 [01:56<10:39, 9.14s/it]
18%|█▊ | 15/84 [02:02<09:27, 8.22s/it]
19%|█▉ | 16/84 [02:08<08:41, 7.67s/it]
20%|██ | 17/84 [02:15<08:21, 7.49s/it]
21%|██▏ | 18/84 [02:20<07:12, 6.56s/it]
23%|██▎ | 19/84 [02:27<07:21, 6.79s/it]
24%|██▍ | 20/84 [02:35<07:49, 7.34s/it]
25%|██▌ | 21/84 [02:44<08:04, 7.68s/it]
26%|██▌ | 22/84 [02:49<07:02, 6.82s/it]
27%|██▋ | 23/84 [02:58<07:34, 7.45s/it]
29%|██▊ | 24/84 [03:06<07:45, 7.75s/it]
30%|██▉ | 25/84 [03:12<07:05, 7.22s/it]
31%|███ | 26/84 [03:17<06:14, 6.45s/it]
32%|███▏ | 27/84 [03:24<06:18, 6.65s/it]
33%|███▎ | 28/84 [03:29<05:54, 6.33s/it]
35%|███▍ | 29/84 [03:53<10:33, 11.51s/it]
36%|███▌ | 30/84 [04:00<09:05, 10.10s/it]
37%|███▋ | 31/84 [04:07<08:11, 9.28s/it]
38%|███▊ | 32/84 [04:12<06:51, 7.92s/it]
39%|███▉ | 33/84 [04:35<10:41, 12.59s/it]
40%|████ | 34/84 [04:43<09:07, 10.96s/it]
42%|████▏ | 35/84 [05:06<12:02, 14.75s/it]
43%|████▎ | 36/84 [05:12<09:45, 12.20s/it]
44%|████▍ | 37/84 [05:17<07:45, 9.90s/it]
45%|████▌ | 38/84 [05:24<06:57, 9.07s/it]
46%|████▋ | 39/84 [05:32<06:26, 8.58s/it]
48%|████▊ | 40/84 [05:55<09:37, 13.12s/it]
49%|████▉ | 41/84 [06:01<07:54, 11.03s/it]
50%|█████ | 42/84 [06:13<07:49, 11.19s/it]
51%|█████ | 43/84 [06:36<10:09, 14.86s/it]
52%|█████▏ | 44/84 [06:43<08:09, 12.24s/it]
54%|█████▎ | 45/84 [07:07<10:14, 15.77s/it]
55%|█████▍ | 46/84 [07:14<08:22, 13.21s/it]
56%|█████▌ | 47/84 [07:21<06:57, 11.27s/it]
57%|█████▋ | 48/84 [07:29<06:18, 10.53s/it]
58%|█████▊ | 49/84 [07:35<05:15, 9.01s/it]
60%|█████▉ | 50/84 [07:59<07:37, 13.46s/it]
61%|██████ | 51/84 [08:06<06:24, 11.64s/it]
62%|██████▏ | 52/84 [08:14<05:39, 10.62s/it]
63%|██████▎ | 53/84 [08:19<04:34, 8.86s/it]
64%|██████▍ | 54/84 [08:23<03:41, 7.37s/it]
65%|██████▌ | 55/84 [08:46<05:54, 12.21s/it]
67%|██████▋ | 56/84 [08:50<04:29, 9.63s/it]
68%|██████▊ | 57/84 [09:00<04:18, 9.58s/it]
69%|██████▉ | 58/84 [09:03<03:24, 7.87s/it]
70%|███████ | 59/84 [09:13<03:29, 8.38s/it]
71%|███████▏ | 60/84 [09:21<03:16, 8.19s/it]
73%|███████▎ | 61/84 [09:26<02:47, 7.29s/it]
74%|███████▍ | 62/84 [09:35<02:49, 7.70s/it]
75%|███████▌ | 63/84 [09:45<02:56, 8.40s/it]
76%|███████▌ | 64/84 [09:54<02:54, 8.72s/it]
77%|███████▋ | 65/84 [10:04<02:50, 8.95s/it]
79%|███████▊ | 66/84 [10:11<02:30, 8.38s/it]
80%|███████▉ | 67/84 [10:16<02:07, 7.51s/it]
81%|████████ | 68/84 [10:26<02:10, 8.15s/it]
82%|████████▏ | 69/84 [10:32<01:54, 7.65s/it]
83%|████████▎ | 70/84 [10:41<01:50, 7.87s/it]
85%|████████▍ | 71/84 [10:45<01:29, 6.88s/it]
86%|████████▌ | 72/84 [10:51<01:18, 6.53s/it]
87%|████████▋ | 73/84 [10:57<01:10, 6.42s/it]
88%|████████▊ | 74/84 [11:04<01:04, 6.47s/it]
89%|████████▉ | 75/84 [11:11<01:01, 6.82s/it]
90%|█████████ | 76/84 [11:24<01:07, 8.47s/it]
92%|█████████▏| 77/84 [11:30<00:54, 7.81s/it]
93%|█████████▎| 78/84 [11:38<00:47, 7.92s/it]
94%|█████████▍| 79/84 [11:43<00:34, 6.95s/it]
95%|█████████▌| 80/84 [11:52<00:30, 7.52s/it]
96%|█████████▋| 81/84 [11:57<00:21, 7.03s/it]
98%|█████████▊| 82/84 [12:08<00:16, 8.19s/it]
99%|█████████▉| 83/84 [12:16<00:07, 7.99s/it]
100%|██████████| 84/84 [12:39<00:00, 12.69s/it]
{'eval_loss': 0.0, 'eval_runtime': 767.1044, 'eval_samples_per_second': 0.652, 'eval_steps_per_second': 0.11, 'eval_num_tokens': 158932678.0, 'eval_completions/mean_length': 626.029779343378, 'eval_completions/min_length': 269.6309523809524, 'eval_completions/max_length': 1239.75, 'eval_completions/clipped_ratio': 0.023809524519102915, 'eval_completions/mean_terminated_length': 567.4938645135788, 'eval_completions/min_terminated_length': 269.6309523809524, 'eval_completions/max_terminated_length': 1011.1071428571429, 'eval_rewards/reward_correctness/mean': 0.7182539676626524, 'eval_rewards/reward_correctness/std': 0.41254346995126634, 'eval_reward': 0.7182539875308672, 'eval_reward_std': 0.4125434688868977, 'eval_frac_reward_zero_std': 1.0, 'eval_sampling/sampling_logp_difference/mean': 0.037226093312104545, 'eval_sampling/sampling_logp_difference/max': 2.3605878239586238, 'eval_sampling/importance_sampling_ratio/min': 0.00021769897526041198, 'eval_sampling/importance_sampling_ratio/mean': 0.08027123272439071, 'eval_sampling/importance_sampling_ratio/max': 0.3713289199783633, 'eval_entropy': 0.14750546021830468, 'eval_clip_ratio/low_mean': 0.0, 'eval_clip_ratio/low_min': 0.0, 'eval_clip_ratio/high_mean': 0.0, 'eval_clip_ratio/high_max': 0.0, 'eval_clip_ratio/region_mean': 0.0, 'epoch': 1.62}
81%|████████ | 110/136 [11:51:48<2:23:07, 330.30s/it]
100%|██████████| 84/84 [12:40<00:00, 12.69s/it]
[2026-08-13 10:22:26] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,13)
82%|████████▏ | 111/136 [11:57:01<3:51:24, 555.38s/it]
{'loss': -0.0211, 'grad_norm': 0.04335790136718892, 'learning_rate': 5.540337027170566e-07, 'num_tokens': 160358860.0, 'completions/mean_length': 836.53515625, 'completions/min_length': 169.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.0904947966337204, 'completions/mean_terminated_length': 614.1087646484375, 'completions/min_terminated_length': 169.0, 'completions/max_terminated_length': 1792.0, 'rewards/reward_correctness/mean': 0.6647135615348816, 'rewards/reward_correctness/std': 0.4722442626953125, 'reward': 0.6647135615348816, 'reward_std': 0.4722442328929901, 'frac_reward_zero_std': 0.53125, 'sampling/sampling_logp_difference/mean': 0.022612713277339935, 'sampling/sampling_logp_difference/max': 1.4412472248077393, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5987992882728577, 'sampling/importance_sampling_ratio/max': 2.9934427738189697, 'entropy': 0.36708237300626934, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 305.83953939471394, 'epoch': 1.63}
82%|████████▏ | 111/136 [11:57:01<3:51:24, 555.38s/it][2026-08-13 10:27:35] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,12)
[2026-08-13 10:27:35] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,193)
[2026-08-13 10:27:35] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,2)
[2026-08-13 10:27:35] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,4)
[2026-08-13 10:27:36] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,18)
[2026-08-13 10:27:36] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,314)
82%|████████▏ | 112/136 [12:02:10<3:12:34, 481.45s/it]
{'loss': -0.0358, 'grad_norm': 0.04585958475342479, 'learning_rate': 5.354468387008236e-07, 'num_tokens': 161854512.0, 'completions/mean_length': 878.3255615234375, 'completions/min_length': 148.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1041666716337204, 'completions/mean_terminated_length': 623.2470703125, 'completions/min_terminated_length': 148.0, 'completions/max_terminated_length': 2685.0, 'rewards/reward_correctness/mean': 0.6119791865348816, 'rewards/reward_correctness/std': 0.48745808005332947, 'reward': 0.6119791865348816, 'reward_std': 0.4874580502510071, 'frac_reward_zero_std': 0.5234375, 'sampling/sampling_logp_difference/mean': 0.024161145091056824, 'sampling/sampling_logp_difference/max': 1.0098832845687866, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5624591112136841, 'sampling/importance_sampling_ratio/max': 2.983163833618164, 'entropy': 0.4083844933193177, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 307.8114897273481, 'epoch': 1.65}
82%|████████▏ | 112/136 [12:02:10<3:12:34, 481.45s/it]
83%|████████▎ | 113/136 [12:07:12<2:43:55, 427.61s/it]
{'loss': -0.0253, 'grad_norm': 0.053624327974202725, 'learning_rate': 5.175009438352725e-07, 'num_tokens': 163208983.0, 'completions/mean_length': 788.8092651367188, 'completions/min_length': 154.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.0768229216337204, 'completions/mean_terminated_length': 598.8201904296875, 'completions/min_terminated_length': 154.0, 'completions/max_terminated_length': 2183.0, 'rewards/reward_correctness/mean': 0.65234375, 'rewards/reward_correctness/std': 0.4763812720775604, 'reward': 0.65234375, 'reward_std': 0.4763812720775604, 'frac_reward_zero_std': 0.546875, 'sampling/sampling_logp_difference/mean': 0.021032704040408134, 'sampling/sampling_logp_difference/max': 1.14524507522583, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.606112539768219, 'sampling/importance_sampling_ratio/max': 2.9941608905792236, 'entropy': 0.34125655790558085, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 300.7747645783238, 'epoch': 1.66}
83%|████████▎ | 113/136 [12:07:12<2:43:55, 427.61s/it][2026-08-13 10:37:52] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\)' @(1,0)
84%|████████▍ | 114/136 [12:12:26<2:24:17, 393.54s/it]
{'loss': -0.0191, 'grad_norm': 0.049732445330309516, 'learning_rate': 5.002063386375302e-07, 'num_tokens': 164723246.0, 'completions/mean_length': 895.5358276367188, 'completions/min_length': 150.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.09375, 'completions/mean_terminated_length': 670.3843383789062, 'completions/min_terminated_length': 150.0, 'completions/max_terminated_length': 2575.0, 'rewards/reward_correctness/mean': 0.6321614384651184, 'rewards/reward_correctness/std': 0.4823741614818573, 'reward': 0.6321614980697632, 'reward_std': 0.4823741316795349, 'frac_reward_zero_std': 0.4296875, 'sampling/sampling_logp_difference/mean': 0.021676061674952507, 'sampling/sampling_logp_difference/max': 0.9949932098388672, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5702483057975769, 'sampling/importance_sampling_ratio/max': 2.99418044090271, 'entropy': 0.3677748629706912, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 312.8656594133936, 'epoch': 1.68}
84%|████████▍ | 114/136 [12:12:26<2:24:17, 393.54s/it]
85%|████████▍ | 115/136 [12:17:34<2:08:42, 367.76s/it]
{'loss': -0.0117, 'grad_norm': 0.0583397216318952, 'learning_rate': 4.835729690740971e-07, 'num_tokens': 166113511.0, 'completions/mean_length': 818.5814208984375, 'completions/min_length': 161.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.0930989608168602, 'completions/mean_terminated_length': 587.254150390625, 'completions/min_terminated_length': 161.0, 'completions/max_terminated_length': 2274.0, 'rewards/reward_correctness/mean': 0.6751301884651184, 'rewards/reward_correctness/std': 0.4684787094593048, 'reward': 0.6751302480697632, 'reward_std': 0.4684787094593048, 'frac_reward_zero_std': 0.5625, 'sampling/sampling_logp_difference/mean': 0.02168840914964676, 'sampling/sampling_logp_difference/max': 5.4962921142578125, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.6631502509117126, 'sampling/importance_sampling_ratio/max': 2.9827520847320557, 'entropy': 0.35392041283193976, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 306.4326215046458, 'epoch': 1.69}
85%|████████▍ | 115/136 [12:17:34<2:08:42, 367.76s/it]
85%|████████▌ | 116/136 [12:22:46<1:56:59, 351.00s/it]
{'loss': -0.0112, 'grad_norm': 0.05383767546988474, 'learning_rate': 4.6761040084101373e-07, 'num_tokens': 167656493.0, 'completions/mean_length': 907.9440307617188, 'completions/min_length': 173.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1145833358168602, 'completions/mean_terminated_length': 627.8897094726562, 'completions/min_terminated_length': 173.0, 'completions/max_terminated_length': 2316.0, 'rewards/reward_correctness/mean': 0.5774739384651184, 'rewards/reward_correctness/std': 0.494122177362442, 'reward': 0.5774739980697632, 'reward_std': 0.4941222071647644, 'frac_reward_zero_std': 0.5078125, 'sampling/sampling_logp_difference/mean': 0.025177408009767532, 'sampling/sampling_logp_difference/max': 2.678741693496704, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5863032341003418, 'sampling/importance_sampling_ratio/max': 2.9951207637786865, 'entropy': 0.44132739264750853, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 310.71802368201315, 'epoch': 1.71}
85%|████████▌ | 116/136 [12:22:46<1:56:59, 351.00s/it]
86%|████████▌ | 117/136 [12:27:53<1:46:59, 337.87s/it]
{'loss': -0.0322, 'grad_norm': 0.05001580981100338, 'learning_rate': 4.523278138627179e-07, 'num_tokens': 169211412.0, 'completions/mean_length': 927.2233276367188, 'completions/min_length': 163.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1243489608168602, 'completions/mean_terminated_length': 622.8141479492188, 'completions/min_terminated_length': 163.0, 'completions/max_terminated_length': 2546.0, 'rewards/reward_correctness/mean': 0.591796875, 'rewards/reward_correctness/std': 0.4916611909866333, 'reward': 0.591796875, 'reward_std': 0.4916611611843109, 'frac_reward_zero_std': 0.5234375, 'sampling/sampling_logp_difference/mean': 0.026477819308638573, 'sampling/sampling_logp_difference/max': 1.4523508548736572, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5662473440170288, 'sampling/importance_sampling_ratio/max': 2.994056224822998, 'entropy': 0.4399852986098267, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 305.99031915096566, 'epoch': 1.72}
86%|████████▌ | 117/136 [12:27:53<1:46:59, 337.87s/it][2026-08-13 10:58:40] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,1132)
[2026-08-13 10:58:40] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,122)
[2026-08-13 10:58:40] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,5)
[2026-08-13 10:58:41] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,57)
[2026-08-13 10:58:41] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,283)
[2026-08-13 10:58:44] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,22)
87%|████████▋ | 118/136 [12:33:47<1:42:45, 342.53s/it]
{'loss': -0.0445, 'grad_norm': 0.05290927155308445, 'learning_rate': 4.377339970127567e-07, 'num_tokens': 170908298.0, 'completions/mean_length': 996.1654052734375, 'completions/min_length': 124.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1295572966337204, 'completions/mean_terminated_length': 689.387451171875, 'completions/min_terminated_length': 124.0, 'completions/max_terminated_length': 2326.0, 'rewards/reward_correctness/mean': 0.5748698115348816, 'rewards/reward_correctness/std': 0.4945237338542938, 'reward': 0.5748698115348816, 'reward_std': 0.4945237338542938, 'frac_reward_zero_std': 0.4609375, 'sampling/sampling_logp_difference/mean': 0.04102567210793495, 'sampling/sampling_logp_difference/max': 27.493650436401367, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.4558072090148926, 'sampling/importance_sampling_ratio/max': 2.981295108795166, 'entropy': 0.4670035591116175, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 352.26924694515765, 'epoch': 1.74}
87%|████████▋ | 118/136 [12:33:47<1:42:45, 342.53s/it][2026-08-13 11:04:42] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,0)
[2026-08-13 11:04:42] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,16)
[2026-08-13 11:04:47] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,62)
88%|████████▊ | 119/136 [12:39:15<1:35:52, 338.40s/it]
{'loss': -0.0303, 'grad_norm': 0.049328964216854024, 'learning_rate': 4.238373430593857e-07, 'num_tokens': 172437473.0, 'completions/mean_length': 912.095703125, 'completions/min_length': 143.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1126302108168602, 'completions/mean_terminated_length': 637.9479370117188, 'completions/min_terminated_length': 143.0, 'completions/max_terminated_length': 2728.0, 'rewards/reward_correctness/mean': 0.646484375, 'rewards/reward_correctness/std': 0.4782167077064514, 'reward': 0.646484375, 'reward_std': 0.47821667790412903, 'frac_reward_zero_std': 0.4921875, 'sampling/sampling_logp_difference/mean': 0.02208506129682064, 'sampling/sampling_logp_difference/max': 1.0320615768432617, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5879413485527039, 'sampling/importance_sampling_ratio/max': 2.9586801528930664, 'entropy': 0.38580389926210046, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 327.5872627142817, 'epoch': 1.75}
88%|████████▊ | 119/136 [12:39:15<1:35:52, 338.40s/it][2026-08-13 11:09:59] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,356)
[2026-08-13 11:10:00] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,83)
88%|████████▊ | 120/136 [12:44:43<1:29:21, 335.08s/it]
{'loss': -0.0215, 'grad_norm': 0.055668120244077884, 'learning_rate': 4.1064584383896707e-07, 'num_tokens': 174037783.0, 'completions/mean_length': 930.1575927734375, 'completions/min_length': 156.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1178385466337204, 'completions/mean_terminated_length': 644.065673828125, 'completions/min_terminated_length': 156.0, 'completions/max_terminated_length': 2830.0, 'rewards/reward_correctness/mean': 0.6236979365348816, 'rewards/reward_correctness/std': 0.4846150279045105, 'reward': 0.6236979365348816, 'reward_std': 0.4846149981021881, 'frac_reward_zero_std': 0.46875, 'sampling/sampling_logp_difference/mean': 0.024076789617538452, 'sampling/sampling_logp_difference/max': 6.9161248207092285, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5683541893959045, 'sampling/importance_sampling_ratio/max': 2.9956037998199463, 'entropy': 0.41531214263523, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 326.06551372399554, 'epoch': 1.76}
88%|████████▊ | 120/136 [12:44:43<1:29:21, 335.08s/it]
0%| | 0/84 [00:00<?, ?it/s]
2%|▏ | 2/84 [00:04<03:16, 2.39s/it]
4%|▎ | 3/84 [00:11<05:36, 4.15s/it]
5%|▍ | 4/84 [00:18<07:15, 5.44s/it]
6%|▌ | 5/84 [00:42<15:33, 11.81s/it]
7%|▋ | 6/84 [01:06<20:32, 15.80s/it]
8%|▊ | 7/84 [01:15<17:36, 13.72s/it]
10%|▉ | 8/84 [01:19<13:17, 10.49s/it]
11%|█ | 9/84 [01:27<12:17, 9.83s/it]
12%|█▏ | 10/84 [01:32<10:14, 8.30s/it]
13%|█▎ | 11/84 [01:40<09:48, 8.07s/it]
14%|█▍ | 12/84 [01:46<09:06, 7.59s/it]
15%|█▌ | 13/84 [02:10<14:40, 12.41s/it]
17%|█▋ | 14/84 [02:18<13:05, 11.22s/it]
18%|█▊ | 15/84 [02:27<11:58, 10.42s/it]
19%|█▉ | 16/84 [02:31<09:47, 8.64s/it]
20%|██ | 17/84 [02:37<08:51, 7.93s/it]
21%|██▏ | 18/84 [02:44<08:06, 7.38s/it]
23%|██▎ | 19/84 [02:54<08:53, 8.21s/it]
24%|██▍ | 20/84 [03:01<08:22, 7.85s/it]
25%|██▌ | 21/84 [03:07<07:43, 7.36s/it]
26%|██▌ | 22/84 [03:12<06:55, 6.69s/it]
27%|██▋ | 23/84 [03:17<06:13, 6.12s/it]
29%|██▊ | 24/84 [03:26<07:00, 7.01s/it]
30%|██▉ | 25/84 [03:33<06:58, 7.10s/it]
31%|███ | 26/84 [03:38<06:09, 6.36s/it]
32%|███▏ | 27/84 [03:47<06:49, 7.18s/it]
33%|███▎ | 28/84 [03:53<06:20, 6.79s/it]
35%|███▍ | 29/84 [04:02<06:54, 7.53s/it]
36%|███▌ | 30/84 [04:08<06:27, 7.18s/it]
37%|███▋ | 31/84 [04:32<10:40, 12.08s/it]
38%|███▊ | 32/84 [04:37<08:38, 9.97s/it]
39%|███▉ | 33/84 [04:44<07:41, 9.06s/it]
40%|████ | 34/84 [05:07<11:09, 13.39s/it]
42%|████▏ | 35/84 [05:31<13:25, 16.44s/it]
43%|████▎ | 36/84 [05:36<10:26, 13.05s/it]
44%|████▍ | 37/84 [05:41<08:19, 10.62s/it]
45%|████▌ | 38/84 [05:47<07:07, 9.30s/it]
46%|████▋ | 39/84 [05:56<06:49, 9.11s/it]
48%|████▊ | 40/84 [06:20<09:55, 13.54s/it]
49%|████▉ | 41/84 [06:25<07:57, 11.10s/it]
50%|█████ | 42/84 [06:33<07:00, 10.01s/it][2026-08-13 11:20:11] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "frac" @(1,1)
Open LaTeX blocks:
@(1,1) arguments of macro "frac"
[2026-08-13 11:20:11] WARNING __init__.py:1389: WARNING: Error in configuration: macro '\frac' failed its substitution!
51%|█████ | 43/84 [06:56<09:37, 14.09s/it]
52%|█████▏ | 44/84 [07:02<07:36, 11.42s/it]
54%|█████▎ | 45/84 [07:09<06:39, 10.25s/it]
55%|█████▍ | 46/84 [07:18<06:19, 9.98s/it]
56%|█████▌ | 47/84 [07:25<05:32, 8.98s/it]
57%|█████▋ | 48/84 [07:34<05:23, 9.00s/it]
58%|█████▊ | 49/84 [07:58<07:54, 13.55s/it]
60%|█████▉ | 50/84 [08:12<07:41, 13.57s/it]
61%|██████ | 51/84 [08:35<09:06, 16.55s/it]
62%|██████▏ | 52/84 [08:44<07:31, 14.11s/it]
63%|██████▎ | 53/84 [08:47<05:38, 10.92s/it]
64%|██████▍ | 54/84 [08:54<04:48, 9.61s/it]
65%|██████▌ | 55/84 [09:08<05:15, 10.89s/it]
67%|██████▋ | 56/84 [09:11<04:00, 8.59s/it]
68%|██████▊ | 57/84 [09:17<03:28, 7.73s/it]
69%|██████▉ | 58/84 [09:21<02:52, 6.62s/it]
70%|███████ | 59/84 [09:29<02:58, 7.12s/it]
71%|███████▏ | 60/84 [09:35<02:45, 6.89s/it]
73%|███████▎ | 61/84 [09:45<02:57, 7.71s/it]
74%|███████▍ | 62/84 [09:54<03:00, 8.19s/it]
75%|███████▌ | 63/84 [10:04<02:59, 8.55s/it]
76%|███████▌ | 64/84 [10:28<04:23, 13.20s/it]
77%|███████▋ | 65/84 [10:39<04:01, 12.72s/it]
79%|███████▊ | 66/84 [10:47<03:21, 11.20s/it]
80%|███████▉ | 67/84 [10:54<02:48, 9.91s/it]
81%|████████ | 68/84 [11:04<02:38, 9.94s/it]
82%|████████▏ | 69/84 [11:11<02:14, 8.99s/it]
83%|████████▎ | 70/84 [11:18<01:59, 8.51s/it]
85%|████████▍ | 71/84 [11:22<01:33, 7.23s/it]
86%|████████▌ | 72/84 [11:29<01:25, 7.09s/it]
87%|████████▋ | 73/84 [11:37<01:19, 7.26s/it]
88%|████████▊ | 74/84 [11:43<01:10, 7.08s/it]
89%|████████▉ | 75/84 [11:50<01:02, 6.89s/it]
90%|█████████ | 76/84 [11:59<00:59, 7.44s/it]
92%|█████████▏| 77/84 [12:05<00:49, 7.07s/it]
93%|█████████▎| 78/84 [12:14<00:46, 7.79s/it]
94%|█████████▍| 79/84 [12:25<00:44, 8.81s/it]
95%|█████████▌| 80/84 [12:49<00:52, 13.23s/it]
96%|█████████▋| 81/84 [13:13<00:49, 16.37s/it]
98%|█████████▊| 82/84 [13:23<00:29, 14.55s/it]
99%|█████████▉| 83/84 [13:31<00:12, 12.73s/it]
100%|██████████| 84/84 [13:36<00:00, 10.33s/it]
{'eval_loss': 0.0, 'eval_runtime': 822.9839, 'eval_samples_per_second': 0.608, 'eval_steps_per_second': 0.102, 'eval_num_tokens': 174037783.0, 'eval_completions/mean_length': 642.468273344494, 'eval_completions/min_length': 269.6904761904762, 'eval_completions/max_length': 1323.1785714285713, 'eval_completions/clipped_ratio': 0.027777778605620067, 'eval_completions/mean_terminated_length': 575.1000159127371, 'eval_completions/min_terminated_length': 269.6904761904762, 'eval_completions/max_terminated_length': 1018.6309523809524, 'eval_rewards/reward_correctness/mean': 0.7242063489698228, 'eval_rewards/reward_correctness/std': 0.41008960774966646, 'eval_reward': 0.7242063674188796, 'eval_reward_std': 0.41008960952361423, 'eval_frac_reward_zero_std': 1.0, 'eval_sampling/sampling_logp_difference/mean': 0.03615121694193, 'eval_sampling/sampling_logp_difference/max': 2.3760013296490623, 'eval_sampling/importance_sampling_ratio/min': 0.00024189939948790175, 'eval_sampling/importance_sampling_ratio/mean': 0.086530597872687, 'eval_sampling/importance_sampling_ratio/max': 0.4101124081748172, 'eval_entropy': 0.14365598683555922, 'eval_clip_ratio/low_mean': 0.0, 'eval_clip_ratio/low_min': 0.0, 'eval_clip_ratio/high_mean': 0.0, 'eval_clip_ratio/high_max': 0.0, 'eval_clip_ratio/region_mean': 0.0, 'epoch': 1.76}
88%|████████▊ | 120/136 [12:58:26<1:29:21, 335.08s/it]
100%|██████████| 84/84 [13:36<00:00, 10.33s/it]
[2026-08-13 11:29:01] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,318)
[2026-08-13 11:29:01] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,356)
[2026-08-13 11:29:03] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,311)
[2026-08-13 11:29:03] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "\" @(1,440)
Open LaTeX blocks:
@(1,440) arguments of macro "\"
89%|████████▉ | 121/136 [13:03:46<2:24:25, 577.69s/it]
{'loss': -0.0101, 'grad_norm': 0.043414378713222734, 'learning_rate': 3.9816708565993797e-07, 'num_tokens': 175493701.0, 'completions/mean_length': 862.91796875, 'completions/min_length': 107.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.10546875, 'completions/mean_terminated_length': 602.45849609375, 'completions/min_terminated_length': 107.0, 'completions/max_terminated_length': 2786.0, 'rewards/reward_correctness/mean': 0.6829426884651184, 'rewards/reward_correctness/std': 0.4654814898967743, 'reward': 0.6829427480697632, 'reward_std': 0.4654814600944519, 'frac_reward_zero_std': 0.546875, 'sampling/sampling_logp_difference/mean': 0.023906981572508812, 'sampling/sampling_logp_difference/max': 1.9039257764816284, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5633504390716553, 'sampling/importance_sampling_ratio/max': 2.988877534866333, 'entropy': 0.39749378041597083, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 313.3631044062786, 'epoch': 1.78}
89%|████████▉ | 121/136 [13:03:46<2:24:25, 577.69s/it][2026-08-13 11:34:17] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,732)
90%|████████▉ | 122/136 [13:08:55<1:55:57, 496.96s/it]
{'loss': -0.014, 'grad_norm': 0.067411019205484, 'learning_rate': 3.864082449399963e-07, 'num_tokens': 176947506.0, 'completions/mean_length': 852.6361083984375, 'completions/min_length': 158.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.0983072966337204, 'completions/mean_terminated_length': 610.8281860351562, 'completions/min_terminated_length': 158.0, 'completions/max_terminated_length': 2892.0, 'rewards/reward_correctness/mean': 0.65234375, 'rewards/reward_correctness/std': 0.4763812720775604, 'reward': 0.65234375, 'reward_std': 0.4763812720775604, 'frac_reward_zero_std': 0.5078125, 'sampling/sampling_logp_difference/mean': 0.0226468276232481, 'sampling/sampling_logp_difference/max': 1.482698917388916, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.6137019991874695, 'sampling/importance_sampling_ratio/max': 2.9987151622772217, 'entropy': 0.35477442893898115, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 307.41807026276365, 'epoch': 1.79}
90%|████████▉ | 122/136 [13:08:55<1:55:57, 496.96s/it][2026-08-13 11:39:24] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,25)
90%|█████████ | 123/136 [13:14:01<1:35:15, 439.69s/it]
{'loss': -0.0265, 'grad_norm': 0.0529649070159592, 'learning_rate': 3.753760840790081e-07, 'num_tokens': 178387007.0, 'completions/mean_length': 834.6829833984375, 'completions/min_length': 126.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1080729216337204, 'completions/mean_terminated_length': 563.7233276367188, 'completions/min_terminated_length': 126.0, 'completions/max_terminated_length': 2622.0, 'rewards/reward_correctness/mean': 0.625, 'rewards/reward_correctness/std': 0.4842805862426758, 'reward': 0.625, 'reward_std': 0.4842805564403534, 'frac_reward_zero_std': 0.453125, 'sampling/sampling_logp_difference/mean': 0.025114521384239197, 'sampling/sampling_logp_difference/max': 1.3054018020629883, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5968267917633057, 'sampling/importance_sampling_ratio/max': 2.9606382846832275, 'entropy': 0.40907679829979315, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 304.87532119173557, 'epoch': 1.81}
90%|█████████ | 123/136 [13:14:01<1:35:15, 439.69s/it][2026-08-13 11:44:29] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,847)
[2026-08-13 11:44:30] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\)' @(1,2577)
[2026-08-13 11:44:32] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\)' @(1,2)
91%|█████████ | 124/136 [13:19:11<1:20:09, 400.83s/it]
{'loss': -0.0468, 'grad_norm': 0.05567172477716058, 'learning_rate': 3.650769475700163e-07, 'num_tokens': 179818959.0, 'completions/mean_length': 843.2526245117188, 'completions/min_length': 147.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1028645858168602, 'completions/mean_terminated_length': 587.8563232421875, 'completions/min_terminated_length': 147.0, 'completions/max_terminated_length': 2743.0, 'rewards/reward_correctness/mean': 0.67578125, 'rewards/reward_correctness/std': 0.468234658241272, 'reward': 0.67578125, 'reward_std': 0.4682346284389496, 'frac_reward_zero_std': 0.5859375, 'sampling/sampling_logp_difference/mean': 0.02423350140452385, 'sampling/sampling_logp_difference/max': 2.201457977294922, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.6072058081626892, 'sampling/importance_sampling_ratio/max': 2.9983177185058594, 'entropy': 0.3993370668613352, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 308.9692654227838, 'epoch': 1.82}
91%|█████████ | 124/136 [13:19:11<1:20:09, 400.83s/it][2026-08-13 11:49:51] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "frac" @(1,1)
Open LaTeX blocks:
@(1,1) arguments of macro "frac"
[2026-08-13 11:49:51] WARNING __init__.py:1389: WARNING: Error in configuration: macro '\frac' failed its substitution!
[2026-08-13 11:49:51] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,13)
92%|█████████▏| 125/136 [13:24:28<1:08:53, 375.73s/it]
{'loss': -0.0201, 'grad_norm': 0.04641479563046326, 'learning_rate': 3.5551675835057994e-07, 'num_tokens': 181378256.0, 'completions/mean_length': 918.4173583984375, 'completions/min_length': 141.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1171875, 'completions/mean_terminated_length': 632.5435180664062, 'completions/min_terminated_length': 141.0, 'completions/max_terminated_length': 2944.0, 'rewards/reward_correctness/mean': 0.6080729365348816, 'rewards/reward_correctness/std': 0.48833954334259033, 'reward': 0.6080729365348816, 'reward_std': 0.48833951354026794, 'frac_reward_zero_std': 0.46875, 'sampling/sampling_logp_difference/mean': 0.024576259776949883, 'sampling/sampling_logp_difference/max': 2.0342092514038086, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5756347179412842, 'sampling/importance_sampling_ratio/max': 2.930337429046631, 'entropy': 0.41694220766657963, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 316.0315662450157, 'epoch': 1.84}
92%|█████████▏| 125/136 [13:24:28<1:08:53, 375.73s/it][2026-08-13 11:54:56] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,243)
[2026-08-13 11:54:56] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,35)
93%|█████████▎| 126/136 [13:29:32<59:01, 354.12s/it]
{'loss': -0.0423, 'grad_norm': 0.04851792726196639, 'learning_rate': 3.4670101439654904e-07, 'num_tokens': 182828080.0, 'completions/mean_length': 848.2474365234375, 'completions/min_length': 179.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.095703125, 'completions/mean_terminated_length': 612.9042358398438, 'completions/min_terminated_length': 179.0, 'completions/max_terminated_length': 2250.0, 'rewards/reward_correctness/mean': 0.5989583134651184, 'rewards/reward_correctness/std': 0.4902690351009369, 'reward': 0.5989583730697632, 'reward_std': 0.4902690052986145, 'frac_reward_zero_std': 0.5546875, 'sampling/sampling_logp_difference/mean': 0.02267294190824032, 'sampling/sampling_logp_difference/max': 1.0113334655761719, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5886579155921936, 'sampling/importance_sampling_ratio/max': 2.94360613822937, 'entropy': 0.3706806362606585, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 302.60534020280465, 'epoch': 1.85}
93%|█████████▎| 126/136 [13:29:32<59:01, 354.12s/it]
93%|█████████▎| 127/136 [13:34:40<51:01, 340.15s/it]
{'loss': -0.0353, 'grad_norm': 0.06029741056238967, 'learning_rate': 3.3863478556022955e-07, 'num_tokens': 184296533.0, 'completions/mean_length': 876.0866088867188, 'completions/min_length': 168.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1041666716337204, 'completions/mean_terminated_length': 620.747802734375, 'completions/min_terminated_length': 168.0, 'completions/max_terminated_length': 2757.0, 'rewards/reward_correctness/mean': 0.642578125, 'rewards/reward_correctness/std': 0.47939661145210266, 'reward': 0.642578125, 'reward_std': 0.47939661145210266, 'frac_reward_zero_std': 0.5234375, 'sampling/sampling_logp_difference/mean': 0.023947691544890404, 'sampling/sampling_logp_difference/max': 1.4283764362335205, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5787960886955261, 'sampling/importance_sampling_ratio/max': 2.996951103210449, 'entropy': 0.3983296254882589, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 306.45986785087734, 'epoch': 1.87}
93%|█████████▎| 127/136 [13:34:40<51:01, 340.15s/it][2026-08-13 12:05:21] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,1)
[2026-08-13 12:05:23] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\)' @(1,33)
[2026-08-13 12:05:23] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\]' @(1,126)
94%|█████████▍| 128/136 [13:40:07<44:51, 336.44s/it]
{'loss': -0.0345, 'grad_norm': 0.042861531662967446, 'learning_rate': 3.313227106547582e-07, 'num_tokens': 185860104.0, 'completions/mean_length': 918.0983276367188, 'completions/min_length': 180.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1243489608168602, 'completions/mean_terminated_length': 612.228271484375, 'completions/min_terminated_length': 180.0, 'completions/max_terminated_length': 2824.0, 'rewards/reward_correctness/mean': 0.611328125, 'rewards/reward_correctness/std': 0.48760727047920227, 'reward': 0.611328125, 'reward_std': 0.4876072406768799, 'frac_reward_zero_std': 0.453125, 'sampling/sampling_logp_difference/mean': 0.025863494724035263, 'sampling/sampling_logp_difference/max': 1.1543688774108887, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5670685172080994, 'sampling/importance_sampling_ratio/max': 2.9391822814941406, 'entropy': 0.44061603699810803, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 326.5879848697223, 'epoch': 1.88}
94%|█████████▍| 128/136 [13:40:07<44:51, 336.44s/it][2026-08-13 12:10:50] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "frac" @(1,1)
Open LaTeX blocks:
@(1,1) arguments of macro "frac"
[2026-08-13 12:10:50] WARNING __init__.py:1389: WARNING: Error in configuration: macro '\frac' failed its substitution!
[2026-08-13 12:10:50] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,0)
95%|█████████▍| 129/136 [13:45:36<38:58, 334.01s/it]
{'loss': -0.029, 'grad_norm': 0.044436777773561156, 'learning_rate': 3.247689947863649e-07, 'num_tokens': 187473223.0, 'completions/mean_length': 937.4264526367188, 'completions/min_length': 157.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1334635466337204, 'completions/mean_terminated_length': 610.2637329101562, 'completions/min_terminated_length': 157.0, 'completions/max_terminated_length': 3031.0, 'rewards/reward_correctness/mean': 0.580078125, 'rewards/reward_correctness/std': 0.4937065839767456, 'reward': 0.580078125, 'reward_std': 0.4937065541744232, 'frac_reward_zero_std': 0.5234375, 'sampling/sampling_logp_difference/mean': 0.027563076466321945, 'sampling/sampling_logp_difference/max': 1.9644856452941895, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.6023716330528259, 'sampling/importance_sampling_ratio/max': 2.9967188835144043, 'entropy': 0.4768651365302503, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 327.2371936151758, 'epoch': 1.9}
95%|█████████▍| 129/136 [13:45:36<38:58, 334.01s/it]
96%|█████████▌| 130/136 [13:50:40<32:30, 325.04s/it]
{'loss': -0.0174, 'grad_norm': 0.04517983424365822, 'learning_rate': 3.1897740693605444e-07, 'num_tokens': 188954447.0, 'completions/mean_length': 883.4869995117188, 'completions/min_length': 158.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1002604216337204, 'completions/mean_terminated_length': 639.6150512695312, 'completions/min_terminated_length': 158.0, 'completions/max_terminated_length': 2910.0, 'rewards/reward_correctness/mean': 0.6204426884651184, 'rewards/reward_correctness/std': 0.4854348301887512, 'reward': 0.6204427480697632, 'reward_std': 0.4854348301887512, 'frac_reward_zero_std': 0.5234375, 'sampling/sampling_logp_difference/mean': 0.021819358691573143, 'sampling/sampling_logp_difference/max': 1.9377039670944214, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5933597683906555, 'sampling/importance_sampling_ratio/max': 2.9764018058776855, 'entropy': 0.3736990636680275, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 302.9436299605295, 'epoch': 1.91}
96%|█████████▌| 130/136 [13:50:40<32:30, 325.04s/it]
0%| | 0/84 [00:00<?, ?it/s]
2%|▏ | 2/84 [00:08<05:54, 4.32s/it]
4%|▎ | 3/84 [00:13<06:22, 4.72s/it]
5%|▍ | 4/84 [00:24<09:16, 6.96s/it]
6%|▌ | 5/84 [00:48<16:50, 12.79s/it]
7%|▋ | 6/84 [01:12<21:25, 16.49s/it]
8%|▊ | 7/84 [01:20<17:32, 13.67s/it]
10%|▉ | 8/84 [01:24<13:36, 10.74s/it]
11%|█ | 9/84 [01:35<13:30, 10.81s/it]
12%|█▏ | 10/84 [01:40<11:08, 9.03s/it]
13%|█▎ | 11/84 [01:48<10:35, 8.71s/it]
14%|█▍ | 12/84 [01:56<10:17, 8.57s/it]
15%|█▌ | 13/84 [02:17<14:35, 12.34s/it]
17%|█▋ | 14/84 [02:41<18:31, 15.88s/it]
18%|█▊ | 15/84 [02:49<15:28, 13.46s/it]
19%|█▉ | 16/84 [02:57<13:22, 11.80s/it]
20%|██ | 17/84 [03:03<11:10, 10.01s/it]
21%|██▏ | 18/84 [03:07<09:09, 8.32s/it]
23%|██▎ | 19/84 [03:15<08:47, 8.12s/it]
24%|██▍ | 20/84 [03:21<08:01, 7.52s/it]
25%|██▌ | 21/84 [03:45<12:56, 12.33s/it]
26%|██▌ | 22/84 [03:49<10:10, 9.85s/it]
27%|██▋ | 23/84 [03:55<08:53, 8.74s/it]
29%|██▊ | 24/84 [04:19<13:18, 13.30s/it]
30%|██▉ | 25/84 [04:23<10:23, 10.57s/it]
31%|███ | 26/84 [04:47<14:01, 14.51s/it]
32%|███▏ | 27/84 [04:54<11:51, 12.48s/it]
33%|███▎ | 28/84 [05:00<09:39, 10.34s/it]
35%|███▍ | 29/84 [05:08<08:47, 9.60s/it]
36%|███▌ | 30/84 [05:16<08:24, 9.35s/it]
37%|███▋ | 31/84 [05:40<12:01, 13.62s/it]
38%|███▊ | 32/84 [05:45<09:35, 11.06s/it]
39%|███▉ | 33/84 [06:09<12:35, 14.81s/it]
40%|████ | 34/84 [06:20<11:32, 13.85s/it]
42%|████▏ | 35/84 [06:36<11:42, 14.33s/it]
43%|████▎ | 36/84 [06:42<09:32, 11.92s/it]
44%|████▍ | 37/84 [06:46<07:30, 9.58s/it]
45%|████▌ | 38/84 [06:54<06:51, 8.95s/it]
46%|████▋ | 39/84 [07:03<06:52, 9.17s/it]
48%|████▊ | 40/84 [07:19<08:13, 11.21s/it]
49%|████▉ | 41/84 [07:28<07:29, 10.45s/it]
50%|█████ | 42/84 [07:39<07:26, 10.63s/it]
51%|█████ | 43/84 [07:46<06:31, 9.56s/it]
52%|█████▏ | 44/84 [07:51<05:29, 8.23s/it]
54%|█████▎ | 45/84 [08:15<08:24, 12.95s/it]
55%|█████▍ | 46/84 [08:39<10:14, 16.17s/it]
56%|█████▌ | 47/84 [08:47<08:34, 13.90s/it]
57%|█████▋ | 48/84 [08:58<07:41, 12.81s/it]
58%|█████▊ | 49/84 [09:05<06:33, 11.25s/it]
60%|█████▉ | 50/84 [09:14<06:01, 10.64s/it]
61%|██████ | 51/84 [09:38<07:58, 14.49s/it]
62%|██████▏ | 52/84 [09:48<06:57, 13.06s/it]
63%|██████▎ | 53/84 [09:54<05:39, 10.95s/it]
64%|██████▍ | 54/84 [09:57<04:22, 8.76s/it]
65%|██████▌ | 55/84 [10:06<04:16, 8.83s/it]
67%|██████▋ | 56/84 [10:10<03:25, 7.33s/it]
68%|██████▊ | 57/84 [10:17<03:16, 7.28s/it]
69%|██████▉ | 58/84 [10:21<02:44, 6.33s/it]
70%|███████ | 59/84 [10:29<02:46, 6.64s/it]
71%|███████▏ | 60/84 [10:34<02:29, 6.22s/it]
73%|███████▎ | 61/84 [10:39<02:17, 5.99s/it]
74%|███████▍ | 62/84 [10:50<02:38, 7.22s/it]
75%|███████▌ | 63/84 [10:57<02:31, 7.21s/it]
76%|███████▌ | 64/84 [11:09<02:55, 8.75s/it]
77%|███████▋ | 65/84 [11:33<04:13, 13.32s/it]
79%|███████▊ | 66/84 [11:38<03:16, 10.89s/it]
80%|███████▉ | 67/84 [11:45<02:45, 9.71s/it]
81%|████████ | 68/84 [11:56<02:38, 9.93s/it]
82%|████████▏ | 69/84 [12:02<02:11, 8.76s/it]
83%|████████▎ | 70/84 [12:25<03:05, 13.26s/it]
85%|████████▍ | 71/84 [12:30<02:16, 10.53s/it]
86%|████████▌ | 72/84 [12:36<01:49, 9.13s/it]
87%|████████▋ | 73/84 [12:42<01:31, 8.30s/it]
88%|████████▊ | 74/84 [12:48<01:17, 7.80s/it]
89%|████████▉ | 75/84 [12:54<01:04, 7.18s/it]
90%|█████████ | 76/84 [13:04<01:04, 8.07s/it]
92%|█████████▏| 77/84 [13:09<00:49, 7.14s/it]
93%|█████████▎| 78/84 [13:15<00:40, 6.78s/it]
94%|█████████▍| 79/84 [13:20<00:30, 6.03s/it]
95%|█████████▌| 80/84 [13:28<00:27, 6.84s/it]
96%|█████████▋| 81/84 [13:35<00:19, 6.66s/it]
98%|█████████▊| 82/84 [13:42<00:13, 6.92s/it]
99%|█████████▉| 83/84 [13:48<00:06, 6.56s/it]
100%|██████████| 84/84 [13:53<00:00, 6.18s/it]
{'eval_loss': 0.0, 'eval_runtime': 839.8586, 'eval_samples_per_second': 0.595, 'eval_steps_per_second': 0.1, 'eval_num_tokens': 188954447.0, 'eval_completions/mean_length': 639.4682747977121, 'eval_completions/min_length': 265.5595238095238, 'eval_completions/max_length': 1345.5, 'eval_completions/clipped_ratio': 0.02579365156236149, 'eval_completions/mean_terminated_length': 575.7262064615885, 'eval_completions/min_terminated_length': 265.5595238095238, 'eval_completions/max_terminated_length': 1026.9285714285713, 'eval_rewards/reward_correctness/mean': 0.7123015891937983, 'eval_rewards/reward_correctness/std': 0.41657470166683197, 'eval_reward': 0.7123016062236968, 'eval_reward_std': 0.4165747023764111, 'eval_frac_reward_zero_std': 1.0, 'eval_sampling/sampling_logp_difference/mean': 0.0360281661685024, 'eval_sampling/sampling_logp_difference/max': 2.4157612039929344, 'eval_sampling/importance_sampling_ratio/min': 0.00011745825359944689, 'eval_sampling/importance_sampling_ratio/mean': 0.08415341998373888, 'eval_sampling/importance_sampling_ratio/max': 0.42427862378411063, 'eval_entropy': 0.14163078057269254, 'eval_clip_ratio/low_mean': 0.0, 'eval_clip_ratio/low_min': 0.0, 'eval_clip_ratio/high_mean': 0.0, 'eval_clip_ratio/high_max': 0.0, 'eval_clip_ratio/region_mean': 0.0, 'epoch': 1.91}
96%|█████████▌| 130/136 [14:04:40<32:30, 325.04s/it]
100%|██████████| 84/84 [13:53<00:00, 6.18s/it]
[2026-08-13 12:35:15] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,56)
[2026-08-13 12:35:17] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,3)
96%|█████████▋| 131/136 [14:09:50<47:42, 572.50s/it]
{'loss': -0.002, 'grad_norm': 0.05146181509009875, 'learning_rate': 3.1395127779210154e-07, 'num_tokens': 190413235.0, 'completions/mean_length': 856.2318115234375, 'completions/min_length': 147.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.09375, 'completions/mean_terminated_length': 627.0143432617188, 'completions/min_terminated_length': 147.0, 'completions/max_terminated_length': 3019.0, 'rewards/reward_correctness/mean': 0.6484375, 'rewards/reward_correctness/std': 0.4776136875152588, 'reward': 0.6484375, 'reward_std': 0.4776136577129364, 'frac_reward_zero_std': 0.5390625, 'sampling/sampling_logp_difference/mean': 0.021607233211398125, 'sampling/sampling_logp_difference/max': 1.4557209014892578, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.6245337128639221, 'sampling/importance_sampling_ratio/max': 2.9930741786956787, 'entropy': 0.35391040629474446, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 302.00397709198296, 'epoch': 1.93}
96%|█████████▋| 131/136 [14:09:50<47:42, 572.50s/it][2026-08-13 12:40:39] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,1)
[2026-08-13 12:40:39] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Bad \end macro: expected {environmentname} @(1,103)
[2026-08-13 12:40:39] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,1)
[2026-08-13 12:40:39] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Bad \end macro: expected {environmentname} @(1,110)
97%|█████████▋| 132/136 [14:15:43<33:46, 506.70s/it]
{'loss': -0.0302, 'grad_norm': 0.05004972084846401, 'learning_rate': 3.0969349783460157e-07, 'num_tokens': 192014218.0, 'completions/mean_length': 940.916015625, 'completions/min_length': 141.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1041666716337204, 'completions/mean_terminated_length': 695.9607543945312, 'completions/min_terminated_length': 141.0, 'completions/max_terminated_length': 2876.0, 'rewards/reward_correctness/mean': 0.6432291865348816, 'rewards/reward_correctness/std': 0.47920235991477966, 'reward': 0.6432291865348816, 'reward_std': 0.47920235991477966, 'frac_reward_zero_std': 0.59375, 'sampling/sampling_logp_difference/mean': 0.03865577280521393, 'sampling/sampling_logp_difference/max': 18.620969772338867, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5011322498321533, 'sampling/importance_sampling_ratio/max': 2.9860193729400635, 'entropy': 0.38355993037112057, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 352.00507430778816, 'epoch': 1.94}
97%|█████████▋| 132/136 [14:15:43<33:46, 506.70s/it][2026-08-13 12:46:13] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "frac" @(1,1)
Open LaTeX blocks:
@(1,1) arguments of macro "frac"
[2026-08-13 12:46:13] WARNING __init__.py:1389: WARNING: Error in configuration: macro '\frac' failed its substitution!
[2026-08-13 12:46:13] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "frac" @(1,2)
Open LaTeX blocks:
@(1,2) arguments of macro "frac"
[2026-08-13 12:46:13] WARNING __init__.py:1389: WARNING: Error in configuration: macro '\frac' failed its substitution!
[2026-08-13 12:46:14] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,0)
[2026-08-13 12:46:14] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,6)
[2026-08-13 12:46:15] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected closing math mode: '\)' @(1,4195)
98%|█████████▊| 133/136 [14:21:00<22:29, 449.77s/it]
{'loss': -0.0244, 'grad_norm': 0.07158180367926063, 'learning_rate': 3.0620651567318436e-07, 'num_tokens': 193466782.0, 'completions/mean_length': 842.8984375, 'completions/min_length': 133.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.08984375, 'completions/mean_terminated_length': 622.9170532226562, 'completions/min_terminated_length': 133.0, 'completions/max_terminated_length': 2799.0, 'rewards/reward_correctness/mean': 0.6575520634651184, 'rewards/reward_correctness/std': 0.47468310594558716, 'reward': 0.6575521230697632, 'reward_std': 0.47468307614326477, 'frac_reward_zero_std': 0.515625, 'sampling/sampling_logp_difference/mean': 0.022393621504306793, 'sampling/sampling_logp_difference/max': 1.8104618787765503, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5758584141731262, 'sampling/importance_sampling_ratio/max': 2.993088722229004, 'entropy': 0.36497818934731185, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 315.8437795029022, 'epoch': 1.96}
98%|█████████▊| 133/136 [14:21:00<22:29, 449.77s/it][2026-08-13 12:51:37] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): End of input while parsing arguments of macro "frac" @(1,8)
Open LaTeX blocks:
@(1,8) arguments of macro "frac"
[2026-08-13 12:51:37] WARNING __init__.py:1389: WARNING: Error in configuration: macro '\frac' failed its substitution!
99%|█████████▊| 134/136 [14:26:24<13:43, 411.98s/it]
{'loss': -0.0342, 'grad_norm': 0.06221349869734758, 'learning_rate': 3.0349233663883985e-07, 'num_tokens': 195044597.0, 'completions/mean_length': 923.3717651367188, 'completions/min_length': 126.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.119140625, 'completions/mean_terminated_length': 633.404296875, 'completions/min_terminated_length': 126.0, 'completions/max_terminated_length': 2043.0, 'rewards/reward_correctness/mean': 0.6217448115348816, 'rewards/reward_correctness/std': 0.48510968685150146, 'reward': 0.6217448115348816, 'reward_std': 0.48510968685150146, 'frac_reward_zero_std': 0.5234375, 'sampling/sampling_logp_difference/mean': 0.02538115344941616, 'sampling/sampling_logp_difference/max': 3.544827461242676, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5720989108085632, 'sampling/importance_sampling_ratio/max': 2.9823083877563477, 'entropy': 0.42708491726079956, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 322.6157622030005, 'epoch': 1.97}
99%|█████████▊| 134/136 [14:26:24<13:43, 411.98s/it][2026-08-13 12:56:58] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,0)
[2026-08-13 12:56:58] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,21)
[2026-08-13 12:56:58] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,22)
[2026-08-13 12:56:58] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,0)
99%|█████████▉| 135/136 [14:31:40<06:23, 383.12s/it]
{'loss': 0.0012, 'grad_norm': 0.05529387617320365, 'learning_rate': 3.015525216306716e-07, 'num_tokens': 196534291.0, 'completions/mean_length': 868.0169677734375, 'completions/min_length': 163.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1015625, 'completions/mean_terminated_length': 618.9080200195312, 'completions/min_terminated_length': 163.0, 'completions/max_terminated_length': 2785.0, 'rewards/reward_correctness/mean': 0.6087239384651184, 'rewards/reward_correctness/std': 0.48819488286972046, 'reward': 0.6087239980697632, 'reward_std': 0.48819488286972046, 'frac_reward_zero_std': 0.53125, 'sampling/sampling_logp_difference/mean': 0.024546820670366287, 'sampling/sampling_logp_difference/max': 1.6037323474884033, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5620114803314209, 'sampling/importance_sampling_ratio/max': 2.9879226684570312, 'entropy': 0.4124655578052625, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 314.64262982131913, 'epoch': 1.99}
99%|█████████▉| 135/136 [14:31:40<06:23, 383.12s/it][2026-08-13 13:02:09] INFO __init__.py:1249: Ignoring parse error (tolerant parsing mode): Unexpected mismatching closing brace: '}' @(1,6247)
100%|██████████| 136/136 [14:36:42<00:00, 358.81s/it]
{'loss': -0.0307, 'grad_norm': 0.04742592778780676, 'learning_rate': 3.003881862182383e-07, 'num_tokens': 198033861.0, 'completions/mean_length': 891.4622802734375, 'completions/min_length': 175.0, 'completions/max_length': 3072.0, 'completions/clipped_ratio': 0.1067708358168602, 'completions/mean_terminated_length': 630.8148803710938, 'completions/min_terminated_length': 175.0, 'completions/max_terminated_length': 3046.0, 'rewards/reward_correctness/mean': 0.6223958134651184, 'rewards/reward_correctness/std': 0.48494574427604675, 'reward': 0.6223958730697632, 'reward_std': 0.484945684671402, 'frac_reward_zero_std': 0.53125, 'sampling/sampling_logp_difference/mean': 0.02337666228413582, 'sampling/sampling_logp_difference/max': 4.9380927085876465, 'sampling/importance_sampling_ratio/min': 0.0, 'sampling/importance_sampling_ratio/mean': 0.5779362320899963, 'sampling/importance_sampling_ratio/max': 2.9746880531311035, 'entropy': 0.4176048967638053, 'clip_ratio/low_mean': 0.0, 'clip_ratio/low_min': 0.0, 'clip_ratio/high_mean': 0.0, 'clip_ratio/high_max': 0.0, 'clip_ratio/region_mean': 0.0, 'step_time': 300.921415225137, 'epoch': 2.0}
100%|██████████| 136/136 [14:36:42<00:00, 358.81s/it]
{'train_runtime': 52608.9994, 'train_samples_per_second': 0.332, 'train_steps_per_second': 0.003, 'train_loss': -0.02310945923197912, 'epoch': 2.0}
100%|██████████| 136/136 [14:36:48<00:00, 358.81s/it]
100%|██████████| 136/136 [14:36:48<00:00, 386.83s/it]
[rank4]:[W813 13:05:18.671711696 ProcessGroupNCCL.cpp:1553] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
[rank2]:[W813 13:05:18.702290790 ProcessGroupNCCL.cpp:1553] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
[rank3]:[W813 13:05:18.712095197 ProcessGroupNCCL.cpp:1553] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
[rank1]:[W813 13:05:18.742272413 ProcessGroupNCCL.cpp:1553] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
[rank5]:[W813 13:05:18.822800037 ProcessGroupNCCL.cpp:1553] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())
wandb:
wandb: 🚀 View run gt__microsoft_Phi-4-mini-instruct__20260812_222659_lr3e-6_bs1536 at: https://wandb.ai/logan-yang2002-johns-hopkins-university/grpo-training/runs/5cwwrnh0
wandb: Find logs at: wandb/run-20260812_222730-5cwwrnh0/logs
[rank0]:[W813 13:05:21.418453192 ProcessGroupNCCL.cpp:1553] Warning: WARNING: destroy_process_group() was not called before program exit, which can leak resources. For more info, please see https://pytorch.org/docs/stable/distributed.html#shutdown (function operator())