Model save
Browse files- README.md +68 -0
- all_results.json +8 -0
- generation_config.json +14 -0
- train_results.json +8 -0
- trainer_state.json +897 -0
README.md
ADDED
|
@@ -0,0 +1,68 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
base_model: Qwen/Qwen2.5-1.5B-Instruct
|
| 3 |
+
library_name: transformers
|
| 4 |
+
model_name: Qwen2.5-1.5B-Open-R1-Math-GRPO
|
| 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-Math-GRPO
|
| 13 |
+
|
| 14 |
+
This model is a fine-tuned version of [Qwen/Qwen2.5-1.5B-Instruct](https://huggingface.co/Qwen/Qwen2.5-1.5B-Instruct).
|
| 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="jonatatyska/Qwen2.5-1.5B-Open-R1-Math-GRPO", 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/ine-ufsc/huggingface/runs/57dxyp6x)
|
| 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: 4.49.0
|
| 39 |
+
- Pytorch: 2.5.1
|
| 40 |
+
- Datasets: 3.6.0
|
| 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": -0.11113634665512986,
|
| 4 |
+
"train_runtime": 2287.8706,
|
| 5 |
+
"train_samples": 1600,
|
| 6 |
+
"train_samples_per_second": 0.699,
|
| 7 |
+
"train_steps_per_second": 0.025
|
| 8 |
+
}
|
generation_config.json
ADDED
|
@@ -0,0 +1,14 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bos_token_id": 151643,
|
| 3 |
+
"do_sample": true,
|
| 4 |
+
"eos_token_id": [
|
| 5 |
+
151645,
|
| 6 |
+
151643
|
| 7 |
+
],
|
| 8 |
+
"pad_token_id": 151643,
|
| 9 |
+
"repetition_penalty": 1.1,
|
| 10 |
+
"temperature": 0.7,
|
| 11 |
+
"top_k": 20,
|
| 12 |
+
"top_p": 0.8,
|
| 13 |
+
"transformers_version": "4.49.0"
|
| 14 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"total_flos": 0.0,
|
| 3 |
+
"train_loss": -0.11113634665512986,
|
| 4 |
+
"train_runtime": 2287.8706,
|
| 5 |
+
"train_samples": 1600,
|
| 6 |
+
"train_samples_per_second": 0.699,
|
| 7 |
+
"train_steps_per_second": 0.025
|
| 8 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,897 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 0.9956331877729258,
|
| 5 |
+
"eval_steps": 500,
|
| 6 |
+
"global_step": 57,
|
| 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": 611.8683166503906,
|
| 14 |
+
"epoch": 0.017467248908296942,
|
| 15 |
+
"grad_norm": 0.44490134716033936,
|
| 16 |
+
"kl": 0.0,
|
| 17 |
+
"learning_rate": 3.3333333333333333e-06,
|
| 18 |
+
"loss": 0.01,
|
| 19 |
+
"reward": 0.1757812611758709,
|
| 20 |
+
"reward_std": 0.32868771255016327,
|
| 21 |
+
"rewards/format_reward": 0.03125000232830644,
|
| 22 |
+
"rewards/math_reward": 0.0022321429569274187,
|
| 23 |
+
"rewards/tag_count_reward": 0.1422991156578064,
|
| 24 |
+
"step": 1
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"clip_ratio": 0.0,
|
| 28 |
+
"completion_length": 684.4286041259766,
|
| 29 |
+
"epoch": 0.034934497816593885,
|
| 30 |
+
"grad_norm": 0.367574542760849,
|
| 31 |
+
"kl": 0.0,
|
| 32 |
+
"learning_rate": 6.666666666666667e-06,
|
| 33 |
+
"loss": 0.0233,
|
| 34 |
+
"reward": 0.1356026828289032,
|
| 35 |
+
"reward_std": 0.3039904832839966,
|
| 36 |
+
"rewards/format_reward": 0.026785715948790312,
|
| 37 |
+
"rewards/math_reward": 0.006696428870782256,
|
| 38 |
+
"rewards/tag_count_reward": 0.102120541036129,
|
| 39 |
+
"step": 2
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"clip_ratio": 0.0,
|
| 43 |
+
"completion_length": 696.294677734375,
|
| 44 |
+
"epoch": 0.05240174672489083,
|
| 45 |
+
"grad_norm": 0.38345867395401,
|
| 46 |
+
"kl": 0.0002143383026123047,
|
| 47 |
+
"learning_rate": 1e-05,
|
| 48 |
+
"loss": 0.006,
|
| 49 |
+
"reward": 0.12109375558793545,
|
| 50 |
+
"reward_std": 0.29076967388391495,
|
| 51 |
+
"rewards/format_reward": 0.020089286379516125,
|
| 52 |
+
"rewards/math_reward": 0.0,
|
| 53 |
+
"rewards/tag_count_reward": 0.10100446827709675,
|
| 54 |
+
"step": 3
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"clip_ratio": 0.0,
|
| 58 |
+
"completion_length": 695.0536041259766,
|
| 59 |
+
"epoch": 0.06986899563318777,
|
| 60 |
+
"grad_norm": 233.4116973876953,
|
| 61 |
+
"kl": 5.1015625,
|
| 62 |
+
"learning_rate": 1.3333333333333333e-05,
|
| 63 |
+
"loss": 0.2516,
|
| 64 |
+
"reward": 0.5195312798023224,
|
| 65 |
+
"reward_std": 0.6752565056085587,
|
| 66 |
+
"rewards/format_reward": 0.1562500074505806,
|
| 67 |
+
"rewards/math_reward": 0.006696428870782256,
|
| 68 |
+
"rewards/tag_count_reward": 0.3565848395228386,
|
| 69 |
+
"step": 4
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"clip_ratio": 0.0,
|
| 73 |
+
"completion_length": 646.7767944335938,
|
| 74 |
+
"epoch": 0.08733624454148471,
|
| 75 |
+
"grad_norm": 0.9780572652816772,
|
| 76 |
+
"kl": 0.06536865234375,
|
| 77 |
+
"learning_rate": 1.6666666666666667e-05,
|
| 78 |
+
"loss": 0.0314,
|
| 79 |
+
"reward": 0.8236607611179352,
|
| 80 |
+
"reward_std": 0.7809945940971375,
|
| 81 |
+
"rewards/format_reward": 0.2745535895228386,
|
| 82 |
+
"rewards/math_reward": 0.015625000465661287,
|
| 83 |
+
"rewards/tag_count_reward": 0.5334821715950966,
|
| 84 |
+
"step": 5
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"clip_ratio": 0.0,
|
| 88 |
+
"completion_length": 571.350471496582,
|
| 89 |
+
"epoch": 0.10480349344978165,
|
| 90 |
+
"grad_norm": 0.4738185703754425,
|
| 91 |
+
"kl": 0.049560546875,
|
| 92 |
+
"learning_rate": 2e-05,
|
| 93 |
+
"loss": 0.0527,
|
| 94 |
+
"reward": 1.0898438096046448,
|
| 95 |
+
"reward_std": 0.8620707839727402,
|
| 96 |
+
"rewards/format_reward": 0.3861607313156128,
|
| 97 |
+
"rewards/math_reward": 0.046875001629814506,
|
| 98 |
+
"rewards/tag_count_reward": 0.6568080633878708,
|
| 99 |
+
"step": 6
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"clip_ratio": 0.0,
|
| 103 |
+
"completion_length": 601.9687805175781,
|
| 104 |
+
"epoch": 0.1222707423580786,
|
| 105 |
+
"grad_norm": 0.3717152178287506,
|
| 106 |
+
"kl": 0.03515625,
|
| 107 |
+
"learning_rate": 1.9981033287370443e-05,
|
| 108 |
+
"loss": 0.017,
|
| 109 |
+
"reward": 1.471540242433548,
|
| 110 |
+
"reward_std": 0.7875999510288239,
|
| 111 |
+
"rewards/format_reward": 0.5691964626312256,
|
| 112 |
+
"rewards/math_reward": 0.10937500488944352,
|
| 113 |
+
"rewards/tag_count_reward": 0.7929687798023224,
|
| 114 |
+
"step": 7
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"clip_ratio": 0.0,
|
| 118 |
+
"completion_length": 595.5692291259766,
|
| 119 |
+
"epoch": 0.13973799126637554,
|
| 120 |
+
"grad_norm": 3.1607468128204346,
|
| 121 |
+
"kl": 0.07501220703125,
|
| 122 |
+
"learning_rate": 1.992420509671936e-05,
|
| 123 |
+
"loss": -0.015,
|
| 124 |
+
"reward": 1.5334822237491608,
|
| 125 |
+
"reward_std": 0.7315323203802109,
|
| 126 |
+
"rewards/format_reward": 0.6294643133878708,
|
| 127 |
+
"rewards/math_reward": 0.06473214784637094,
|
| 128 |
+
"rewards/tag_count_reward": 0.839285746216774,
|
| 129 |
+
"step": 8
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"clip_ratio": 0.0,
|
| 133 |
+
"completion_length": 630.6585083007812,
|
| 134 |
+
"epoch": 0.1572052401746725,
|
| 135 |
+
"grad_norm": 0.46272966265678406,
|
| 136 |
+
"kl": 0.09783935546875,
|
| 137 |
+
"learning_rate": 1.982973099683902e-05,
|
| 138 |
+
"loss": 0.0339,
|
| 139 |
+
"reward": 1.6484375596046448,
|
| 140 |
+
"reward_std": 0.6518489569425583,
|
| 141 |
+
"rewards/format_reward": 0.729910746216774,
|
| 142 |
+
"rewards/math_reward": 0.04241071455180645,
|
| 143 |
+
"rewards/tag_count_reward": 0.8761161118745804,
|
| 144 |
+
"step": 9
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"clip_ratio": 0.0,
|
| 148 |
+
"completion_length": 587.7053909301758,
|
| 149 |
+
"epoch": 0.17467248908296942,
|
| 150 |
+
"grad_norm": 3.4270660877227783,
|
| 151 |
+
"kl": 0.3292236328125,
|
| 152 |
+
"learning_rate": 1.9697969360350098e-05,
|
| 153 |
+
"loss": 0.0064,
|
| 154 |
+
"reward": 1.860491156578064,
|
| 155 |
+
"reward_std": 0.6494010090827942,
|
| 156 |
+
"rewards/format_reward": 0.7633928954601288,
|
| 157 |
+
"rewards/math_reward": 0.196428582072258,
|
| 158 |
+
"rewards/tag_count_reward": 0.9006696790456772,
|
| 159 |
+
"step": 10
|
| 160 |
+
},
|
| 161 |
+
{
|
| 162 |
+
"clip_ratio": 0.0,
|
| 163 |
+
"completion_length": 625.0759124755859,
|
| 164 |
+
"epoch": 0.19213973799126638,
|
| 165 |
+
"grad_norm": 3.397077798843384,
|
| 166 |
+
"kl": 0.296875,
|
| 167 |
+
"learning_rate": 1.9529420004271568e-05,
|
| 168 |
+
"loss": 0.1006,
|
| 169 |
+
"reward": 1.2667411267757416,
|
| 170 |
+
"reward_std": 0.8302740603685379,
|
| 171 |
+
"rewards/format_reward": 0.4977678805589676,
|
| 172 |
+
"rewards/math_reward": 0.0535714291036129,
|
| 173 |
+
"rewards/tag_count_reward": 0.7154018133878708,
|
| 174 |
+
"step": 11
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"clip_ratio": 0.0,
|
| 178 |
+
"completion_length": 660.763427734375,
|
| 179 |
+
"epoch": 0.2096069868995633,
|
| 180 |
+
"grad_norm": 1.088446855545044,
|
| 181 |
+
"kl": 0.34130859375,
|
| 182 |
+
"learning_rate": 1.932472229404356e-05,
|
| 183 |
+
"loss": 0.1301,
|
| 184 |
+
"reward": 0.9073661118745804,
|
| 185 |
+
"reward_std": 0.8017744421958923,
|
| 186 |
+
"rewards/format_reward": 0.305803582072258,
|
| 187 |
+
"rewards/math_reward": 0.02455357206054032,
|
| 188 |
+
"rewards/tag_count_reward": 0.5770089626312256,
|
| 189 |
+
"step": 12
|
| 190 |
+
},
|
| 191 |
+
{
|
| 192 |
+
"clip_ratio": 0.0,
|
| 193 |
+
"completion_length": 697.044677734375,
|
| 194 |
+
"epoch": 0.22707423580786026,
|
| 195 |
+
"grad_norm": 35.29081344604492,
|
| 196 |
+
"kl": 3.5400390625,
|
| 197 |
+
"learning_rate": 1.9084652718195237e-05,
|
| 198 |
+
"loss": 0.0956,
|
| 199 |
+
"reward": 0.8772321939468384,
|
| 200 |
+
"reward_std": 0.8175660222768784,
|
| 201 |
+
"rewards/format_reward": 0.294642873108387,
|
| 202 |
+
"rewards/math_reward": 0.02901785750873387,
|
| 203 |
+
"rewards/tag_count_reward": 0.5535714700818062,
|
| 204 |
+
"step": 13
|
| 205 |
+
},
|
| 206 |
+
{
|
| 207 |
+
"clip_ratio": 0.0,
|
| 208 |
+
"completion_length": 714.5536041259766,
|
| 209 |
+
"epoch": 0.2445414847161572,
|
| 210 |
+
"grad_norm": 6.326420307159424,
|
| 211 |
+
"kl": 1.01171875,
|
| 212 |
+
"learning_rate": 1.8810121942857848e-05,
|
| 213 |
+
"loss": 0.1219,
|
| 214 |
+
"reward": 0.6640625298023224,
|
| 215 |
+
"reward_std": 0.7399376779794693,
|
| 216 |
+
"rewards/format_reward": 0.1785714402794838,
|
| 217 |
+
"rewards/math_reward": 0.04017857299186289,
|
| 218 |
+
"rewards/tag_count_reward": 0.4453125298023224,
|
| 219 |
+
"step": 14
|
| 220 |
+
},
|
| 221 |
+
{
|
| 222 |
+
"clip_ratio": 0.0,
|
| 223 |
+
"completion_length": 732.5803985595703,
|
| 224 |
+
"epoch": 0.26200873362445415,
|
| 225 |
+
"grad_norm": 67.30467224121094,
|
| 226 |
+
"kl": 2.271484375,
|
| 227 |
+
"learning_rate": 1.8502171357296144e-05,
|
| 228 |
+
"loss": 0.1286,
|
| 229 |
+
"reward": 0.6646205633878708,
|
| 230 |
+
"reward_std": 0.7443441152572632,
|
| 231 |
+
"rewards/format_reward": 0.2053571529686451,
|
| 232 |
+
"rewards/math_reward": 0.013392857741564512,
|
| 233 |
+
"rewards/tag_count_reward": 0.4458705559372902,
|
| 234 |
+
"step": 15
|
| 235 |
+
},
|
| 236 |
+
{
|
| 237 |
+
"clip_ratio": 0.0,
|
| 238 |
+
"completion_length": 720.5268096923828,
|
| 239 |
+
"epoch": 0.2794759825327511,
|
| 240 |
+
"grad_norm": 5.1966753005981445,
|
| 241 |
+
"kl": 1.416015625,
|
| 242 |
+
"learning_rate": 1.816196912356222e-05,
|
| 243 |
+
"loss": 0.0392,
|
| 244 |
+
"reward": 0.683035746216774,
|
| 245 |
+
"reward_std": 0.7767317444086075,
|
| 246 |
+
"rewards/format_reward": 0.232142873108387,
|
| 247 |
+
"rewards/math_reward": 0.02455357275903225,
|
| 248 |
+
"rewards/tag_count_reward": 0.4263393059372902,
|
| 249 |
+
"step": 16
|
| 250 |
+
},
|
| 251 |
+
{
|
| 252 |
+
"clip_ratio": 0.0,
|
| 253 |
+
"completion_length": 660.1183471679688,
|
| 254 |
+
"epoch": 0.29694323144104806,
|
| 255 |
+
"grad_norm": 1619.3175048828125,
|
| 256 |
+
"kl": 8.3203125,
|
| 257 |
+
"learning_rate": 1.7790805745256703e-05,
|
| 258 |
+
"loss": 0.5223,
|
| 259 |
+
"reward": 0.4051339477300644,
|
| 260 |
+
"reward_std": 0.5999295711517334,
|
| 261 |
+
"rewards/format_reward": 0.11607143096625805,
|
| 262 |
+
"rewards/math_reward": 0.01562500116415322,
|
| 263 |
+
"rewards/tag_count_reward": 0.2734375074505806,
|
| 264 |
+
"step": 17
|
| 265 |
+
},
|
| 266 |
+
{
|
| 267 |
+
"clip_ratio": 0.0,
|
| 268 |
+
"completion_length": 555.8906631469727,
|
| 269 |
+
"epoch": 0.314410480349345,
|
| 270 |
+
"grad_norm": 5.135190486907959,
|
| 271 |
+
"kl": 4.6875,
|
| 272 |
+
"learning_rate": 1.7390089172206594e-05,
|
| 273 |
+
"loss": -0.1531,
|
| 274 |
+
"reward": 0.2232142984867096,
|
| 275 |
+
"reward_std": 0.39768724143505096,
|
| 276 |
+
"rewards/format_reward": 0.05580357275903225,
|
| 277 |
+
"rewards/math_reward": 0.0,
|
| 278 |
+
"rewards/tag_count_reward": 0.1674107201397419,
|
| 279 |
+
"step": 18
|
| 280 |
+
},
|
| 281 |
+
{
|
| 282 |
+
"clip_ratio": 0.0,
|
| 283 |
+
"completion_length": 390.2209930419922,
|
| 284 |
+
"epoch": 0.3318777292576419,
|
| 285 |
+
"grad_norm": 1.996619462966919,
|
| 286 |
+
"kl": 2.265625,
|
| 287 |
+
"learning_rate": 1.696133945962927e-05,
|
| 288 |
+
"loss": -0.3963,
|
| 289 |
+
"reward": 0.12667411379516125,
|
| 290 |
+
"reward_std": 0.2982850968837738,
|
| 291 |
+
"rewards/format_reward": 0.03348214388824999,
|
| 292 |
+
"rewards/math_reward": 0.0,
|
| 293 |
+
"rewards/tag_count_reward": 0.0931919664144516,
|
| 294 |
+
"step": 19
|
| 295 |
+
},
|
| 296 |
+
{
|
| 297 |
+
"clip_ratio": 0.0,
|
| 298 |
+
"completion_length": 278.82814025878906,
|
| 299 |
+
"epoch": 0.34934497816593885,
|
| 300 |
+
"grad_norm": 1.6652499437332153,
|
| 301 |
+
"kl": 2.1640625,
|
| 302 |
+
"learning_rate": 1.650618300204242e-05,
|
| 303 |
+
"loss": -0.7212,
|
| 304 |
+
"reward": 0.1657366119325161,
|
| 305 |
+
"reward_std": 0.3662828281521797,
|
| 306 |
+
"rewards/format_reward": 0.040178573690354824,
|
| 307 |
+
"rewards/math_reward": 0.008928572060540318,
|
| 308 |
+
"rewards/tag_count_reward": 0.1166294701397419,
|
| 309 |
+
"step": 20
|
| 310 |
+
},
|
| 311 |
+
{
|
| 312 |
+
"clip_ratio": 0.0,
|
| 313 |
+
"completion_length": 375.1116256713867,
|
| 314 |
+
"epoch": 0.36681222707423583,
|
| 315 |
+
"grad_norm": 1.9009864330291748,
|
| 316 |
+
"kl": 1.830078125,
|
| 317 |
+
"learning_rate": 1.6026346363792565e-05,
|
| 318 |
+
"loss": -0.5307,
|
| 319 |
+
"reward": 0.2220982201397419,
|
| 320 |
+
"reward_std": 0.43379347771406174,
|
| 321 |
+
"rewards/format_reward": 0.040178573690354824,
|
| 322 |
+
"rewards/math_reward": 0.004464285913854837,
|
| 323 |
+
"rewards/tag_count_reward": 0.1774553656578064,
|
| 324 |
+
"step": 21
|
| 325 |
+
},
|
| 326 |
+
{
|
| 327 |
+
"clip_ratio": 0.0,
|
| 328 |
+
"completion_length": 447.9419860839844,
|
| 329 |
+
"epoch": 0.38427947598253276,
|
| 330 |
+
"grad_norm": 1.1156517267227173,
|
| 331 |
+
"kl": 1.85546875,
|
| 332 |
+
"learning_rate": 1.552364972960506e-05,
|
| 333 |
+
"loss": -0.4126,
|
| 334 |
+
"reward": 0.2204241156578064,
|
| 335 |
+
"reward_std": 0.377070814371109,
|
| 336 |
+
"rewards/format_reward": 0.03125000209547579,
|
| 337 |
+
"rewards/math_reward": 0.0,
|
| 338 |
+
"rewards/tag_count_reward": 0.1891741119325161,
|
| 339 |
+
"step": 22
|
| 340 |
+
},
|
| 341 |
+
{
|
| 342 |
+
"clip_ratio": 0.0,
|
| 343 |
+
"completion_length": 526.6406478881836,
|
| 344 |
+
"epoch": 0.4017467248908297,
|
| 345 |
+
"grad_norm": 2.6996726989746094,
|
| 346 |
+
"kl": 2.9921875,
|
| 347 |
+
"learning_rate": 1.5000000000000002e-05,
|
| 348 |
+
"loss": -0.3377,
|
| 349 |
+
"reward": 0.176339291036129,
|
| 350 |
+
"reward_std": 0.2370295263826847,
|
| 351 |
+
"rewards/format_reward": 0.004464285913854837,
|
| 352 |
+
"rewards/math_reward": 0.0,
|
| 353 |
+
"rewards/tag_count_reward": 0.1718750074505806,
|
| 354 |
+
"step": 23
|
| 355 |
+
},
|
| 356 |
+
{
|
| 357 |
+
"clip_ratio": 0.0,
|
| 358 |
+
"completion_length": 651.6183319091797,
|
| 359 |
+
"epoch": 0.4192139737991266,
|
| 360 |
+
"grad_norm": 1.0993598699569702,
|
| 361 |
+
"kl": 2.52734375,
|
| 362 |
+
"learning_rate": 1.4457383557765385e-05,
|
| 363 |
+
"loss": -0.3695,
|
| 364 |
+
"reward": 0.1322544738650322,
|
| 365 |
+
"reward_std": 0.14882266148924828,
|
| 366 |
+
"rewards/format_reward": 0.0,
|
| 367 |
+
"rewards/math_reward": 0.0,
|
| 368 |
+
"rewards/tag_count_reward": 0.1322544738650322,
|
| 369 |
+
"step": 24
|
| 370 |
+
},
|
| 371 |
+
{
|
| 372 |
+
"clip_ratio": 0.0,
|
| 373 |
+
"completion_length": 736.7522735595703,
|
| 374 |
+
"epoch": 0.4366812227074236,
|
| 375 |
+
"grad_norm": 179.8589324951172,
|
| 376 |
+
"kl": 3.986328125,
|
| 377 |
+
"learning_rate": 1.3897858732926794e-05,
|
| 378 |
+
"loss": -0.2141,
|
| 379 |
+
"reward": 0.1523437574505806,
|
| 380 |
+
"reward_std": 0.1332450620830059,
|
| 381 |
+
"rewards/format_reward": 0.0,
|
| 382 |
+
"rewards/math_reward": 0.0,
|
| 383 |
+
"rewards/tag_count_reward": 0.1523437574505806,
|
| 384 |
+
"step": 25
|
| 385 |
+
},
|
| 386 |
+
{
|
| 387 |
+
"clip_ratio": 0.0,
|
| 388 |
+
"completion_length": 678.4129943847656,
|
| 389 |
+
"epoch": 0.45414847161572053,
|
| 390 |
+
"grad_norm": 2.871249198913574,
|
| 391 |
+
"kl": 3.12109375,
|
| 392 |
+
"learning_rate": 1.3323547994796597e-05,
|
| 393 |
+
"loss": -0.2502,
|
| 394 |
+
"reward": 0.1439732238650322,
|
| 395 |
+
"reward_std": 0.1400768794119358,
|
| 396 |
+
"rewards/format_reward": 0.0,
|
| 397 |
+
"rewards/math_reward": 0.0,
|
| 398 |
+
"rewards/tag_count_reward": 0.1439732238650322,
|
| 399 |
+
"step": 26
|
| 400 |
+
},
|
| 401 |
+
{
|
| 402 |
+
"clip_ratio": 0.0,
|
| 403 |
+
"completion_length": 615.0134124755859,
|
| 404 |
+
"epoch": 0.47161572052401746,
|
| 405 |
+
"grad_norm": 3.0205397605895996,
|
| 406 |
+
"kl": 1.2001953125,
|
| 407 |
+
"learning_rate": 1.2736629900720832e-05,
|
| 408 |
+
"loss": -0.2349,
|
| 409 |
+
"reward": 0.1171875037252903,
|
| 410 |
+
"reward_std": 0.12809346616268158,
|
| 411 |
+
"rewards/format_reward": 0.0,
|
| 412 |
+
"rewards/math_reward": 0.0,
|
| 413 |
+
"rewards/tag_count_reward": 0.1171875037252903,
|
| 414 |
+
"step": 27
|
| 415 |
+
},
|
| 416 |
+
{
|
| 417 |
+
"clip_ratio": 0.0,
|
| 418 |
+
"completion_length": 594.3660888671875,
|
| 419 |
+
"epoch": 0.4890829694323144,
|
| 420 |
+
"grad_norm": 14.793048858642578,
|
| 421 |
+
"kl": 1.2587890625,
|
| 422 |
+
"learning_rate": 1.2139330832064975e-05,
|
| 423 |
+
"loss": -0.2831,
|
| 424 |
+
"reward": 0.12276786006987095,
|
| 425 |
+
"reward_std": 0.1267802119255066,
|
| 426 |
+
"rewards/format_reward": 0.0,
|
| 427 |
+
"rewards/math_reward": 0.0,
|
| 428 |
+
"rewards/tag_count_reward": 0.12276786006987095,
|
| 429 |
+
"step": 28
|
| 430 |
+
},
|
| 431 |
+
{
|
| 432 |
+
"clip_ratio": 0.0,
|
| 433 |
+
"completion_length": 551.0401992797852,
|
| 434 |
+
"epoch": 0.5065502183406113,
|
| 435 |
+
"grad_norm": 1.9193952083587646,
|
| 436 |
+
"kl": 1.31640625,
|
| 437 |
+
"learning_rate": 1.1533916548786856e-05,
|
| 438 |
+
"loss": -0.2434,
|
| 439 |
+
"reward": 0.11607143469154835,
|
| 440 |
+
"reward_std": 0.1321311704814434,
|
| 441 |
+
"rewards/format_reward": 0.0,
|
| 442 |
+
"rewards/math_reward": 0.0,
|
| 443 |
+
"rewards/tag_count_reward": 0.11607143469154835,
|
| 444 |
+
"step": 29
|
| 445 |
+
},
|
| 446 |
+
{
|
| 447 |
+
"clip_ratio": 0.0,
|
| 448 |
+
"completion_length": 569.2522583007812,
|
| 449 |
+
"epoch": 0.5240174672489083,
|
| 450 |
+
"grad_norm": 1.2092971801757812,
|
| 451 |
+
"kl": 1.2158203125,
|
| 452 |
+
"learning_rate": 1.092268359463302e-05,
|
| 453 |
+
"loss": -0.3131,
|
| 454 |
+
"reward": 0.1160714328289032,
|
| 455 |
+
"reward_std": 0.13385851308703423,
|
| 456 |
+
"rewards/format_reward": 0.0,
|
| 457 |
+
"rewards/math_reward": 0.0,
|
| 458 |
+
"rewards/tag_count_reward": 0.1160714328289032,
|
| 459 |
+
"step": 30
|
| 460 |
+
},
|
| 461 |
+
{
|
| 462 |
+
"clip_ratio": 0.0,
|
| 463 |
+
"completion_length": 609.825927734375,
|
| 464 |
+
"epoch": 0.5414847161572053,
|
| 465 |
+
"grad_norm": 2.0326733589172363,
|
| 466 |
+
"kl": 2.9453125,
|
| 467 |
+
"learning_rate": 1.0307950585561705e-05,
|
| 468 |
+
"loss": -0.2521,
|
| 469 |
+
"reward": 0.1132812574505806,
|
| 470 |
+
"reward_std": 0.1328713335096836,
|
| 471 |
+
"rewards/format_reward": 0.0,
|
| 472 |
+
"rewards/math_reward": 0.0,
|
| 473 |
+
"rewards/tag_count_reward": 0.1132812574505806,
|
| 474 |
+
"step": 31
|
| 475 |
+
},
|
| 476 |
+
{
|
| 477 |
+
"clip_ratio": 0.0,
|
| 478 |
+
"completion_length": 595.2344055175781,
|
| 479 |
+
"epoch": 0.5589519650655022,
|
| 480 |
+
"grad_norm": 0.9160346984863281,
|
| 481 |
+
"kl": 1.41796875,
|
| 482 |
+
"learning_rate": 9.692049414438298e-06,
|
| 483 |
+
"loss": -0.197,
|
| 484 |
+
"reward": 0.11328125558793545,
|
| 485 |
+
"reward_std": 0.13070819526910782,
|
| 486 |
+
"rewards/format_reward": 0.0,
|
| 487 |
+
"rewards/math_reward": 0.0,
|
| 488 |
+
"rewards/tag_count_reward": 0.11328125558793545,
|
| 489 |
+
"step": 32
|
| 490 |
+
},
|
| 491 |
+
{
|
| 492 |
+
"clip_ratio": 0.0,
|
| 493 |
+
"completion_length": 561.7076110839844,
|
| 494 |
+
"epoch": 0.5764192139737991,
|
| 495 |
+
"grad_norm": 20.22221565246582,
|
| 496 |
+
"kl": 2.130859375,
|
| 497 |
+
"learning_rate": 9.07731640536698e-06,
|
| 498 |
+
"loss": -0.1843,
|
| 499 |
+
"reward": 0.1188616156578064,
|
| 500 |
+
"reward_std": 0.13797599077224731,
|
| 501 |
+
"rewards/format_reward": 0.0,
|
| 502 |
+
"rewards/math_reward": 0.0,
|
| 503 |
+
"rewards/tag_count_reward": 0.1188616156578064,
|
| 504 |
+
"step": 33
|
| 505 |
+
},
|
| 506 |
+
{
|
| 507 |
+
"clip_ratio": 0.0,
|
| 508 |
+
"completion_length": 517.8482360839844,
|
| 509 |
+
"epoch": 0.5938864628820961,
|
| 510 |
+
"grad_norm": 16.17409896850586,
|
| 511 |
+
"kl": 2.591796875,
|
| 512 |
+
"learning_rate": 8.466083451213145e-06,
|
| 513 |
+
"loss": -0.0359,
|
| 514 |
+
"reward": 0.10993303917348385,
|
| 515 |
+
"reward_std": 0.1322279404848814,
|
| 516 |
+
"rewards/format_reward": 0.0,
|
| 517 |
+
"rewards/math_reward": 0.0,
|
| 518 |
+
"rewards/tag_count_reward": 0.10993303917348385,
|
| 519 |
+
"step": 34
|
| 520 |
+
},
|
| 521 |
+
{
|
| 522 |
+
"clip_ratio": 0.0,
|
| 523 |
+
"completion_length": 458.41297149658203,
|
| 524 |
+
"epoch": 0.611353711790393,
|
| 525 |
+
"grad_norm": 1.844921588897705,
|
| 526 |
+
"kl": 1.6328125,
|
| 527 |
+
"learning_rate": 7.860669167935028e-06,
|
| 528 |
+
"loss": -0.0605,
|
| 529 |
+
"reward": 0.10825893469154835,
|
| 530 |
+
"reward_std": 0.13310172408819199,
|
| 531 |
+
"rewards/format_reward": 0.0,
|
| 532 |
+
"rewards/math_reward": 0.0,
|
| 533 |
+
"rewards/tag_count_reward": 0.10825893469154835,
|
| 534 |
+
"step": 35
|
| 535 |
+
},
|
| 536 |
+
{
|
| 537 |
+
"clip_ratio": 0.0,
|
| 538 |
+
"completion_length": 408.12947845458984,
|
| 539 |
+
"epoch": 0.62882096069869,
|
| 540 |
+
"grad_norm": 3.107694149017334,
|
| 541 |
+
"kl": 1.501953125,
|
| 542 |
+
"learning_rate": 7.263370099279173e-06,
|
| 543 |
+
"loss": -0.1977,
|
| 544 |
+
"reward": 0.12667411379516125,
|
| 545 |
+
"reward_std": 0.14315131679177284,
|
| 546 |
+
"rewards/format_reward": 0.0,
|
| 547 |
+
"rewards/math_reward": 0.0,
|
| 548 |
+
"rewards/tag_count_reward": 0.12667411379516125,
|
| 549 |
+
"step": 36
|
| 550 |
+
},
|
| 551 |
+
{
|
| 552 |
+
"clip_ratio": 0.0,
|
| 553 |
+
"completion_length": 397.15403747558594,
|
| 554 |
+
"epoch": 0.6462882096069869,
|
| 555 |
+
"grad_norm": 3.9368510246276855,
|
| 556 |
+
"kl": 2.07421875,
|
| 557 |
+
"learning_rate": 6.6764520052034054e-06,
|
| 558 |
+
"loss": -0.1921,
|
| 559 |
+
"reward": 0.13950893469154835,
|
| 560 |
+
"reward_std": 0.15414584055542946,
|
| 561 |
+
"rewards/format_reward": 0.0,
|
| 562 |
+
"rewards/math_reward": 0.0,
|
| 563 |
+
"rewards/tag_count_reward": 0.13950893469154835,
|
| 564 |
+
"step": 37
|
| 565 |
+
},
|
| 566 |
+
{
|
| 567 |
+
"clip_ratio": 0.0,
|
| 568 |
+
"completion_length": 419.35269927978516,
|
| 569 |
+
"epoch": 0.6637554585152838,
|
| 570 |
+
"grad_norm": 1.3178058862686157,
|
| 571 |
+
"kl": 1.17578125,
|
| 572 |
+
"learning_rate": 6.102141267073207e-06,
|
| 573 |
+
"loss": -0.178,
|
| 574 |
+
"reward": 0.141183041036129,
|
| 575 |
+
"reward_std": 0.16641410812735558,
|
| 576 |
+
"rewards/format_reward": 0.0,
|
| 577 |
+
"rewards/math_reward": 0.0,
|
| 578 |
+
"rewards/tag_count_reward": 0.141183041036129,
|
| 579 |
+
"step": 38
|
| 580 |
+
},
|
| 581 |
+
{
|
| 582 |
+
"clip_ratio": 0.0,
|
| 583 |
+
"completion_length": 438.5692138671875,
|
| 584 |
+
"epoch": 0.6812227074235808,
|
| 585 |
+
"grad_norm": 2.4538495540618896,
|
| 586 |
+
"kl": 0.7109375,
|
| 587 |
+
"learning_rate": 5.542616442234618e-06,
|
| 588 |
+
"loss": -0.0449,
|
| 589 |
+
"reward": 0.1612723283469677,
|
| 590 |
+
"reward_std": 0.15356499329209328,
|
| 591 |
+
"rewards/format_reward": 0.0,
|
| 592 |
+
"rewards/math_reward": 0.0,
|
| 593 |
+
"rewards/tag_count_reward": 0.1612723283469677,
|
| 594 |
+
"step": 39
|
| 595 |
+
},
|
| 596 |
+
{
|
| 597 |
+
"clip_ratio": 0.0,
|
| 598 |
+
"completion_length": 460.3995819091797,
|
| 599 |
+
"epoch": 0.6986899563318777,
|
| 600 |
+
"grad_norm": 0.8929517865180969,
|
| 601 |
+
"kl": 0.7744140625,
|
| 602 |
+
"learning_rate": 5.000000000000003e-06,
|
| 603 |
+
"loss": -0.2372,
|
| 604 |
+
"reward": 0.1568080447614193,
|
| 605 |
+
"reward_std": 0.17059754580259323,
|
| 606 |
+
"rewards/format_reward": 0.0,
|
| 607 |
+
"rewards/math_reward": 0.0,
|
| 608 |
+
"rewards/tag_count_reward": 0.1568080447614193,
|
| 609 |
+
"step": 40
|
| 610 |
+
},
|
| 611 |
+
{
|
| 612 |
+
"clip_ratio": 0.0,
|
| 613 |
+
"completion_length": 496.6250305175781,
|
| 614 |
+
"epoch": 0.7161572052401747,
|
| 615 |
+
"grad_norm": 1.5635244846343994,
|
| 616 |
+
"kl": 0.75,
|
| 617 |
+
"learning_rate": 4.476350270394942e-06,
|
| 618 |
+
"loss": -0.1742,
|
| 619 |
+
"reward": 0.1674107201397419,
|
| 620 |
+
"reward_std": 0.1668030135333538,
|
| 621 |
+
"rewards/format_reward": 0.0,
|
| 622 |
+
"rewards/math_reward": 0.0,
|
| 623 |
+
"rewards/tag_count_reward": 0.1674107201397419,
|
| 624 |
+
"step": 41
|
| 625 |
+
},
|
| 626 |
+
{
|
| 627 |
+
"clip_ratio": 0.0,
|
| 628 |
+
"completion_length": 544.7656707763672,
|
| 629 |
+
"epoch": 0.7336244541484717,
|
| 630 |
+
"grad_norm": 1.9236276149749756,
|
| 631 |
+
"kl": 0.7568359375,
|
| 632 |
+
"learning_rate": 3.973653636207437e-06,
|
| 633 |
+
"loss": -0.1978,
|
| 634 |
+
"reward": 0.1841517947614193,
|
| 635 |
+
"reward_std": 0.17343608289957047,
|
| 636 |
+
"rewards/format_reward": 0.0,
|
| 637 |
+
"rewards/math_reward": 0.0,
|
| 638 |
+
"rewards/tag_count_reward": 0.1841517947614193,
|
| 639 |
+
"step": 42
|
| 640 |
+
},
|
| 641 |
+
{
|
| 642 |
+
"clip_ratio": 0.0,
|
| 643 |
+
"completion_length": 542.763427734375,
|
| 644 |
+
"epoch": 0.7510917030567685,
|
| 645 |
+
"grad_norm": 1.9667587280273438,
|
| 646 |
+
"kl": 1.2421875,
|
| 647 |
+
"learning_rate": 3.493816997957582e-06,
|
| 648 |
+
"loss": -0.1336,
|
| 649 |
+
"reward": 0.1830357238650322,
|
| 650 |
+
"reward_std": 0.17494625598192215,
|
| 651 |
+
"rewards/format_reward": 0.0,
|
| 652 |
+
"rewards/math_reward": 0.0,
|
| 653 |
+
"rewards/tag_count_reward": 0.1830357238650322,
|
| 654 |
+
"step": 43
|
| 655 |
+
},
|
| 656 |
+
{
|
| 657 |
+
"clip_ratio": 0.0,
|
| 658 |
+
"completion_length": 575.8772735595703,
|
| 659 |
+
"epoch": 0.7685589519650655,
|
| 660 |
+
"grad_norm": 1.682228922843933,
|
| 661 |
+
"kl": 1.52734375,
|
| 662 |
+
"learning_rate": 3.0386605403707347e-06,
|
| 663 |
+
"loss": -0.0949,
|
| 664 |
+
"reward": 0.1908482238650322,
|
| 665 |
+
"reward_std": 0.17441384866833687,
|
| 666 |
+
"rewards/format_reward": 0.0,
|
| 667 |
+
"rewards/math_reward": 0.0,
|
| 668 |
+
"rewards/tag_count_reward": 0.1908482238650322,
|
| 669 |
+
"step": 44
|
| 670 |
+
},
|
| 671 |
+
{
|
| 672 |
+
"clip_ratio": 0.0,
|
| 673 |
+
"completion_length": 602.6518096923828,
|
| 674 |
+
"epoch": 0.7860262008733624,
|
| 675 |
+
"grad_norm": 1.8132331371307373,
|
| 676 |
+
"kl": 1.7109375,
|
| 677 |
+
"learning_rate": 2.6099108277934105e-06,
|
| 678 |
+
"loss": -0.095,
|
| 679 |
+
"reward": 0.1958705447614193,
|
| 680 |
+
"reward_std": 0.17568127438426018,
|
| 681 |
+
"rewards/format_reward": 0.0,
|
| 682 |
+
"rewards/math_reward": 0.0,
|
| 683 |
+
"rewards/tag_count_reward": 0.1958705447614193,
|
| 684 |
+
"step": 45
|
| 685 |
+
},
|
| 686 |
+
{
|
| 687 |
+
"clip_ratio": 0.0,
|
| 688 |
+
"completion_length": 596.0089569091797,
|
| 689 |
+
"epoch": 0.8034934497816594,
|
| 690 |
+
"grad_norm": 1.854785442352295,
|
| 691 |
+
"kl": 1.494140625,
|
| 692 |
+
"learning_rate": 2.209194254743295e-06,
|
| 693 |
+
"loss": -0.118,
|
| 694 |
+
"reward": 0.1791294738650322,
|
| 695 |
+
"reward_std": 0.17129607126116753,
|
| 696 |
+
"rewards/format_reward": 0.0,
|
| 697 |
+
"rewards/math_reward": 0.0,
|
| 698 |
+
"rewards/tag_count_reward": 0.1791294738650322,
|
| 699 |
+
"step": 46
|
| 700 |
+
},
|
| 701 |
+
{
|
| 702 |
+
"clip_ratio": 0.0,
|
| 703 |
+
"completion_length": 649.0737152099609,
|
| 704 |
+
"epoch": 0.8209606986899564,
|
| 705 |
+
"grad_norm": 1.340461015701294,
|
| 706 |
+
"kl": 1.1064453125,
|
| 707 |
+
"learning_rate": 1.8380308764377841e-06,
|
| 708 |
+
"loss": -0.1276,
|
| 709 |
+
"reward": 0.2148437611758709,
|
| 710 |
+
"reward_std": 0.1898629516363144,
|
| 711 |
+
"rewards/format_reward": 0.0,
|
| 712 |
+
"rewards/math_reward": 0.0,
|
| 713 |
+
"rewards/tag_count_reward": 0.2148437611758709,
|
| 714 |
+
"step": 47
|
| 715 |
+
},
|
| 716 |
+
{
|
| 717 |
+
"clip_ratio": 0.0,
|
| 718 |
+
"completion_length": 607.2924346923828,
|
| 719 |
+
"epoch": 0.8384279475982532,
|
| 720 |
+
"grad_norm": 1.8552510738372803,
|
| 721 |
+
"kl": 0.98046875,
|
| 722 |
+
"learning_rate": 1.4978286427038602e-06,
|
| 723 |
+
"loss": -0.1139,
|
| 724 |
+
"reward": 0.2053571529686451,
|
| 725 |
+
"reward_std": 0.193530160933733,
|
| 726 |
+
"rewards/format_reward": 0.0,
|
| 727 |
+
"rewards/math_reward": 0.0,
|
| 728 |
+
"rewards/tag_count_reward": 0.2053571529686451,
|
| 729 |
+
"step": 48
|
| 730 |
+
},
|
| 731 |
+
{
|
| 732 |
+
"clip_ratio": 0.0,
|
| 733 |
+
"completion_length": 606.4910888671875,
|
| 734 |
+
"epoch": 0.8558951965065502,
|
| 735 |
+
"grad_norm": 3.160490036010742,
|
| 736 |
+
"kl": 0.767578125,
|
| 737 |
+
"learning_rate": 1.1898780571421554e-06,
|
| 738 |
+
"loss": -0.0985,
|
| 739 |
+
"reward": 0.1914062574505806,
|
| 740 |
+
"reward_std": 0.18279264867305756,
|
| 741 |
+
"rewards/format_reward": 0.0,
|
| 742 |
+
"rewards/math_reward": 0.0,
|
| 743 |
+
"rewards/tag_count_reward": 0.1914062574505806,
|
| 744 |
+
"step": 49
|
| 745 |
+
},
|
| 746 |
+
{
|
| 747 |
+
"clip_ratio": 0.0,
|
| 748 |
+
"completion_length": 619.294677734375,
|
| 749 |
+
"epoch": 0.8733624454148472,
|
| 750 |
+
"grad_norm": 2.6852455139160156,
|
| 751 |
+
"kl": 0.8779296875,
|
| 752 |
+
"learning_rate": 9.153472818047627e-07,
|
| 753 |
+
"loss": -0.1051,
|
| 754 |
+
"reward": 0.1986607238650322,
|
| 755 |
+
"reward_std": 0.184061449021101,
|
| 756 |
+
"rewards/format_reward": 0.0,
|
| 757 |
+
"rewards/math_reward": 0.0,
|
| 758 |
+
"rewards/tag_count_reward": 0.1986607238650322,
|
| 759 |
+
"step": 50
|
| 760 |
+
},
|
| 761 |
+
{
|
| 762 |
+
"clip_ratio": 0.0,
|
| 763 |
+
"completion_length": 577.8460083007812,
|
| 764 |
+
"epoch": 0.8908296943231441,
|
| 765 |
+
"grad_norm": 0.8228152394294739,
|
| 766 |
+
"kl": 0.8427734375,
|
| 767 |
+
"learning_rate": 6.752777059564431e-07,
|
| 768 |
+
"loss": -0.0832,
|
| 769 |
+
"reward": 0.2159598357975483,
|
| 770 |
+
"reward_std": 0.20024307817220688,
|
| 771 |
+
"rewards/format_reward": 0.0,
|
| 772 |
+
"rewards/math_reward": 0.0,
|
| 773 |
+
"rewards/tag_count_reward": 0.2159598357975483,
|
| 774 |
+
"step": 51
|
| 775 |
+
},
|
| 776 |
+
{
|
| 777 |
+
"clip_ratio": 0.0,
|
| 778 |
+
"completion_length": 581.6049346923828,
|
| 779 |
+
"epoch": 0.9082969432314411,
|
| 780 |
+
"grad_norm": 1.3566585779190063,
|
| 781 |
+
"kl": 0.7158203125,
|
| 782 |
+
"learning_rate": 4.7057999572843516e-07,
|
| 783 |
+
"loss": -0.1733,
|
| 784 |
+
"reward": 0.1975446529686451,
|
| 785 |
+
"reward_std": 0.18785196170210838,
|
| 786 |
+
"rewards/format_reward": 0.0,
|
| 787 |
+
"rewards/math_reward": 0.0,
|
| 788 |
+
"rewards/tag_count_reward": 0.1975446529686451,
|
| 789 |
+
"step": 52
|
| 790 |
+
},
|
| 791 |
+
{
|
| 792 |
+
"clip_ratio": 0.0,
|
| 793 |
+
"completion_length": 609.1339416503906,
|
| 794 |
+
"epoch": 0.925764192139738,
|
| 795 |
+
"grad_norm": 2.234724283218384,
|
| 796 |
+
"kl": 0.6767578125,
|
| 797 |
+
"learning_rate": 3.020306396499062e-07,
|
| 798 |
+
"loss": -0.0538,
|
| 799 |
+
"reward": 0.2193080447614193,
|
| 800 |
+
"reward_std": 0.19074836000800133,
|
| 801 |
+
"rewards/format_reward": 0.0,
|
| 802 |
+
"rewards/math_reward": 0.0,
|
| 803 |
+
"rewards/tag_count_reward": 0.2193080447614193,
|
| 804 |
+
"step": 53
|
| 805 |
+
},
|
| 806 |
+
{
|
| 807 |
+
"clip_ratio": 0.0,
|
| 808 |
+
"completion_length": 596.1317291259766,
|
| 809 |
+
"epoch": 0.9432314410480349,
|
| 810 |
+
"grad_norm": 1.5052021741867065,
|
| 811 |
+
"kl": 0.810546875,
|
| 812 |
+
"learning_rate": 1.7026900316098217e-07,
|
| 813 |
+
"loss": -0.0754,
|
| 814 |
+
"reward": 0.2243303656578064,
|
| 815 |
+
"reward_std": 0.18597954511642456,
|
| 816 |
+
"rewards/format_reward": 0.0,
|
| 817 |
+
"rewards/math_reward": 0.0,
|
| 818 |
+
"rewards/tag_count_reward": 0.2243303656578064,
|
| 819 |
+
"step": 54
|
| 820 |
+
},
|
| 821 |
+
{
|
| 822 |
+
"clip_ratio": 0.0,
|
| 823 |
+
"completion_length": 550.7991333007812,
|
| 824 |
+
"epoch": 0.9606986899563319,
|
| 825 |
+
"grad_norm": 1.7257078886032104,
|
| 826 |
+
"kl": 0.822265625,
|
| 827 |
+
"learning_rate": 7.579490328064265e-08,
|
| 828 |
+
"loss": -0.07,
|
| 829 |
+
"reward": 0.2031250074505806,
|
| 830 |
+
"reward_std": 0.19252333417534828,
|
| 831 |
+
"rewards/format_reward": 0.0,
|
| 832 |
+
"rewards/math_reward": 0.0,
|
| 833 |
+
"rewards/tag_count_reward": 0.2031250074505806,
|
| 834 |
+
"step": 55
|
| 835 |
+
},
|
| 836 |
+
{
|
| 837 |
+
"clip_ratio": 0.0,
|
| 838 |
+
"completion_length": 588.982177734375,
|
| 839 |
+
"epoch": 0.9781659388646288,
|
| 840 |
+
"grad_norm": 1.599360704421997,
|
| 841 |
+
"kl": 0.736328125,
|
| 842 |
+
"learning_rate": 1.896671262955896e-08,
|
| 843 |
+
"loss": -0.0884,
|
| 844 |
+
"reward": 0.188616082072258,
|
| 845 |
+
"reward_std": 0.18718989938497543,
|
| 846 |
+
"rewards/format_reward": 0.0,
|
| 847 |
+
"rewards/math_reward": 0.0,
|
| 848 |
+
"rewards/tag_count_reward": 0.188616082072258,
|
| 849 |
+
"step": 56
|
| 850 |
+
},
|
| 851 |
+
{
|
| 852 |
+
"clip_ratio": 0.0,
|
| 853 |
+
"completion_length": 562.84375,
|
| 854 |
+
"epoch": 0.9956331877729258,
|
| 855 |
+
"grad_norm": 1.2704195976257324,
|
| 856 |
+
"kl": 0.8037109375,
|
| 857 |
+
"learning_rate": 0.0,
|
| 858 |
+
"loss": -0.0482,
|
| 859 |
+
"reward": 0.2237723357975483,
|
| 860 |
+
"reward_std": 0.20235323160886765,
|
| 861 |
+
"rewards/format_reward": 0.0,
|
| 862 |
+
"rewards/math_reward": 0.0,
|
| 863 |
+
"rewards/tag_count_reward": 0.2237723357975483,
|
| 864 |
+
"step": 57
|
| 865 |
+
},
|
| 866 |
+
{
|
| 867 |
+
"epoch": 0.9956331877729258,
|
| 868 |
+
"step": 57,
|
| 869 |
+
"total_flos": 0.0,
|
| 870 |
+
"train_loss": -0.11113634665512986,
|
| 871 |
+
"train_runtime": 2287.8706,
|
| 872 |
+
"train_samples_per_second": 0.699,
|
| 873 |
+
"train_steps_per_second": 0.025
|
| 874 |
+
}
|
| 875 |
+
],
|
| 876 |
+
"logging_steps": 1,
|
| 877 |
+
"max_steps": 57,
|
| 878 |
+
"num_input_tokens_seen": 0,
|
| 879 |
+
"num_train_epochs": 1,
|
| 880 |
+
"save_steps": 500,
|
| 881 |
+
"stateful_callbacks": {
|
| 882 |
+
"TrainerControl": {
|
| 883 |
+
"args": {
|
| 884 |
+
"should_epoch_stop": false,
|
| 885 |
+
"should_evaluate": false,
|
| 886 |
+
"should_log": false,
|
| 887 |
+
"should_save": true,
|
| 888 |
+
"should_training_stop": true
|
| 889 |
+
},
|
| 890 |
+
"attributes": {}
|
| 891 |
+
}
|
| 892 |
+
},
|
| 893 |
+
"total_flos": 0.0,
|
| 894 |
+
"train_batch_size": 16,
|
| 895 |
+
"trial_name": null,
|
| 896 |
+
"trial_params": null
|
| 897 |
+
}
|