Model save
Browse files- README.md +68 -0
- all_results.json +8 -0
- generation_config.json +9 -0
- train_results.json +8 -0
- trainer_state.json +1229 -0
README.md
ADDED
|
@@ -0,0 +1,68 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
base_model: deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B
|
| 3 |
+
library_name: transformers
|
| 4 |
+
model_name: Qwen2.5-1.5B-Open-R1-GRPO-cot-v3
|
| 5 |
+
tags:
|
| 6 |
+
- generated_from_trainer
|
| 7 |
+
- trl
|
| 8 |
+
- grpo
|
| 9 |
+
licence: license
|
| 10 |
+
---
|
| 11 |
+
|
| 12 |
+
# Model Card for Qwen2.5-1.5B-Open-R1-GRPO-cot-v3
|
| 13 |
+
|
| 14 |
+
This model is a fine-tuned version of [deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B](https://huggingface.co/deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B).
|
| 15 |
+
It has been trained using [TRL](https://github.com/huggingface/trl).
|
| 16 |
+
|
| 17 |
+
## Quick start
|
| 18 |
+
|
| 19 |
+
```python
|
| 20 |
+
from transformers import pipeline
|
| 21 |
+
|
| 22 |
+
question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
|
| 23 |
+
generator = pipeline("text-generation", model="weltonwang88/Qwen2.5-1.5B-Open-R1-GRPO-cot-v3", device="cuda")
|
| 24 |
+
output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
|
| 25 |
+
print(output["generated_text"])
|
| 26 |
+
```
|
| 27 |
+
|
| 28 |
+
## Training procedure
|
| 29 |
+
|
| 30 |
+
[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/weltonwang88-stanford/huggingface/runs/ef81zz98)
|
| 31 |
+
|
| 32 |
+
|
| 33 |
+
This model was trained with GRPO, a method introduced in [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://huggingface.co/papers/2402.03300).
|
| 34 |
+
|
| 35 |
+
### Framework versions
|
| 36 |
+
|
| 37 |
+
- TRL: 0.16.0.dev0
|
| 38 |
+
- Transformers: 4.50.0.dev0
|
| 39 |
+
- Pytorch: 2.5.1+cu121
|
| 40 |
+
- Datasets: 3.3.2
|
| 41 |
+
- Tokenizers: 0.21.1
|
| 42 |
+
|
| 43 |
+
## Citations
|
| 44 |
+
|
| 45 |
+
Cite GRPO as:
|
| 46 |
+
|
| 47 |
+
```bibtex
|
| 48 |
+
@article{zhihong2024deepseekmath,
|
| 49 |
+
title = {{DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models}},
|
| 50 |
+
author = {Zhihong Shao and Peiyi Wang and Qihao Zhu and Runxin Xu and Junxiao Song and Mingchuan Zhang and Y. K. Li and Y. Wu and Daya Guo},
|
| 51 |
+
year = 2024,
|
| 52 |
+
eprint = {arXiv:2402.03300},
|
| 53 |
+
}
|
| 54 |
+
|
| 55 |
+
```
|
| 56 |
+
|
| 57 |
+
Cite TRL as:
|
| 58 |
+
|
| 59 |
+
```bibtex
|
| 60 |
+
@misc{vonwerra2022trl,
|
| 61 |
+
title = {{TRL: Transformer Reinforcement Learning}},
|
| 62 |
+
author = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang and Kashif Rasul and Quentin Gallouédec},
|
| 63 |
+
year = 2020,
|
| 64 |
+
journal = {GitHub repository},
|
| 65 |
+
publisher = {GitHub},
|
| 66 |
+
howpublished = {\url{https://github.com/huggingface/trl}}
|
| 67 |
+
}
|
| 68 |
+
```
|
all_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"total_flos": 0.0,
|
| 3 |
+
"train_loss": 0.6775813137989773,
|
| 4 |
+
"train_runtime": 20674.7257,
|
| 5 |
+
"train_samples": 50,
|
| 6 |
+
"train_samples_per_second": 0.087,
|
| 7 |
+
"train_steps_per_second": 0.005
|
| 8 |
+
}
|
generation_config.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"bos_token_id": 151646,
|
| 4 |
+
"do_sample": true,
|
| 5 |
+
"eos_token_id": 151643,
|
| 6 |
+
"temperature": 0.6,
|
| 7 |
+
"top_p": 0.95,
|
| 8 |
+
"transformers_version": "4.50.0.dev0"
|
| 9 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"total_flos": 0.0,
|
| 3 |
+
"train_loss": 0.6775813137989773,
|
| 4 |
+
"train_runtime": 20674.7257,
|
| 5 |
+
"train_samples": 50,
|
| 6 |
+
"train_samples_per_second": 0.087,
|
| 7 |
+
"train_steps_per_second": 0.005
|
| 8 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,1229 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 3.9955555555555557,
|
| 5 |
+
"eval_steps": 30,
|
| 6 |
+
"global_step": 112,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"clip_ratio": 0.0,
|
| 13 |
+
"completion_length": 1992.2813301086426,
|
| 14 |
+
"epoch": 0.035555555555555556,
|
| 15 |
+
"grad_norm": 0.11222778239149554,
|
| 16 |
+
"kl": 0.0,
|
| 17 |
+
"learning_rate": 1.6666666666666667e-06,
|
| 18 |
+
"loss": 0.0509,
|
| 19 |
+
"reward": -8.534029252827168,
|
| 20 |
+
"reward_std": 3.1286671087145805,
|
| 21 |
+
"rewards/cot_length_penalty_reward": -8.792958237230778,
|
| 22 |
+
"rewards/math_latex_accuracy_reward": 0.2589285857975483,
|
| 23 |
+
"step": 1
|
| 24 |
+
},
|
| 25 |
+
{
|
| 26 |
+
"clip_ratio": 0.0,
|
| 27 |
+
"epoch": 0.07111111111111111,
|
| 28 |
+
"grad_norm": 0.11224900964736327,
|
| 29 |
+
"kl": 0.0,
|
| 30 |
+
"learning_rate": 3.3333333333333333e-06,
|
| 31 |
+
"loss": 0.0509,
|
| 32 |
+
"step": 2
|
| 33 |
+
},
|
| 34 |
+
{
|
| 35 |
+
"clip_ratio": 0.002639908329001628,
|
| 36 |
+
"epoch": 0.10666666666666667,
|
| 37 |
+
"grad_norm": 0.11190265883625335,
|
| 38 |
+
"kl": 0.0004132986068725586,
|
| 39 |
+
"learning_rate": 5e-06,
|
| 40 |
+
"loss": 0.051,
|
| 41 |
+
"step": 3
|
| 42 |
+
},
|
| 43 |
+
{
|
| 44 |
+
"clip_ratio": 0.0026859724457608536,
|
| 45 |
+
"epoch": 0.14222222222222222,
|
| 46 |
+
"grad_norm": 0.10842287053917446,
|
| 47 |
+
"kl": 0.00042808055877685547,
|
| 48 |
+
"learning_rate": 6.666666666666667e-06,
|
| 49 |
+
"loss": 0.0506,
|
| 50 |
+
"step": 4
|
| 51 |
+
},
|
| 52 |
+
{
|
| 53 |
+
"clip_ratio": 0.0,
|
| 54 |
+
"completion_length": 2354.6050567626953,
|
| 55 |
+
"epoch": 0.17777777777777778,
|
| 56 |
+
"grad_norm": 0.11660083282365401,
|
| 57 |
+
"kl": 0.0005452632904052734,
|
| 58 |
+
"learning_rate": 8.333333333333334e-06,
|
| 59 |
+
"loss": 0.0532,
|
| 60 |
+
"reward": -9.576663568615913,
|
| 61 |
+
"reward_std": 3.4961936213076115,
|
| 62 |
+
"rewards/cot_length_penalty_reward": -9.817734986543655,
|
| 63 |
+
"rewards/math_latex_accuracy_reward": 0.24107144074514508,
|
| 64 |
+
"step": 5
|
| 65 |
+
},
|
| 66 |
+
{
|
| 67 |
+
"clip_ratio": 0.004271271725883707,
|
| 68 |
+
"epoch": 0.21333333333333335,
|
| 69 |
+
"grad_norm": 0.15124528039054966,
|
| 70 |
+
"kl": 0.0023946762084960938,
|
| 71 |
+
"learning_rate": 1e-05,
|
| 72 |
+
"loss": 0.0521,
|
| 73 |
+
"step": 6
|
| 74 |
+
},
|
| 75 |
+
{
|
| 76 |
+
"clip_ratio": 0.00593576196115464,
|
| 77 |
+
"epoch": 0.24888888888888888,
|
| 78 |
+
"grad_norm": 0.22845939154064746,
|
| 79 |
+
"kl": 0.0017180442810058594,
|
| 80 |
+
"learning_rate": 1.1666666666666668e-05,
|
| 81 |
+
"loss": 0.0519,
|
| 82 |
+
"step": 7
|
| 83 |
+
},
|
| 84 |
+
{
|
| 85 |
+
"clip_ratio": 0.00681446076487191,
|
| 86 |
+
"epoch": 0.28444444444444444,
|
| 87 |
+
"grad_norm": 0.25815482225017694,
|
| 88 |
+
"kl": 0.002631664276123047,
|
| 89 |
+
"learning_rate": 1.3333333333333333e-05,
|
| 90 |
+
"loss": 0.0484,
|
| 91 |
+
"step": 8
|
| 92 |
+
},
|
| 93 |
+
{
|
| 94 |
+
"clip_ratio": 0.0,
|
| 95 |
+
"completion_length": 2251.4844703674316,
|
| 96 |
+
"epoch": 0.32,
|
| 97 |
+
"grad_norm": 0.11515864264586664,
|
| 98 |
+
"kl": 0.0024547576904296875,
|
| 99 |
+
"learning_rate": 1.5000000000000002e-05,
|
| 100 |
+
"loss": 0.0261,
|
| 101 |
+
"reward": -8.878705874085426,
|
| 102 |
+
"reward_std": 3.5140193179249763,
|
| 103 |
+
"rewards/cot_length_penalty_reward": -9.128705888986588,
|
| 104 |
+
"rewards/math_latex_accuracy_reward": 0.2500000149011612,
|
| 105 |
+
"step": 9
|
| 106 |
+
},
|
| 107 |
+
{
|
| 108 |
+
"clip_ratio": 0.0070763813419034705,
|
| 109 |
+
"epoch": 0.35555555555555557,
|
| 110 |
+
"grad_norm": 0.2974695944541913,
|
| 111 |
+
"kl": 0.005417823791503906,
|
| 112 |
+
"learning_rate": 1.6666666666666667e-05,
|
| 113 |
+
"loss": 0.0255,
|
| 114 |
+
"step": 10
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"clip_ratio": 0.010105093329912052,
|
| 118 |
+
"epoch": 0.39111111111111113,
|
| 119 |
+
"grad_norm": 85.04432926044146,
|
| 120 |
+
"kl": 0.007180213928222656,
|
| 121 |
+
"learning_rate": 1.8333333333333333e-05,
|
| 122 |
+
"loss": 19.5224,
|
| 123 |
+
"step": 11
|
| 124 |
+
},
|
| 125 |
+
{
|
| 126 |
+
"clip_ratio": 0.017627036664634943,
|
| 127 |
+
"epoch": 0.4266666666666667,
|
| 128 |
+
"grad_norm": 2.432235493998052,
|
| 129 |
+
"kl": 0.0702056884765625,
|
| 130 |
+
"learning_rate": 2e-05,
|
| 131 |
+
"loss": 0.0247,
|
| 132 |
+
"step": 12
|
| 133 |
+
},
|
| 134 |
+
{
|
| 135 |
+
"clip_ratio": 0.0,
|
| 136 |
+
"completion_length": 1870.0737648010254,
|
| 137 |
+
"epoch": 0.4622222222222222,
|
| 138 |
+
"grad_norm": 0.32693917550649865,
|
| 139 |
+
"kl": 0.022480010986328125,
|
| 140 |
+
"learning_rate": 1.9995065603657317e-05,
|
| 141 |
+
"loss": 0.0103,
|
| 142 |
+
"reward": -9.444286078214645,
|
| 143 |
+
"reward_std": 3.207320176064968,
|
| 144 |
+
"rewards/cot_length_penalty_reward": -9.848303943872452,
|
| 145 |
+
"rewards/math_latex_accuracy_reward": 0.4040178805589676,
|
| 146 |
+
"step": 13
|
| 147 |
+
},
|
| 148 |
+
{
|
| 149 |
+
"clip_ratio": 0.004404508654261008,
|
| 150 |
+
"epoch": 0.49777777777777776,
|
| 151 |
+
"grad_norm": 1.6311442510668763,
|
| 152 |
+
"kl": 0.010528564453125,
|
| 153 |
+
"learning_rate": 1.9980267284282718e-05,
|
| 154 |
+
"loss": 0.0093,
|
| 155 |
+
"step": 14
|
| 156 |
+
},
|
| 157 |
+
{
|
| 158 |
+
"clip_ratio": 0.0062026621017139405,
|
| 159 |
+
"epoch": 0.5333333333333333,
|
| 160 |
+
"grad_norm": 0.5615009008596469,
|
| 161 |
+
"kl": 0.051082611083984375,
|
| 162 |
+
"learning_rate": 1.99556196460308e-05,
|
| 163 |
+
"loss": 0.0076,
|
| 164 |
+
"step": 15
|
| 165 |
+
},
|
| 166 |
+
{
|
| 167 |
+
"clip_ratio": 0.006701507809339091,
|
| 168 |
+
"epoch": 0.5688888888888889,
|
| 169 |
+
"grad_norm": 0.15171374489080983,
|
| 170 |
+
"kl": 0.018802642822265625,
|
| 171 |
+
"learning_rate": 1.9921147013144782e-05,
|
| 172 |
+
"loss": 0.0041,
|
| 173 |
+
"step": 16
|
| 174 |
+
},
|
| 175 |
+
{
|
| 176 |
+
"clip_ratio": 0.0,
|
| 177 |
+
"completion_length": 2222.4309005737305,
|
| 178 |
+
"epoch": 0.6044444444444445,
|
| 179 |
+
"grad_norm": 0.11283429637288535,
|
| 180 |
+
"kl": 0.01442718505859375,
|
| 181 |
+
"learning_rate": 1.9876883405951378e-05,
|
| 182 |
+
"loss": 0.081,
|
| 183 |
+
"reward": -9.554554164409637,
|
| 184 |
+
"reward_std": 4.235630825161934,
|
| 185 |
+
"rewards/cot_length_penalty_reward": -9.844732716679573,
|
| 186 |
+
"rewards/math_latex_accuracy_reward": 0.29017858393490314,
|
| 187 |
+
"step": 17
|
| 188 |
+
},
|
| 189 |
+
{
|
| 190 |
+
"clip_ratio": 0.004120954225072637,
|
| 191 |
+
"epoch": 0.64,
|
| 192 |
+
"grad_norm": 0.10899171004441378,
|
| 193 |
+
"kl": 0.015628814697265625,
|
| 194 |
+
"learning_rate": 1.982287250728689e-05,
|
| 195 |
+
"loss": 0.2482,
|
| 196 |
+
"step": 18
|
| 197 |
+
},
|
| 198 |
+
{
|
| 199 |
+
"clip_ratio": 0.005419444481958635,
|
| 200 |
+
"epoch": 0.6755555555555556,
|
| 201 |
+
"grad_norm": 0.1157331857796372,
|
| 202 |
+
"kl": 0.01764678955078125,
|
| 203 |
+
"learning_rate": 1.9759167619387474e-05,
|
| 204 |
+
"loss": 0.2459,
|
| 205 |
+
"step": 19
|
| 206 |
+
},
|
| 207 |
+
{
|
| 208 |
+
"clip_ratio": 0.005958295805612579,
|
| 209 |
+
"epoch": 0.7111111111111111,
|
| 210 |
+
"grad_norm": 0.10652991525129403,
|
| 211 |
+
"kl": 0.01905059814453125,
|
| 212 |
+
"learning_rate": 1.9685831611286312e-05,
|
| 213 |
+
"loss": 0.2434,
|
| 214 |
+
"step": 20
|
| 215 |
+
},
|
| 216 |
+
{
|
| 217 |
+
"clip_ratio": 0.0,
|
| 218 |
+
"completion_length": 2501.2255668640137,
|
| 219 |
+
"epoch": 0.7466666666666667,
|
| 220 |
+
"grad_norm": 0.12065925342674215,
|
| 221 |
+
"kl": 0.020366668701171875,
|
| 222 |
+
"learning_rate": 1.9602936856769432e-05,
|
| 223 |
+
"loss": 0.033,
|
| 224 |
+
"reward": -11.581663489341736,
|
| 225 |
+
"reward_std": 4.305310405790806,
|
| 226 |
+
"rewards/cot_length_penalty_reward": -11.86737784743309,
|
| 227 |
+
"rewards/math_latex_accuracy_reward": 0.2857142973225564,
|
| 228 |
+
"step": 21
|
| 229 |
+
},
|
| 230 |
+
{
|
| 231 |
+
"clip_ratio": 0.0039043642027536407,
|
| 232 |
+
"epoch": 0.7822222222222223,
|
| 233 |
+
"grad_norm": 0.36463686705700576,
|
| 234 |
+
"kl": 0.019252777099609375,
|
| 235 |
+
"learning_rate": 1.9510565162951538e-05,
|
| 236 |
+
"loss": 0.0325,
|
| 237 |
+
"step": 22
|
| 238 |
+
},
|
| 239 |
+
{
|
| 240 |
+
"clip_ratio": 0.005750590149546042,
|
| 241 |
+
"epoch": 0.8177777777777778,
|
| 242 |
+
"grad_norm": 20283.4909421177,
|
| 243 |
+
"kl": 1147.058982849121,
|
| 244 |
+
"learning_rate": 1.9408807689542257e-05,
|
| 245 |
+
"loss": 46.027,
|
| 246 |
+
"step": 23
|
| 247 |
+
},
|
| 248 |
+
{
|
| 249 |
+
"clip_ratio": 0.008487990504363552,
|
| 250 |
+
"epoch": 0.8533333333333334,
|
| 251 |
+
"grad_norm": 0.17175661916148474,
|
| 252 |
+
"kl": 0.026885986328125,
|
| 253 |
+
"learning_rate": 1.9297764858882516e-05,
|
| 254 |
+
"loss": 0.0289,
|
| 255 |
+
"step": 24
|
| 256 |
+
},
|
| 257 |
+
{
|
| 258 |
+
"clip_ratio": 0.0,
|
| 259 |
+
"completion_length": 2333.062614440918,
|
| 260 |
+
"epoch": 0.8888888888888888,
|
| 261 |
+
"grad_norm": 0.10364225115067384,
|
| 262 |
+
"kl": 0.0201416015625,
|
| 263 |
+
"learning_rate": 1.9177546256839814e-05,
|
| 264 |
+
"loss": 0.0113,
|
| 265 |
+
"reward": -10.619498401880264,
|
| 266 |
+
"reward_std": 3.768970273435116,
|
| 267 |
+
"rewards/cot_length_penalty_reward": -10.869498312473297,
|
| 268 |
+
"rewards/math_latex_accuracy_reward": 0.2500000123400241,
|
| 269 |
+
"step": 25
|
| 270 |
+
},
|
| 271 |
+
{
|
| 272 |
+
"clip_ratio": 0.0032545153953833506,
|
| 273 |
+
"epoch": 0.9244444444444444,
|
| 274 |
+
"grad_norm": 0.10471903488067007,
|
| 275 |
+
"kl": 0.02140045166015625,
|
| 276 |
+
"learning_rate": 1.9048270524660197e-05,
|
| 277 |
+
"loss": 0.0103,
|
| 278 |
+
"step": 26
|
| 279 |
+
},
|
| 280 |
+
{
|
| 281 |
+
"clip_ratio": 0.004354664255515672,
|
| 282 |
+
"epoch": 0.96,
|
| 283 |
+
"grad_norm": 0.0965546219660842,
|
| 284 |
+
"kl": 0.0223541259765625,
|
| 285 |
+
"learning_rate": 1.891006524188368e-05,
|
| 286 |
+
"loss": 0.0085,
|
| 287 |
+
"step": 27
|
| 288 |
+
},
|
| 289 |
+
{
|
| 290 |
+
"clip_ratio": 0.005275880845147185,
|
| 291 |
+
"epoch": 0.9955555555555555,
|
| 292 |
+
"grad_norm": 0.10544368470129743,
|
| 293 |
+
"kl": 0.023712158203125,
|
| 294 |
+
"learning_rate": 1.8763066800438638e-05,
|
| 295 |
+
"loss": 0.0065,
|
| 296 |
+
"step": 28
|
| 297 |
+
},
|
| 298 |
+
{
|
| 299 |
+
"clip_ratio": 0.0,
|
| 300 |
+
"completion_length": 2427.296974182129,
|
| 301 |
+
"epoch": 1.0355555555555556,
|
| 302 |
+
"grad_norm": 0.42983301720943096,
|
| 303 |
+
"kl": 0.03394317626953125,
|
| 304 |
+
"learning_rate": 1.860742027003944e-05,
|
| 305 |
+
"loss": 0.0039,
|
| 306 |
+
"reward": -11.214900106191635,
|
| 307 |
+
"reward_std": 3.760936316102743,
|
| 308 |
+
"rewards/cot_length_penalty_reward": -11.5006143450737,
|
| 309 |
+
"rewards/math_latex_accuracy_reward": 0.285714297555387,
|
| 310 |
+
"step": 29
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"epoch": 1.0711111111111111,
|
| 314 |
+
"grad_norm": 0.1067773530257121,
|
| 315 |
+
"learning_rate": 1.8443279255020153e-05,
|
| 316 |
+
"loss": 0.0071,
|
| 317 |
+
"step": 30
|
| 318 |
+
},
|
| 319 |
+
{
|
| 320 |
+
"epoch": 1.0711111111111111,
|
| 321 |
+
"eval_clip_ratio": 0.0,
|
| 322 |
+
"eval_completion_length": 2303.7637939453125,
|
| 323 |
+
"eval_kl": 0.025606595552884616,
|
| 324 |
+
"eval_loss": 0.04365207254886627,
|
| 325 |
+
"eval_reward": -8.784464891140278,
|
| 326 |
+
"eval_reward_std": 3.7600448498359094,
|
| 327 |
+
"eval_rewards/cot_length_penalty_reward": -9.116882379238422,
|
| 328 |
+
"eval_rewards/math_latex_accuracy_reward": 0.3324175958450024,
|
| 329 |
+
"eval_runtime": 448.0952,
|
| 330 |
+
"eval_samples_per_second": 0.112,
|
| 331 |
+
"eval_steps_per_second": 0.004,
|
| 332 |
+
"step": 30
|
| 333 |
+
},
|
| 334 |
+
{
|
| 335 |
+
"clip_ratio": 0.0037656883359886706,
|
| 336 |
+
"epoch": 1.1066666666666667,
|
| 337 |
+
"grad_norm": 0.7346983824268155,
|
| 338 |
+
"kl": 0.027835845947265625,
|
| 339 |
+
"learning_rate": 1.827080574274562e-05,
|
| 340 |
+
"loss": 0.0033,
|
| 341 |
+
"step": 31
|
| 342 |
+
},
|
| 343 |
+
{
|
| 344 |
+
"clip_ratio": 0.006145871157059446,
|
| 345 |
+
"epoch": 1.1422222222222222,
|
| 346 |
+
"grad_norm": 11.473259751864658,
|
| 347 |
+
"kl": 1.4422760009765625,
|
| 348 |
+
"learning_rate": 1.8090169943749477e-05,
|
| 349 |
+
"loss": 0.0549,
|
| 350 |
+
"step": 32
|
| 351 |
+
},
|
| 352 |
+
{
|
| 353 |
+
"clip_ratio": 0.0,
|
| 354 |
+
"completion_length": 2531.1674995422363,
|
| 355 |
+
"epoch": 1.1777777777777778,
|
| 356 |
+
"grad_norm": 0.11984100416609303,
|
| 357 |
+
"kl": 0.03124237060546875,
|
| 358 |
+
"learning_rate": 1.7901550123756906e-05,
|
| 359 |
+
"loss": 0.0239,
|
| 360 |
+
"reward": -8.361417755484581,
|
| 361 |
+
"reward_std": 4.016169548034668,
|
| 362 |
+
"rewards/cot_length_penalty_reward": -8.689542889595032,
|
| 363 |
+
"rewards/math_latex_accuracy_reward": 0.3281250139698386,
|
| 364 |
+
"step": 33
|
| 365 |
+
},
|
| 366 |
+
{
|
| 367 |
+
"clip_ratio": 0.004058451057062484,
|
| 368 |
+
"epoch": 1.2133333333333334,
|
| 369 |
+
"grad_norm": 0.16684935043495766,
|
| 370 |
+
"kl": 0.03450775146484375,
|
| 371 |
+
"learning_rate": 1.7705132427757895e-05,
|
| 372 |
+
"loss": 0.0232,
|
| 373 |
+
"step": 34
|
| 374 |
+
},
|
| 375 |
+
{
|
| 376 |
+
"clip_ratio": 0.006084064312744886,
|
| 377 |
+
"epoch": 1.248888888888889,
|
| 378 |
+
"grad_norm": 0.11163561617870978,
|
| 379 |
+
"kl": 0.0318145751953125,
|
| 380 |
+
"learning_rate": 1.7501110696304598e-05,
|
| 381 |
+
"loss": 0.0214,
|
| 382 |
+
"step": 35
|
| 383 |
+
},
|
| 384 |
+
{
|
| 385 |
+
"clip_ratio": 0.007263028150191531,
|
| 386 |
+
"epoch": 1.2844444444444445,
|
| 387 |
+
"grad_norm": 0.11128954549532989,
|
| 388 |
+
"kl": 0.03281402587890625,
|
| 389 |
+
"learning_rate": 1.7289686274214116e-05,
|
| 390 |
+
"loss": 0.0197,
|
| 391 |
+
"step": 36
|
| 392 |
+
},
|
| 393 |
+
{
|
| 394 |
+
"clip_ratio": 0.0,
|
| 395 |
+
"completion_length": 1888.4487342834473,
|
| 396 |
+
"epoch": 1.32,
|
| 397 |
+
"grad_norm": 0.6837402463902799,
|
| 398 |
+
"kl": 0.05413055419921875,
|
| 399 |
+
"learning_rate": 1.7071067811865477e-05,
|
| 400 |
+
"loss": 0.1151,
|
| 401 |
+
"reward": -7.783694684505463,
|
| 402 |
+
"reward_std": 3.098730646073818,
|
| 403 |
+
"rewards/cot_length_penalty_reward": -8.16985534131527,
|
| 404 |
+
"rewards/math_latex_accuracy_reward": 0.3861607341095805,
|
| 405 |
+
"step": 37
|
| 406 |
+
},
|
| 407 |
+
{
|
| 408 |
+
"clip_ratio": 0.0028788788622478023,
|
| 409 |
+
"epoch": 1.3555555555555556,
|
| 410 |
+
"grad_norm": 2.464864347264584,
|
| 411 |
+
"kl": 0.04084014892578125,
|
| 412 |
+
"learning_rate": 1.684547105928689e-05,
|
| 413 |
+
"loss": 0.3644,
|
| 414 |
+
"step": 38
|
| 415 |
+
},
|
| 416 |
+
{
|
| 417 |
+
"clip_ratio": 0.004996606716304086,
|
| 418 |
+
"epoch": 1.3911111111111112,
|
| 419 |
+
"grad_norm": 0.32236476835294475,
|
| 420 |
+
"kl": 0.04229736328125,
|
| 421 |
+
"learning_rate": 1.661311865323652e-05,
|
| 422 |
+
"loss": 0.1132,
|
| 423 |
+
"step": 39
|
| 424 |
+
},
|
| 425 |
+
{
|
| 426 |
+
"clip_ratio": 0.005848184140631929,
|
| 427 |
+
"epoch": 1.4266666666666667,
|
| 428 |
+
"grad_norm": 2.236113570550632,
|
| 429 |
+
"kl": 0.180694580078125,
|
| 430 |
+
"learning_rate": 1.63742398974869e-05,
|
| 431 |
+
"loss": 0.116,
|
| 432 |
+
"step": 40
|
| 433 |
+
},
|
| 434 |
+
{
|
| 435 |
+
"clip_ratio": 0.0,
|
| 436 |
+
"completion_length": 1681.7277793884277,
|
| 437 |
+
"epoch": 1.462222222222222,
|
| 438 |
+
"grad_norm": 55.91487468875501,
|
| 439 |
+
"kl": 1.1835174560546875,
|
| 440 |
+
"learning_rate": 1.6129070536529767e-05,
|
| 441 |
+
"loss": 0.0918,
|
| 442 |
+
"reward": -7.822701282799244,
|
| 443 |
+
"reward_std": 2.5297958850860596,
|
| 444 |
+
"rewards/cot_length_penalty_reward": -8.191005058586597,
|
| 445 |
+
"rewards/math_latex_accuracy_reward": 0.3683035862632096,
|
| 446 |
+
"step": 41
|
| 447 |
+
},
|
| 448 |
+
{
|
| 449 |
+
"clip_ratio": 0.003095990905421786,
|
| 450 |
+
"epoch": 1.4977777777777779,
|
| 451 |
+
"grad_norm": 3454.9772869499748,
|
| 452 |
+
"kl": 0.0470123291015625,
|
| 453 |
+
"learning_rate": 1.5877852522924733e-05,
|
| 454 |
+
"loss": 6.8427,
|
| 455 |
+
"step": 42
|
| 456 |
+
},
|
| 457 |
+
{
|
| 458 |
+
"clip_ratio": 0.005094703097711317,
|
| 459 |
+
"epoch": 1.5333333333333332,
|
| 460 |
+
"grad_norm": 15.584761637863307,
|
| 461 |
+
"kl": 1.0414886474609375,
|
| 462 |
+
"learning_rate": 1.5620833778521306e-05,
|
| 463 |
+
"loss": 0.0866,
|
| 464 |
+
"step": 43
|
| 465 |
+
},
|
| 466 |
+
{
|
| 467 |
+
"clip_ratio": 0.008272722363471985,
|
| 468 |
+
"epoch": 1.568888888888889,
|
| 469 |
+
"grad_norm": 1.2550063449844295,
|
| 470 |
+
"kl": 0.04656219482421875,
|
| 471 |
+
"learning_rate": 1.5358267949789968e-05,
|
| 472 |
+
"loss": 0.0502,
|
| 473 |
+
"step": 44
|
| 474 |
+
},
|
| 475 |
+
{
|
| 476 |
+
"clip_ratio": 0.0,
|
| 477 |
+
"completion_length": 2313.542510986328,
|
| 478 |
+
"epoch": 1.6044444444444443,
|
| 479 |
+
"grad_norm": 0.1122939508940078,
|
| 480 |
+
"kl": 0.038604736328125,
|
| 481 |
+
"learning_rate": 1.5090414157503715e-05,
|
| 482 |
+
"loss": 0.0762,
|
| 483 |
+
"reward": -9.223605461418629,
|
| 484 |
+
"reward_std": 3.827972359955311,
|
| 485 |
+
"rewards/cot_length_penalty_reward": -9.58521255850792,
|
| 486 |
+
"rewards/math_latex_accuracy_reward": 0.36160715692676604,
|
| 487 |
+
"step": 45
|
| 488 |
+
},
|
| 489 |
+
{
|
| 490 |
+
"clip_ratio": 0.0038211173960007727,
|
| 491 |
+
"epoch": 1.6400000000000001,
|
| 492 |
+
"grad_norm": 0.12378389214572502,
|
| 493 |
+
"kl": 0.04041290283203125,
|
| 494 |
+
"learning_rate": 1.4817536741017153e-05,
|
| 495 |
+
"loss": 0.0756,
|
| 496 |
+
"step": 46
|
| 497 |
+
},
|
| 498 |
+
{
|
| 499 |
+
"clip_ratio": 0.005922177180764265,
|
| 500 |
+
"epoch": 1.6755555555555555,
|
| 501 |
+
"grad_norm": 0.12771635689812005,
|
| 502 |
+
"kl": 0.04157257080078125,
|
| 503 |
+
"learning_rate": 1.4539904997395468e-05,
|
| 504 |
+
"loss": 0.0745,
|
| 505 |
+
"step": 47
|
| 506 |
+
},
|
| 507 |
+
{
|
| 508 |
+
"clip_ratio": 0.006733638554578647,
|
| 509 |
+
"epoch": 1.7111111111111112,
|
| 510 |
+
"grad_norm": 0.10941185512569215,
|
| 511 |
+
"kl": 0.04193115234375,
|
| 512 |
+
"learning_rate": 1.4257792915650728e-05,
|
| 513 |
+
"loss": 0.0731,
|
| 514 |
+
"step": 48
|
| 515 |
+
},
|
| 516 |
+
{
|
| 517 |
+
"clip_ratio": 0.0,
|
| 518 |
+
"completion_length": 1813.1072387695312,
|
| 519 |
+
"epoch": 1.7466666666666666,
|
| 520 |
+
"grad_norm": 3.0104818633298525,
|
| 521 |
+
"kl": 0.2118988037109375,
|
| 522 |
+
"learning_rate": 1.3971478906347806e-05,
|
| 523 |
+
"loss": -0.0205,
|
| 524 |
+
"reward": -10.289654642343521,
|
| 525 |
+
"reward_std": 3.131831008940935,
|
| 526 |
+
"rewards/cot_length_penalty_reward": -10.762868821620941,
|
| 527 |
+
"rewards/math_latex_accuracy_reward": 0.47321430779993534,
|
| 528 |
+
"step": 49
|
| 529 |
+
},
|
| 530 |
+
{
|
| 531 |
+
"clip_ratio": 0.002227201643108856,
|
| 532 |
+
"epoch": 1.7822222222222224,
|
| 533 |
+
"grad_norm": 0.11902180384589063,
|
| 534 |
+
"kl": 0.04238128662109375,
|
| 535 |
+
"learning_rate": 1.3681245526846782e-05,
|
| 536 |
+
"loss": -0.0276,
|
| 537 |
+
"step": 50
|
| 538 |
+
},
|
| 539 |
+
{
|
| 540 |
+
"clip_ratio": 0.0032200364221353084,
|
| 541 |
+
"epoch": 1.8177777777777777,
|
| 542 |
+
"grad_norm": 0.1256236704087802,
|
| 543 |
+
"kl": 0.04290008544921875,
|
| 544 |
+
"learning_rate": 1.3387379202452917e-05,
|
| 545 |
+
"loss": -0.0286,
|
| 546 |
+
"step": 51
|
| 547 |
+
},
|
| 548 |
+
{
|
| 549 |
+
"clip_ratio": 0.003942100578569807,
|
| 550 |
+
"epoch": 1.8533333333333335,
|
| 551 |
+
"grad_norm": 0.10119215538963353,
|
| 552 |
+
"kl": 0.0430450439453125,
|
| 553 |
+
"learning_rate": 1.3090169943749475e-05,
|
| 554 |
+
"loss": -0.03,
|
| 555 |
+
"step": 52
|
| 556 |
+
},
|
| 557 |
+
{
|
| 558 |
+
"clip_ratio": 0.0,
|
| 559 |
+
"completion_length": 2273.6608276367188,
|
| 560 |
+
"epoch": 1.8888888888888888,
|
| 561 |
+
"grad_norm": 0.17450385200895388,
|
| 562 |
+
"kl": 0.05017852783203125,
|
| 563 |
+
"learning_rate": 1.2789911060392295e-05,
|
| 564 |
+
"loss": 0.005,
|
| 565 |
+
"reward": -7.135257016867399,
|
| 566 |
+
"reward_std": 3.697649233043194,
|
| 567 |
+
"rewards/cot_length_penalty_reward": -7.5548999309539795,
|
| 568 |
+
"rewards/math_latex_accuracy_reward": 0.4196428684517741,
|
| 569 |
+
"step": 53
|
| 570 |
+
},
|
| 571 |
+
{
|
| 572 |
+
"clip_ratio": 0.002888819137297105,
|
| 573 |
+
"epoch": 1.9244444444444444,
|
| 574 |
+
"grad_norm": 0.10967403835477697,
|
| 575 |
+
"kl": 0.04810333251953125,
|
| 576 |
+
"learning_rate": 1.2486898871648552e-05,
|
| 577 |
+
"loss": 0.0038,
|
| 578 |
+
"step": 54
|
| 579 |
+
},
|
| 580 |
+
{
|
| 581 |
+
"clip_ratio": 0.005481840795255266,
|
| 582 |
+
"epoch": 1.96,
|
| 583 |
+
"grad_norm": 0.14987018246678602,
|
| 584 |
+
"kl": 0.05097198486328125,
|
| 585 |
+
"learning_rate": 1.2181432413965428e-05,
|
| 586 |
+
"loss": 0.0028,
|
| 587 |
+
"step": 55
|
| 588 |
+
},
|
| 589 |
+
{
|
| 590 |
+
"clip_ratio": 0.007852705341065302,
|
| 591 |
+
"epoch": 1.9955555555555555,
|
| 592 |
+
"grad_norm": 0.12818843141531608,
|
| 593 |
+
"kl": 0.0562286376953125,
|
| 594 |
+
"learning_rate": 1.187381314585725e-05,
|
| 595 |
+
"loss": 0.0013,
|
| 596 |
+
"step": 56
|
| 597 |
+
},
|
| 598 |
+
{
|
| 599 |
+
"clip_ratio": 0.0,
|
| 600 |
+
"completion_length": 2008.5916328430176,
|
| 601 |
+
"epoch": 2.0355555555555553,
|
| 602 |
+
"grad_norm": 0.1643995846810127,
|
| 603 |
+
"kl": 0.0548858642578125,
|
| 604 |
+
"learning_rate": 1.156434465040231e-05,
|
| 605 |
+
"loss": 0.0118,
|
| 606 |
+
"reward": -7.901306234300137,
|
| 607 |
+
"reward_std": 2.6794423200190067,
|
| 608 |
+
"rewards/cot_length_penalty_reward": -8.209341906011105,
|
| 609 |
+
"rewards/math_latex_accuracy_reward": 0.3080357303842902,
|
| 610 |
+
"step": 57
|
| 611 |
+
},
|
| 612 |
+
{
|
| 613 |
+
"clip_ratio": 0.0029279392474563792,
|
| 614 |
+
"epoch": 2.071111111111111,
|
| 615 |
+
"grad_norm": 0.12045324524183162,
|
| 616 |
+
"kl": 0.0587005615234375,
|
| 617 |
+
"learning_rate": 1.1253332335643043e-05,
|
| 618 |
+
"loss": 0.0108,
|
| 619 |
+
"step": 58
|
| 620 |
+
},
|
| 621 |
+
{
|
| 622 |
+
"clip_ratio": 0.005807226421893574,
|
| 623 |
+
"epoch": 2.1066666666666665,
|
| 624 |
+
"grad_norm": 0.14841036250602688,
|
| 625 |
+
"kl": 0.0640106201171875,
|
| 626 |
+
"learning_rate": 1.0941083133185146e-05,
|
| 627 |
+
"loss": 0.0097,
|
| 628 |
+
"step": 59
|
| 629 |
+
},
|
| 630 |
+
{
|
| 631 |
+
"epoch": 2.1422222222222222,
|
| 632 |
+
"grad_norm": 0.11089468253649072,
|
| 633 |
+
"learning_rate": 1.0627905195293135e-05,
|
| 634 |
+
"loss": 0.0084,
|
| 635 |
+
"step": 60
|
| 636 |
+
},
|
| 637 |
+
{
|
| 638 |
+
"epoch": 2.1422222222222222,
|
| 639 |
+
"eval_clip_ratio": 0.0,
|
| 640 |
+
"eval_completion_length": 2107.960148737981,
|
| 641 |
+
"eval_kl": 0.05983323317307692,
|
| 642 |
+
"eval_loss": -0.00015631201677024364,
|
| 643 |
+
"eval_reward": -8.494354761563814,
|
| 644 |
+
"eval_reward_std": 3.4025442325151882,
|
| 645 |
+
"eval_rewards/cot_length_penalty_reward": -8.876223013951229,
|
| 646 |
+
"eval_rewards/math_latex_accuracy_reward": 0.3818681509448932,
|
| 647 |
+
"eval_runtime": 422.1329,
|
| 648 |
+
"eval_samples_per_second": 0.118,
|
| 649 |
+
"eval_steps_per_second": 0.005,
|
| 650 |
+
"step": 60
|
| 651 |
+
},
|
| 652 |
+
{
|
| 653 |
+
"clip_ratio": 0.003193242686393205,
|
| 654 |
+
"completion_length": 1800.6674766540527,
|
| 655 |
+
"epoch": 2.1777777777777776,
|
| 656 |
+
"grad_norm": 0.1697521453747013,
|
| 657 |
+
"kl": 0.065582275390625,
|
| 658 |
+
"learning_rate": 1.0314107590781284e-05,
|
| 659 |
+
"loss": 0.0174,
|
| 660 |
+
"reward": -8.092556223273277,
|
| 661 |
+
"reward_std": 3.146493151783943,
|
| 662 |
+
"rewards/cot_length_penalty_reward": -8.458627462387085,
|
| 663 |
+
"rewards/math_latex_accuracy_reward": 0.3660714477300644,
|
| 664 |
+
"step": 61
|
| 665 |
+
},
|
| 666 |
+
{
|
| 667 |
+
"clip_ratio": 0.003192656353348866,
|
| 668 |
+
"epoch": 2.2133333333333334,
|
| 669 |
+
"grad_norm": 0.12330763778596482,
|
| 670 |
+
"kl": 0.0718231201171875,
|
| 671 |
+
"learning_rate": 1e-05,
|
| 672 |
+
"loss": 0.0163,
|
| 673 |
+
"step": 62
|
| 674 |
+
},
|
| 675 |
+
{
|
| 676 |
+
"clip_ratio": 0.0062453514110529795,
|
| 677 |
+
"epoch": 2.2488888888888887,
|
| 678 |
+
"grad_norm": 0.16098784282181033,
|
| 679 |
+
"kl": 0.078704833984375,
|
| 680 |
+
"learning_rate": 9.685892409218718e-06,
|
| 681 |
+
"loss": 0.0151,
|
| 682 |
+
"step": 63
|
| 683 |
+
},
|
| 684 |
+
{
|
| 685 |
+
"clip_ratio": 0.006978008910664357,
|
| 686 |
+
"epoch": 2.2844444444444445,
|
| 687 |
+
"grad_norm": 0.1406450810476633,
|
| 688 |
+
"kl": 0.0782470703125,
|
| 689 |
+
"learning_rate": 9.372094804706867e-06,
|
| 690 |
+
"loss": 0.0137,
|
| 691 |
+
"step": 64
|
| 692 |
+
},
|
| 693 |
+
{
|
| 694 |
+
"clip_ratio": 0.0,
|
| 695 |
+
"completion_length": 2595.997859954834,
|
| 696 |
+
"epoch": 2.32,
|
| 697 |
+
"grad_norm": 0.18250322704566987,
|
| 698 |
+
"kl": 0.0649871826171875,
|
| 699 |
+
"learning_rate": 9.058916866814857e-06,
|
| 700 |
+
"loss": 0.0147,
|
| 701 |
+
"reward": -9.348549716174603,
|
| 702 |
+
"reward_std": 3.3840084299445152,
|
| 703 |
+
"rewards/cot_length_penalty_reward": -9.70346000418067,
|
| 704 |
+
"rewards/math_latex_accuracy_reward": 0.35491072852164507,
|
| 705 |
+
"step": 65
|
| 706 |
+
},
|
| 707 |
+
{
|
| 708 |
+
"clip_ratio": 0.0031114893354242668,
|
| 709 |
+
"epoch": 2.3555555555555556,
|
| 710 |
+
"grad_norm": 0.13005553219968966,
|
| 711 |
+
"kl": 0.0695648193359375,
|
| 712 |
+
"learning_rate": 8.746667664356957e-06,
|
| 713 |
+
"loss": 0.014,
|
| 714 |
+
"step": 66
|
| 715 |
+
},
|
| 716 |
+
{
|
| 717 |
+
"clip_ratio": 0.0075038159266114235,
|
| 718 |
+
"epoch": 2.391111111111111,
|
| 719 |
+
"grad_norm": 0.19621512659848725,
|
| 720 |
+
"kl": 0.0780181884765625,
|
| 721 |
+
"learning_rate": 8.43565534959769e-06,
|
| 722 |
+
"loss": 0.0133,
|
| 723 |
+
"step": 67
|
| 724 |
+
},
|
| 725 |
+
{
|
| 726 |
+
"clip_ratio": 0.006932365708053112,
|
| 727 |
+
"epoch": 2.4266666666666667,
|
| 728 |
+
"grad_norm": 0.13215694629988284,
|
| 729 |
+
"kl": 0.07647705078125,
|
| 730 |
+
"learning_rate": 8.126186854142752e-06,
|
| 731 |
+
"loss": 0.0122,
|
| 732 |
+
"step": 68
|
| 733 |
+
},
|
| 734 |
+
{
|
| 735 |
+
"clip_ratio": 0.0,
|
| 736 |
+
"completion_length": 2154.4286880493164,
|
| 737 |
+
"epoch": 2.462222222222222,
|
| 738 |
+
"grad_norm": 0.27197173025048144,
|
| 739 |
+
"kl": 0.086151123046875,
|
| 740 |
+
"learning_rate": 7.818567586034578e-06,
|
| 741 |
+
"loss": 0.0247,
|
| 742 |
+
"reward": -8.337913118302822,
|
| 743 |
+
"reward_std": 3.0677984952926636,
|
| 744 |
+
"rewards/cot_length_penalty_reward": -8.806663155555725,
|
| 745 |
+
"rewards/math_latex_accuracy_reward": 0.4687500186264515,
|
| 746 |
+
"step": 69
|
| 747 |
+
},
|
| 748 |
+
{
|
| 749 |
+
"clip_ratio": 0.005053140237578191,
|
| 750 |
+
"epoch": 2.497777777777778,
|
| 751 |
+
"grad_norm": 0.20964263197545416,
|
| 752 |
+
"kl": 0.0977630615234375,
|
| 753 |
+
"learning_rate": 7.513101128351454e-06,
|
| 754 |
+
"loss": 0.0237,
|
| 755 |
+
"step": 70
|
| 756 |
+
},
|
| 757 |
+
{
|
| 758 |
+
"clip_ratio": 0.005771905358415097,
|
| 759 |
+
"epoch": 2.533333333333333,
|
| 760 |
+
"grad_norm": 0.15787820635605407,
|
| 761 |
+
"kl": 0.0987091064453125,
|
| 762 |
+
"learning_rate": 7.210088939607709e-06,
|
| 763 |
+
"loss": 0.0226,
|
| 764 |
+
"step": 71
|
| 765 |
+
},
|
| 766 |
+
{
|
| 767 |
+
"clip_ratio": 0.0062158564978744835,
|
| 768 |
+
"epoch": 2.568888888888889,
|
| 769 |
+
"grad_norm": 0.4007267449310534,
|
| 770 |
+
"kl": 0.0895538330078125,
|
| 771 |
+
"learning_rate": 6.909830056250527e-06,
|
| 772 |
+
"loss": 0.022,
|
| 773 |
+
"step": 72
|
| 774 |
+
},
|
| 775 |
+
{
|
| 776 |
+
"clip_ratio": 0.0,
|
| 777 |
+
"completion_length": 1935.1675262451172,
|
| 778 |
+
"epoch": 2.6044444444444443,
|
| 779 |
+
"grad_norm": 0.266781012315019,
|
| 780 |
+
"kl": 0.10394287109375,
|
| 781 |
+
"learning_rate": 6.612620797547087e-06,
|
| 782 |
+
"loss": 0.0125,
|
| 783 |
+
"reward": -7.354442303534597,
|
| 784 |
+
"reward_std": 2.94980551302433,
|
| 785 |
+
"rewards/cot_length_penalty_reward": -7.771853107959032,
|
| 786 |
+
"rewards/math_latex_accuracy_reward": 0.41741072852164507,
|
| 787 |
+
"step": 73
|
| 788 |
+
},
|
| 789 |
+
{
|
| 790 |
+
"clip_ratio": 0.01473489188356325,
|
| 791 |
+
"epoch": 2.64,
|
| 792 |
+
"grad_norm": 0.542093788713072,
|
| 793 |
+
"kl": 0.1417083740234375,
|
| 794 |
+
"learning_rate": 6.318754473153221e-06,
|
| 795 |
+
"loss": 0.0132,
|
| 796 |
+
"step": 74
|
| 797 |
+
},
|
| 798 |
+
{
|
| 799 |
+
"clip_ratio": 0.009351018321467564,
|
| 800 |
+
"epoch": 2.6755555555555555,
|
| 801 |
+
"grad_norm": 0.32832820257493534,
|
| 802 |
+
"kl": 0.1302490234375,
|
| 803 |
+
"learning_rate": 6.028521093652195e-06,
|
| 804 |
+
"loss": 0.0111,
|
| 805 |
+
"step": 75
|
| 806 |
+
},
|
| 807 |
+
{
|
| 808 |
+
"clip_ratio": 0.008401441504247487,
|
| 809 |
+
"epoch": 2.7111111111111112,
|
| 810 |
+
"grad_norm": 0.5313671762370776,
|
| 811 |
+
"kl": 0.106719970703125,
|
| 812 |
+
"learning_rate": 5.742207084349274e-06,
|
| 813 |
+
"loss": 0.0105,
|
| 814 |
+
"step": 76
|
| 815 |
+
},
|
| 816 |
+
{
|
| 817 |
+
"clip_ratio": 0.0,
|
| 818 |
+
"completion_length": 1853.8215065002441,
|
| 819 |
+
"epoch": 2.7466666666666666,
|
| 820 |
+
"grad_norm": 0.25601743476635025,
|
| 821 |
+
"kl": 0.128814697265625,
|
| 822 |
+
"learning_rate": 5.460095002604533e-06,
|
| 823 |
+
"loss": -0.018,
|
| 824 |
+
"reward": -7.114141087979078,
|
| 825 |
+
"reward_std": 2.6430138647556305,
|
| 826 |
+
"rewards/cot_length_penalty_reward": -7.493605274707079,
|
| 827 |
+
"rewards/math_latex_accuracy_reward": 0.37946430314332247,
|
| 828 |
+
"step": 77
|
| 829 |
+
},
|
| 830 |
+
{
|
| 831 |
+
"clip_ratio": 0.004884305511950515,
|
| 832 |
+
"epoch": 2.7822222222222224,
|
| 833 |
+
"grad_norm": 0.18667971276259676,
|
| 834 |
+
"kl": 0.1357421875,
|
| 835 |
+
"learning_rate": 5.1824632589828465e-06,
|
| 836 |
+
"loss": -0.019,
|
| 837 |
+
"step": 78
|
| 838 |
+
},
|
| 839 |
+
{
|
| 840 |
+
"clip_ratio": 0.008678867772687227,
|
| 841 |
+
"epoch": 2.8177777777777777,
|
| 842 |
+
"grad_norm": 0.2515967343714355,
|
| 843 |
+
"kl": 0.1392822265625,
|
| 844 |
+
"learning_rate": 4.909585842496287e-06,
|
| 845 |
+
"loss": -0.0199,
|
| 846 |
+
"step": 79
|
| 847 |
+
},
|
| 848 |
+
{
|
| 849 |
+
"clip_ratio": 0.008155457631801255,
|
| 850 |
+
"epoch": 2.8533333333333335,
|
| 851 |
+
"grad_norm": 0.18942366870295294,
|
| 852 |
+
"kl": 0.131805419921875,
|
| 853 |
+
"learning_rate": 4.641732050210032e-06,
|
| 854 |
+
"loss": -0.0211,
|
| 855 |
+
"step": 80
|
| 856 |
+
},
|
| 857 |
+
{
|
| 858 |
+
"clip_ratio": 0.0,
|
| 859 |
+
"completion_length": 2211.9219856262207,
|
| 860 |
+
"epoch": 2.888888888888889,
|
| 861 |
+
"grad_norm": 0.22043905749174672,
|
| 862 |
+
"kl": 0.1049957275390625,
|
| 863 |
+
"learning_rate": 4.379166221478697e-06,
|
| 864 |
+
"loss": -0.0247,
|
| 865 |
+
"reward": -9.63335988484323,
|
| 866 |
+
"reward_std": 2.9292308390140533,
|
| 867 |
+
"rewards/cot_length_penalty_reward": -10.111038556322455,
|
| 868 |
+
"rewards/math_latex_accuracy_reward": 0.4776785969734192,
|
| 869 |
+
"step": 81
|
| 870 |
+
},
|
| 871 |
+
{
|
| 872 |
+
"clip_ratio": 0.002452510700095445,
|
| 873 |
+
"epoch": 2.924444444444444,
|
| 874 |
+
"grad_norm": 0.2264032851218282,
|
| 875 |
+
"kl": 0.1047821044921875,
|
| 876 |
+
"learning_rate": 4.12214747707527e-06,
|
| 877 |
+
"loss": -0.0248,
|
| 878 |
+
"step": 82
|
| 879 |
+
},
|
| 880 |
+
{
|
| 881 |
+
"clip_ratio": 0.0035177832323824987,
|
| 882 |
+
"epoch": 2.96,
|
| 883 |
+
"grad_norm": 0.14675046732213165,
|
| 884 |
+
"kl": 0.1127166748046875,
|
| 885 |
+
"learning_rate": 3.8709294634702374e-06,
|
| 886 |
+
"loss": -0.0259,
|
| 887 |
+
"step": 83
|
| 888 |
+
},
|
| 889 |
+
{
|
| 890 |
+
"clip_ratio": 0.0036856129445368424,
|
| 891 |
+
"epoch": 2.9955555555555557,
|
| 892 |
+
"grad_norm": 0.1630420640767936,
|
| 893 |
+
"kl": 0.09552001953125,
|
| 894 |
+
"learning_rate": 3.625760102513103e-06,
|
| 895 |
+
"loss": -0.0267,
|
| 896 |
+
"step": 84
|
| 897 |
+
},
|
| 898 |
+
{
|
| 899 |
+
"clip_ratio": 0.0,
|
| 900 |
+
"completion_length": 1545.3081169128418,
|
| 901 |
+
"epoch": 3.0355555555555553,
|
| 902 |
+
"grad_norm": 14.400966954596354,
|
| 903 |
+
"kl": 0.286773681640625,
|
| 904 |
+
"learning_rate": 3.3868813467634833e-06,
|
| 905 |
+
"loss": -0.0265,
|
| 906 |
+
"reward": -7.1288284212350845,
|
| 907 |
+
"reward_std": 1.927463386207819,
|
| 908 |
+
"rewards/cot_length_penalty_reward": -7.606507122516632,
|
| 909 |
+
"rewards/math_latex_accuracy_reward": 0.47767859511077404,
|
| 910 |
+
"step": 85
|
| 911 |
+
},
|
| 912 |
+
{
|
| 913 |
+
"clip_ratio": 0.0029736382130067796,
|
| 914 |
+
"epoch": 3.071111111111111,
|
| 915 |
+
"grad_norm": 0.46033460120838376,
|
| 916 |
+
"kl": 0.130645751953125,
|
| 917 |
+
"learning_rate": 3.1545289407131128e-06,
|
| 918 |
+
"loss": -0.0322,
|
| 919 |
+
"step": 86
|
| 920 |
+
},
|
| 921 |
+
{
|
| 922 |
+
"clip_ratio": 0.0043519225146155804,
|
| 923 |
+
"epoch": 3.1066666666666665,
|
| 924 |
+
"grad_norm": 0.2864629379691617,
|
| 925 |
+
"kl": 0.13775634765625,
|
| 926 |
+
"learning_rate": 2.9289321881345257e-06,
|
| 927 |
+
"loss": -0.0338,
|
| 928 |
+
"step": 87
|
| 929 |
+
},
|
| 930 |
+
{
|
| 931 |
+
"clip_ratio": 0.009186911847791635,
|
| 932 |
+
"epoch": 3.1422222222222222,
|
| 933 |
+
"grad_norm": 0.24547967049213823,
|
| 934 |
+
"kl": 0.155426025390625,
|
| 935 |
+
"learning_rate": 2.7103137257858867e-06,
|
| 936 |
+
"loss": -0.0347,
|
| 937 |
+
"step": 88
|
| 938 |
+
},
|
| 939 |
+
{
|
| 940 |
+
"clip_ratio": 0.0,
|
| 941 |
+
"completion_length": 1969.4442825317383,
|
| 942 |
+
"epoch": 3.1777777777777776,
|
| 943 |
+
"grad_norm": 0.433485726828516,
|
| 944 |
+
"kl": 0.1619415283203125,
|
| 945 |
+
"learning_rate": 2.4988893036954045e-06,
|
| 946 |
+
"loss": -0.0013,
|
| 947 |
+
"reward": -8.979866623878479,
|
| 948 |
+
"reward_std": 2.524892296642065,
|
| 949 |
+
"rewards/cot_length_penalty_reward": -9.319152384996414,
|
| 950 |
+
"rewards/math_latex_accuracy_reward": 0.339285729220137,
|
| 951 |
+
"step": 89
|
| 952 |
+
},
|
| 953 |
+
{
|
| 954 |
+
"epoch": 3.2133333333333334,
|
| 955 |
+
"grad_norm": 0.2766265099056435,
|
| 956 |
+
"learning_rate": 2.2948675722421086e-06,
|
| 957 |
+
"loss": -0.003,
|
| 958 |
+
"step": 90
|
| 959 |
+
},
|
| 960 |
+
{
|
| 961 |
+
"epoch": 3.2133333333333334,
|
| 962 |
+
"eval_clip_ratio": 0.0,
|
| 963 |
+
"eval_completion_length": 2205.780292217548,
|
| 964 |
+
"eval_kl": 0.18556565504807693,
|
| 965 |
+
"eval_loss": 0.01654699072241783,
|
| 966 |
+
"eval_reward": -7.743022455332371,
|
| 967 |
+
"eval_reward_std": 2.8156597109941335,
|
| 968 |
+
"eval_rewards/cot_length_penalty_reward": -8.072692573070526,
|
| 969 |
+
"eval_rewards/math_latex_accuracy_reward": 0.32967034670022816,
|
| 970 |
+
"eval_runtime": 490.0675,
|
| 971 |
+
"eval_samples_per_second": 0.102,
|
| 972 |
+
"eval_steps_per_second": 0.004,
|
| 973 |
+
"step": 90
|
| 974 |
+
},
|
| 975 |
+
{
|
| 976 |
+
"clip_ratio": 0.006224101292900741,
|
| 977 |
+
"epoch": 3.2488888888888887,
|
| 978 |
+
"grad_norm": 0.4063813985520313,
|
| 979 |
+
"kl": 0.197174072265625,
|
| 980 |
+
"learning_rate": 2.098449876243096e-06,
|
| 981 |
+
"loss": -0.0037,
|
| 982 |
+
"step": 91
|
| 983 |
+
},
|
| 984 |
+
{
|
| 985 |
+
"clip_ratio": 0.009523139509838074,
|
| 986 |
+
"epoch": 3.2844444444444445,
|
| 987 |
+
"grad_norm": 0.26707887580226697,
|
| 988 |
+
"kl": 0.196014404296875,
|
| 989 |
+
"learning_rate": 1.9098300562505266e-06,
|
| 990 |
+
"loss": -0.0047,
|
| 991 |
+
"step": 92
|
| 992 |
+
},
|
| 993 |
+
{
|
| 994 |
+
"clip_ratio": 0.0,
|
| 995 |
+
"completion_length": 2093.5982818603516,
|
| 996 |
+
"epoch": 3.32,
|
| 997 |
+
"grad_norm": 0.3581002181636425,
|
| 998 |
+
"kl": 0.199798583984375,
|
| 999 |
+
"learning_rate": 1.7291942572543806e-06,
|
| 1000 |
+
"loss": 0.0178,
|
| 1001 |
+
"reward": -7.027399688959122,
|
| 1002 |
+
"reward_std": 2.7810670882463455,
|
| 1003 |
+
"rewards/cot_length_penalty_reward": -7.435881897807121,
|
| 1004 |
+
"rewards/math_latex_accuracy_reward": 0.40848216507583857,
|
| 1005 |
+
"step": 93
|
| 1006 |
+
},
|
| 1007 |
+
{
|
| 1008 |
+
"clip_ratio": 0.002175023495510686,
|
| 1009 |
+
"epoch": 3.3555555555555556,
|
| 1010 |
+
"grad_norm": 0.35335279380104717,
|
| 1011 |
+
"kl": 0.19476318359375,
|
| 1012 |
+
"learning_rate": 1.5567207449798517e-06,
|
| 1013 |
+
"loss": 0.017,
|
| 1014 |
+
"step": 94
|
| 1015 |
+
},
|
| 1016 |
+
{
|
| 1017 |
+
"clip_ratio": 0.0038238488195929676,
|
| 1018 |
+
"epoch": 3.391111111111111,
|
| 1019 |
+
"grad_norm": 0.26816647206060557,
|
| 1020 |
+
"kl": 0.21075439453125,
|
| 1021 |
+
"learning_rate": 1.3925797299605649e-06,
|
| 1022 |
+
"loss": 0.0159,
|
| 1023 |
+
"step": 95
|
| 1024 |
+
},
|
| 1025 |
+
{
|
| 1026 |
+
"clip_ratio": 0.006063876280677505,
|
| 1027 |
+
"epoch": 3.4266666666666667,
|
| 1028 |
+
"grad_norm": 0.3302248399941887,
|
| 1029 |
+
"kl": 0.22137451171875,
|
| 1030 |
+
"learning_rate": 1.2369331995613664e-06,
|
| 1031 |
+
"loss": 0.0151,
|
| 1032 |
+
"step": 96
|
| 1033 |
+
},
|
| 1034 |
+
{
|
| 1035 |
+
"clip_ratio": 0.0,
|
| 1036 |
+
"completion_length": 2630.805938720703,
|
| 1037 |
+
"epoch": 3.462222222222222,
|
| 1038 |
+
"grad_norm": 1.7166272647509115,
|
| 1039 |
+
"kl": 0.281219482421875,
|
| 1040 |
+
"learning_rate": 1.0899347581163222e-06,
|
| 1041 |
+
"loss": 0.0839,
|
| 1042 |
+
"reward": -7.589185383694712,
|
| 1043 |
+
"reward_std": 3.134066376835108,
|
| 1044 |
+
"rewards/cot_length_penalty_reward": -7.917310604825616,
|
| 1045 |
+
"rewards/math_latex_accuracy_reward": 0.32812501839362085,
|
| 1046 |
+
"step": 97
|
| 1047 |
+
},
|
| 1048 |
+
{
|
| 1049 |
+
"clip_ratio": 0.007515597055316903,
|
| 1050 |
+
"epoch": 3.497777777777778,
|
| 1051 |
+
"grad_norm": 4.303565090937016,
|
| 1052 |
+
"kl": 0.204925537109375,
|
| 1053 |
+
"learning_rate": 9.517294753398066e-07,
|
| 1054 |
+
"loss": 0.0869,
|
| 1055 |
+
"step": 98
|
| 1056 |
+
},
|
| 1057 |
+
{
|
| 1058 |
+
"clip_ratio": 0.0075942349576507695,
|
| 1059 |
+
"epoch": 3.533333333333333,
|
| 1060 |
+
"grad_norm": 2.9889664346961444,
|
| 1061 |
+
"kl": 0.2061920166015625,
|
| 1062 |
+
"learning_rate": 8.224537431601886e-07,
|
| 1063 |
+
"loss": 0.0841,
|
| 1064 |
+
"step": 99
|
| 1065 |
+
},
|
| 1066 |
+
{
|
| 1067 |
+
"clip_ratio": 0.0051619461009977385,
|
| 1068 |
+
"epoch": 3.568888888888889,
|
| 1069 |
+
"grad_norm": 0.4431396626395719,
|
| 1070 |
+
"kl": 0.22930908203125,
|
| 1071 |
+
"learning_rate": 7.022351411174866e-07,
|
| 1072 |
+
"loss": 0.0814,
|
| 1073 |
+
"step": 100
|
| 1074 |
+
},
|
| 1075 |
+
{
|
| 1076 |
+
"clip_ratio": 0.0,
|
| 1077 |
+
"completion_length": 2545.8371925354004,
|
| 1078 |
+
"epoch": 3.6044444444444443,
|
| 1079 |
+
"grad_norm": 0.4301809872674547,
|
| 1080 |
+
"kl": 0.171051025390625,
|
| 1081 |
+
"learning_rate": 5.911923104577455e-07,
|
| 1082 |
+
"loss": 0.0196,
|
| 1083 |
+
"reward": -10.28242233581841,
|
| 1084 |
+
"reward_std": 3.0546065159142017,
|
| 1085 |
+
"rewards/cot_length_penalty_reward": -10.666351079940796,
|
| 1086 |
+
"rewards/math_latex_accuracy_reward": 0.3839285857975483,
|
| 1087 |
+
"step": 101
|
| 1088 |
+
},
|
| 1089 |
+
{
|
| 1090 |
+
"clip_ratio": 0.0024334693371201865,
|
| 1091 |
+
"epoch": 3.64,
|
| 1092 |
+
"grad_norm": 0.40563497309844976,
|
| 1093 |
+
"kl": 0.19537353515625,
|
| 1094 |
+
"learning_rate": 4.894348370484648e-07,
|
| 1095 |
+
"loss": 0.0191,
|
| 1096 |
+
"step": 102
|
| 1097 |
+
},
|
| 1098 |
+
{
|
| 1099 |
+
"clip_ratio": 0.00391879488597624,
|
| 1100 |
+
"epoch": 3.6755555555555555,
|
| 1101 |
+
"grad_norm": 0.5720892424115457,
|
| 1102 |
+
"kl": 0.21160888671875,
|
| 1103 |
+
"learning_rate": 3.9706314323056936e-07,
|
| 1104 |
+
"loss": 0.0191,
|
| 1105 |
+
"step": 103
|
| 1106 |
+
},
|
| 1107 |
+
{
|
| 1108 |
+
"clip_ratio": 0.005467013252200559,
|
| 1109 |
+
"epoch": 3.7111111111111112,
|
| 1110 |
+
"grad_norm": 0.5349890131537904,
|
| 1111 |
+
"kl": 0.21337890625,
|
| 1112 |
+
"learning_rate": 3.1416838871368925e-07,
|
| 1113 |
+
"loss": 0.0189,
|
| 1114 |
+
"step": 104
|
| 1115 |
+
},
|
| 1116 |
+
{
|
| 1117 |
+
"clip_ratio": 0.0,
|
| 1118 |
+
"completion_length": 2160.7389335632324,
|
| 1119 |
+
"epoch": 3.7466666666666666,
|
| 1120 |
+
"grad_norm": 13.207091106199918,
|
| 1121 |
+
"kl": 0.7435302734375,
|
| 1122 |
+
"learning_rate": 2.4083238061252565e-07,
|
| 1123 |
+
"loss": 0.0564,
|
| 1124 |
+
"reward": -7.942288625985384,
|
| 1125 |
+
"reward_std": 2.594830472022295,
|
| 1126 |
+
"rewards/cot_length_penalty_reward": -8.37309193611145,
|
| 1127 |
+
"rewards/math_latex_accuracy_reward": 0.43080358672887087,
|
| 1128 |
+
"step": 105
|
| 1129 |
+
},
|
| 1130 |
+
{
|
| 1131 |
+
"clip_ratio": 0.0028747237083734944,
|
| 1132 |
+
"epoch": 3.7822222222222224,
|
| 1133 |
+
"grad_norm": 3.5352628792853067,
|
| 1134 |
+
"kl": 0.472412109375,
|
| 1135 |
+
"learning_rate": 1.7712749271311392e-07,
|
| 1136 |
+
"loss": 0.0465,
|
| 1137 |
+
"step": 106
|
| 1138 |
+
},
|
| 1139 |
+
{
|
| 1140 |
+
"clip_ratio": 0.004760361814987846,
|
| 1141 |
+
"epoch": 3.8177777777777777,
|
| 1142 |
+
"grad_norm": 0.94419108473764,
|
| 1143 |
+
"kl": 0.3760833740234375,
|
| 1144 |
+
"learning_rate": 1.231165940486234e-07,
|
| 1145 |
+
"loss": 0.0439,
|
| 1146 |
+
"step": 107
|
| 1147 |
+
},
|
| 1148 |
+
{
|
| 1149 |
+
"clip_ratio": 0.0061231208674144,
|
| 1150 |
+
"epoch": 3.8533333333333335,
|
| 1151 |
+
"grad_norm": 1.7315584230873846,
|
| 1152 |
+
"kl": 0.3495025634765625,
|
| 1153 |
+
"learning_rate": 7.885298685522235e-08,
|
| 1154 |
+
"loss": 0.044,
|
| 1155 |
+
"step": 108
|
| 1156 |
+
},
|
| 1157 |
+
{
|
| 1158 |
+
"clip_ratio": 0.0,
|
| 1159 |
+
"completion_length": 1982.8750839233398,
|
| 1160 |
+
"epoch": 3.888888888888889,
|
| 1161 |
+
"grad_norm": 0.7007233776562277,
|
| 1162 |
+
"kl": 0.2796783447265625,
|
| 1163 |
+
"learning_rate": 4.438035396920004e-08,
|
| 1164 |
+
"loss": 0.0207,
|
| 1165 |
+
"reward": -9.675179054960608,
|
| 1166 |
+
"reward_std": 2.574063938111067,
|
| 1167 |
+
"rewards/cot_length_penalty_reward": -9.989911276847124,
|
| 1168 |
+
"rewards/math_latex_accuracy_reward": 0.314732160884887,
|
| 1169 |
+
"step": 109
|
| 1170 |
+
},
|
| 1171 |
+
{
|
| 1172 |
+
"clip_ratio": 0.001981915433134418,
|
| 1173 |
+
"epoch": 3.924444444444444,
|
| 1174 |
+
"grad_norm": 0.6787732749959883,
|
| 1175 |
+
"kl": 0.2747039794921875,
|
| 1176 |
+
"learning_rate": 1.973271571728441e-08,
|
| 1177 |
+
"loss": 0.0208,
|
| 1178 |
+
"step": 110
|
| 1179 |
+
},
|
| 1180 |
+
{
|
| 1181 |
+
"clip_ratio": 0.0019334297030582093,
|
| 1182 |
+
"epoch": 3.96,
|
| 1183 |
+
"grad_norm": 0.6415798052337234,
|
| 1184 |
+
"kl": 0.27728271484375,
|
| 1185 |
+
"learning_rate": 4.9343963426840006e-09,
|
| 1186 |
+
"loss": 0.0206,
|
| 1187 |
+
"step": 111
|
| 1188 |
+
},
|
| 1189 |
+
{
|
| 1190 |
+
"clip_ratio": 0.0017770093054423342,
|
| 1191 |
+
"epoch": 3.9955555555555557,
|
| 1192 |
+
"grad_norm": 0.6316769640873855,
|
| 1193 |
+
"kl": 0.30633544921875,
|
| 1194 |
+
"learning_rate": 0.0,
|
| 1195 |
+
"loss": 0.0207,
|
| 1196 |
+
"step": 112
|
| 1197 |
+
},
|
| 1198 |
+
{
|
| 1199 |
+
"epoch": 3.9955555555555557,
|
| 1200 |
+
"step": 112,
|
| 1201 |
+
"total_flos": 0.0,
|
| 1202 |
+
"train_loss": 0.6775813137989773,
|
| 1203 |
+
"train_runtime": 20674.7257,
|
| 1204 |
+
"train_samples_per_second": 0.087,
|
| 1205 |
+
"train_steps_per_second": 0.005
|
| 1206 |
+
}
|
| 1207 |
+
],
|
| 1208 |
+
"logging_steps": 1,
|
| 1209 |
+
"max_steps": 112,
|
| 1210 |
+
"num_input_tokens_seen": 0,
|
| 1211 |
+
"num_train_epochs": 4,
|
| 1212 |
+
"save_steps": 10,
|
| 1213 |
+
"stateful_callbacks": {
|
| 1214 |
+
"TrainerControl": {
|
| 1215 |
+
"args": {
|
| 1216 |
+
"should_epoch_stop": false,
|
| 1217 |
+
"should_evaluate": false,
|
| 1218 |
+
"should_log": false,
|
| 1219 |
+
"should_save": true,
|
| 1220 |
+
"should_training_stop": true
|
| 1221 |
+
},
|
| 1222 |
+
"attributes": {}
|
| 1223 |
+
}
|
| 1224 |
+
},
|
| 1225 |
+
"total_flos": 0.0,
|
| 1226 |
+
"train_batch_size": 4,
|
| 1227 |
+
"trial_name": null,
|
| 1228 |
+
"trial_params": null
|
| 1229 |
+
}
|