Model save
Browse files- README.md +68 -0
- all_results.json +8 -0
- generation_config.json +6 -0
- train_results.json +8 -0
- trainer_state.json +2787 -0
README.md
ADDED
|
@@ -0,0 +1,68 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
base_model: Qwen/Qwen2.5-3B
|
| 3 |
+
library_name: transformers
|
| 4 |
+
model_name: Qwen2.5-3B-Open-R1-GRPO-math-selected-default
|
| 5 |
+
tags:
|
| 6 |
+
- generated_from_trainer
|
| 7 |
+
- trl
|
| 8 |
+
- grpo
|
| 9 |
+
licence: license
|
| 10 |
+
---
|
| 11 |
+
|
| 12 |
+
# Model Card for Qwen2.5-3B-Open-R1-GRPO-math-selected-default
|
| 13 |
+
|
| 14 |
+
This model is a fine-tuned version of [Qwen/Qwen2.5-3B](https://huggingface.co/Qwen/Qwen2.5-3B).
|
| 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="Lansechen/Qwen2.5-3B-Open-R1-GRPO-math-selected-default", 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/chenran1995-the-chinese-university-of-hong-kong/huggingface/runs/wnriict3)
|
| 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
|
| 38 |
+
- Transformers: 4.50.0
|
| 39 |
+
- Pytorch: 2.5.1+cu121
|
| 40 |
+
- Datasets: 3.5.0
|
| 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": 0.02676454978019156,
|
| 4 |
+
"train_runtime": 31339.2496,
|
| 5 |
+
"train_samples": 11040,
|
| 6 |
+
"train_samples_per_second": 0.705,
|
| 7 |
+
"train_steps_per_second": 0.006
|
| 8 |
+
}
|
generation_config.json
ADDED
|
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bos_token_id": 151643,
|
| 3 |
+
"eos_token_id": 151643,
|
| 4 |
+
"max_new_tokens": 2048,
|
| 5 |
+
"transformers_version": "4.50.0"
|
| 6 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"total_flos": 0.0,
|
| 3 |
+
"train_loss": 0.02676454978019156,
|
| 4 |
+
"train_runtime": 31339.2496,
|
| 5 |
+
"train_samples": 11040,
|
| 6 |
+
"train_samples_per_second": 0.705,
|
| 7 |
+
"train_steps_per_second": 0.006
|
| 8 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,2787 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_global_step": null,
|
| 3 |
+
"best_metric": null,
|
| 4 |
+
"best_model_checkpoint": null,
|
| 5 |
+
"epoch": 1.9936628643852978,
|
| 6 |
+
"eval_steps": 100,
|
| 7 |
+
"global_step": 196,
|
| 8 |
+
"is_hyper_param_search": false,
|
| 9 |
+
"is_local_process_zero": true,
|
| 10 |
+
"is_world_process_zero": true,
|
| 11 |
+
"log_history": [
|
| 12 |
+
{
|
| 13 |
+
"clip_ratio": 0.0,
|
| 14 |
+
"completion_length": 544.7745819091797,
|
| 15 |
+
"epoch": 0.010139416983523447,
|
| 16 |
+
"grad_norm": 0.1392974704504013,
|
| 17 |
+
"learning_rate": 5e-08,
|
| 18 |
+
"loss": -0.0292,
|
| 19 |
+
"num_tokens": 636118.0,
|
| 20 |
+
"reward": 0.12946429196745157,
|
| 21 |
+
"reward_std": 0.22425302118062973,
|
| 22 |
+
"rewards/accuracy_reward": 0.1261160708963871,
|
| 23 |
+
"rewards/format_reward": 0.003348214435391128,
|
| 24 |
+
"step": 1
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"clip_ratio": 0.0,
|
| 28 |
+
"completion_length": 549.2812843322754,
|
| 29 |
+
"epoch": 0.020278833967046894,
|
| 30 |
+
"grad_norm": 0.1566777527332306,
|
| 31 |
+
"learning_rate": 1e-07,
|
| 32 |
+
"loss": -0.0074,
|
| 33 |
+
"num_tokens": 1263330.0,
|
| 34 |
+
"reward": 0.152901791036129,
|
| 35 |
+
"reward_std": 0.25637512281537056,
|
| 36 |
+
"rewards/accuracy_reward": 0.14508928637951612,
|
| 37 |
+
"rewards/format_reward": 0.007812500349245965,
|
| 38 |
+
"step": 2
|
| 39 |
+
},
|
| 40 |
+
{
|
| 41 |
+
"clip_ratio": 0.0,
|
| 42 |
+
"completion_length": 585.7969131469727,
|
| 43 |
+
"epoch": 0.030418250950570342,
|
| 44 |
+
"grad_norm": 0.15250442922115326,
|
| 45 |
+
"learning_rate": 1.5e-07,
|
| 46 |
+
"loss": -0.0299,
|
| 47 |
+
"num_tokens": 1928324.0,
|
| 48 |
+
"reward": 0.17745536379516125,
|
| 49 |
+
"reward_std": 0.26186119951307774,
|
| 50 |
+
"rewards/accuracy_reward": 0.16964285913854837,
|
| 51 |
+
"rewards/format_reward": 0.007812500349245965,
|
| 52 |
+
"step": 3
|
| 53 |
+
},
|
| 54 |
+
{
|
| 55 |
+
"clip_ratio": 0.0,
|
| 56 |
+
"completion_length": 550.0346145629883,
|
| 57 |
+
"epoch": 0.04055766793409379,
|
| 58 |
+
"grad_norm": 0.15349256992340088,
|
| 59 |
+
"learning_rate": 2e-07,
|
| 60 |
+
"loss": -0.0336,
|
| 61 |
+
"num_tokens": 2558107.0,
|
| 62 |
+
"reward": 0.17075893748551607,
|
| 63 |
+
"reward_std": 0.2673841342329979,
|
| 64 |
+
"rewards/accuracy_reward": 0.16071428824216127,
|
| 65 |
+
"rewards/format_reward": 0.010044643306173384,
|
| 66 |
+
"step": 4
|
| 67 |
+
},
|
| 68 |
+
{
|
| 69 |
+
"clip_ratio": 0.0,
|
| 70 |
+
"completion_length": 578.6484642028809,
|
| 71 |
+
"epoch": 0.050697084917617236,
|
| 72 |
+
"grad_norm": 0.3040037155151367,
|
| 73 |
+
"learning_rate": 2.5e-07,
|
| 74 |
+
"loss": -0.0205,
|
| 75 |
+
"num_tokens": 3203360.0,
|
| 76 |
+
"reward": 0.15848214831203222,
|
| 77 |
+
"reward_std": 0.2771828528493643,
|
| 78 |
+
"rewards/accuracy_reward": 0.1517857126891613,
|
| 79 |
+
"rewards/format_reward": 0.006696428754366934,
|
| 80 |
+
"step": 5
|
| 81 |
+
},
|
| 82 |
+
{
|
| 83 |
+
"clip_ratio": 0.0,
|
| 84 |
+
"completion_length": 583.7399826049805,
|
| 85 |
+
"epoch": 0.060836501901140684,
|
| 86 |
+
"grad_norm": 0.13601042330265045,
|
| 87 |
+
"learning_rate": 3e-07,
|
| 88 |
+
"loss": -0.031,
|
| 89 |
+
"num_tokens": 3855583.0,
|
| 90 |
+
"reward": 0.15625000465661287,
|
| 91 |
+
"reward_std": 0.25207988917827606,
|
| 92 |
+
"rewards/accuracy_reward": 0.15066964365541935,
|
| 93 |
+
"rewards/format_reward": 0.005580357392318547,
|
| 94 |
+
"step": 6
|
| 95 |
+
},
|
| 96 |
+
{
|
| 97 |
+
"clip_ratio": 0.0,
|
| 98 |
+
"completion_length": 537.3594093322754,
|
| 99 |
+
"epoch": 0.07097591888466413,
|
| 100 |
+
"grad_norm": 0.16291461884975433,
|
| 101 |
+
"learning_rate": 3.5e-07,
|
| 102 |
+
"loss": -0.0137,
|
| 103 |
+
"num_tokens": 4461481.0,
|
| 104 |
+
"reward": 0.17522322200238705,
|
| 105 |
+
"reward_std": 0.2866264618933201,
|
| 106 |
+
"rewards/accuracy_reward": 0.16183035727590322,
|
| 107 |
+
"rewards/format_reward": 0.013392857508733869,
|
| 108 |
+
"step": 7
|
| 109 |
+
},
|
| 110 |
+
{
|
| 111 |
+
"clip_ratio": 0.0,
|
| 112 |
+
"completion_length": 537.2634048461914,
|
| 113 |
+
"epoch": 0.08111533586818757,
|
| 114 |
+
"grad_norm": 0.19215919077396393,
|
| 115 |
+
"learning_rate": 4e-07,
|
| 116 |
+
"loss": -0.0325,
|
| 117 |
+
"num_tokens": 5070421.0,
|
| 118 |
+
"reward": 0.1785714365541935,
|
| 119 |
+
"reward_std": 0.2945236265659332,
|
| 120 |
+
"rewards/accuracy_reward": 0.1696428582072258,
|
| 121 |
+
"rewards/format_reward": 0.008928571827709675,
|
| 122 |
+
"step": 8
|
| 123 |
+
},
|
| 124 |
+
{
|
| 125 |
+
"clip_ratio": 0.0,
|
| 126 |
+
"completion_length": 575.7366256713867,
|
| 127 |
+
"epoch": 0.09125475285171103,
|
| 128 |
+
"grad_norm": 0.14393264055252075,
|
| 129 |
+
"learning_rate": 4.5e-07,
|
| 130 |
+
"loss": -0.014,
|
| 131 |
+
"num_tokens": 5709609.0,
|
| 132 |
+
"reward": 0.14843750465661287,
|
| 133 |
+
"reward_std": 0.22998391091823578,
|
| 134 |
+
"rewards/accuracy_reward": 0.14285714272409678,
|
| 135 |
+
"rewards/format_reward": 0.005580357392318547,
|
| 136 |
+
"step": 9
|
| 137 |
+
},
|
| 138 |
+
{
|
| 139 |
+
"clip_ratio": 0.0,
|
| 140 |
+
"completion_length": 519.7522468566895,
|
| 141 |
+
"epoch": 0.10139416983523447,
|
| 142 |
+
"grad_norm": 0.16316372156143188,
|
| 143 |
+
"learning_rate": 5e-07,
|
| 144 |
+
"loss": -0.0266,
|
| 145 |
+
"num_tokens": 6302035.0,
|
| 146 |
+
"reward": 0.1930803656578064,
|
| 147 |
+
"reward_std": 0.2917625065892935,
|
| 148 |
+
"rewards/accuracy_reward": 0.1863839291036129,
|
| 149 |
+
"rewards/format_reward": 0.006696428870782256,
|
| 150 |
+
"step": 10
|
| 151 |
+
},
|
| 152 |
+
{
|
| 153 |
+
"clip_ratio": 0.0,
|
| 154 |
+
"completion_length": 595.2377395629883,
|
| 155 |
+
"epoch": 0.11153358681875793,
|
| 156 |
+
"grad_norm": 0.14569994807243347,
|
| 157 |
+
"learning_rate": 5.5e-07,
|
| 158 |
+
"loss": -0.0097,
|
| 159 |
+
"num_tokens": 6972904.0,
|
| 160 |
+
"reward": 0.17299108020961285,
|
| 161 |
+
"reward_std": 0.27831926569342613,
|
| 162 |
+
"rewards/accuracy_reward": 0.16183035913854837,
|
| 163 |
+
"rewards/format_reward": 0.01116071455180645,
|
| 164 |
+
"step": 11
|
| 165 |
+
},
|
| 166 |
+
{
|
| 167 |
+
"clip_ratio": 0.0,
|
| 168 |
+
"completion_length": 575.3605155944824,
|
| 169 |
+
"epoch": 0.12167300380228137,
|
| 170 |
+
"grad_norm": 0.15014010667800903,
|
| 171 |
+
"learning_rate": 6e-07,
|
| 172 |
+
"loss": -0.0216,
|
| 173 |
+
"num_tokens": 7619739.0,
|
| 174 |
+
"reward": 0.17187501303851604,
|
| 175 |
+
"reward_std": 0.2696072347462177,
|
| 176 |
+
"rewards/accuracy_reward": 0.16183035634458065,
|
| 177 |
+
"rewards/format_reward": 0.010044643306173384,
|
| 178 |
+
"step": 12
|
| 179 |
+
},
|
| 180 |
+
{
|
| 181 |
+
"clip_ratio": 0.0,
|
| 182 |
+
"completion_length": 579.7835006713867,
|
| 183 |
+
"epoch": 0.13181242078580482,
|
| 184 |
+
"grad_norm": 0.14231476187705994,
|
| 185 |
+
"learning_rate": 6.5e-07,
|
| 186 |
+
"loss": 0.0066,
|
| 187 |
+
"num_tokens": 8267409.0,
|
| 188 |
+
"reward": 0.18191965110599995,
|
| 189 |
+
"reward_std": 0.27643212117254734,
|
| 190 |
+
"rewards/accuracy_reward": 0.17299107369035482,
|
| 191 |
+
"rewards/format_reward": 0.008928571827709675,
|
| 192 |
+
"step": 13
|
| 193 |
+
},
|
| 194 |
+
{
|
| 195 |
+
"clip_ratio": 0.0,
|
| 196 |
+
"completion_length": 560.2623062133789,
|
| 197 |
+
"epoch": 0.14195183776932827,
|
| 198 |
+
"grad_norm": 0.1491479128599167,
|
| 199 |
+
"learning_rate": 7e-07,
|
| 200 |
+
"loss": -0.007,
|
| 201 |
+
"num_tokens": 8901580.0,
|
| 202 |
+
"reward": 0.17745536752045155,
|
| 203 |
+
"reward_std": 0.28200731612741947,
|
| 204 |
+
"rewards/accuracy_reward": 0.1704155197367072,
|
| 205 |
+
"rewards/format_reward": 0.007812500349245965,
|
| 206 |
+
"step": 14
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"clip_ratio": 0.0,
|
| 210 |
+
"completion_length": 563.2969017028809,
|
| 211 |
+
"epoch": 0.1520912547528517,
|
| 212 |
+
"grad_norm": 0.16368582844734192,
|
| 213 |
+
"learning_rate": 7.5e-07,
|
| 214 |
+
"loss": -0.0445,
|
| 215 |
+
"num_tokens": 9534734.0,
|
| 216 |
+
"reward": 0.22544643841683865,
|
| 217 |
+
"reward_std": 0.3410548157989979,
|
| 218 |
+
"rewards/accuracy_reward": 0.20870535541325808,
|
| 219 |
+
"rewards/format_reward": 0.016741071944124997,
|
| 220 |
+
"step": 15
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"clip_ratio": 0.0,
|
| 224 |
+
"completion_length": 575.6607360839844,
|
| 225 |
+
"epoch": 0.16223067173637515,
|
| 226 |
+
"grad_norm": 0.155720517039299,
|
| 227 |
+
"learning_rate": 8e-07,
|
| 228 |
+
"loss": -0.0125,
|
| 229 |
+
"num_tokens": 10177454.0,
|
| 230 |
+
"reward": 0.2332589402794838,
|
| 231 |
+
"reward_std": 0.33312445878982544,
|
| 232 |
+
"rewards/accuracy_reward": 0.21986606996506453,
|
| 233 |
+
"rewards/format_reward": 0.01339285762514919,
|
| 234 |
+
"step": 16
|
| 235 |
+
},
|
| 236 |
+
{
|
| 237 |
+
"clip_ratio": 0.0,
|
| 238 |
+
"completion_length": 605.8471221923828,
|
| 239 |
+
"epoch": 0.17237008871989862,
|
| 240 |
+
"grad_norm": 0.18089956045150757,
|
| 241 |
+
"learning_rate": 8.499999999999999e-07,
|
| 242 |
+
"loss": -0.0189,
|
| 243 |
+
"num_tokens": 10854805.0,
|
| 244 |
+
"reward": 0.22433037124574184,
|
| 245 |
+
"reward_std": 0.3147235903888941,
|
| 246 |
+
"rewards/accuracy_reward": 0.20647321455180645,
|
| 247 |
+
"rewards/format_reward": 0.017857143306173384,
|
| 248 |
+
"step": 17
|
| 249 |
+
},
|
| 250 |
+
{
|
| 251 |
+
"clip_ratio": 0.0,
|
| 252 |
+
"completion_length": 620.2031478881836,
|
| 253 |
+
"epoch": 0.18250950570342206,
|
| 254 |
+
"grad_norm": 0.14688049256801605,
|
| 255 |
+
"learning_rate": 9e-07,
|
| 256 |
+
"loss": -0.0215,
|
| 257 |
+
"num_tokens": 11533851.0,
|
| 258 |
+
"reward": 0.26227680034935474,
|
| 259 |
+
"reward_std": 0.32751886174082756,
|
| 260 |
+
"rewards/accuracy_reward": 0.2522321417927742,
|
| 261 |
+
"rewards/format_reward": 0.010044643189758062,
|
| 262 |
+
"step": 18
|
| 263 |
+
},
|
| 264 |
+
{
|
| 265 |
+
"clip_ratio": 0.0,
|
| 266 |
+
"completion_length": 567.4386444091797,
|
| 267 |
+
"epoch": 0.1926489226869455,
|
| 268 |
+
"grad_norm": 0.18814197182655334,
|
| 269 |
+
"learning_rate": 9.499999999999999e-07,
|
| 270 |
+
"loss": -0.0224,
|
| 271 |
+
"num_tokens": 12174260.0,
|
| 272 |
+
"reward": 0.33370537497103214,
|
| 273 |
+
"reward_std": 0.3732391260564327,
|
| 274 |
+
"rewards/accuracy_reward": 0.3169642798602581,
|
| 275 |
+
"rewards/format_reward": 0.016741071944124997,
|
| 276 |
+
"step": 19
|
| 277 |
+
},
|
| 278 |
+
{
|
| 279 |
+
"clip_ratio": 0.0,
|
| 280 |
+
"completion_length": 620.7399749755859,
|
| 281 |
+
"epoch": 0.20278833967046894,
|
| 282 |
+
"grad_norm": 0.1669970452785492,
|
| 283 |
+
"learning_rate": 1e-06,
|
| 284 |
+
"loss": -0.0292,
|
| 285 |
+
"num_tokens": 12850867.0,
|
| 286 |
+
"reward": 0.3582589402794838,
|
| 287 |
+
"reward_std": 0.3860209658741951,
|
| 288 |
+
"rewards/accuracy_reward": 0.3359375037252903,
|
| 289 |
+
"rewards/format_reward": 0.02232142922002822,
|
| 290 |
+
"step": 20
|
| 291 |
+
},
|
| 292 |
+
{
|
| 293 |
+
"clip_ratio": 0.0,
|
| 294 |
+
"completion_length": 607.200927734375,
|
| 295 |
+
"epoch": 0.21292775665399238,
|
| 296 |
+
"grad_norm": 1.3311952352523804,
|
| 297 |
+
"learning_rate": 9.999203468625015e-07,
|
| 298 |
+
"loss": -0.0164,
|
| 299 |
+
"num_tokens": 13516191.0,
|
| 300 |
+
"reward": 0.3962053768336773,
|
| 301 |
+
"reward_std": 0.3798532336950302,
|
| 302 |
+
"rewards/accuracy_reward": 0.3694196417927742,
|
| 303 |
+
"rewards/format_reward": 0.026785714901052415,
|
| 304 |
+
"step": 21
|
| 305 |
+
},
|
| 306 |
+
{
|
| 307 |
+
"clip_ratio": 0.0,
|
| 308 |
+
"completion_length": 682.4051742553711,
|
| 309 |
+
"epoch": 0.22306717363751585,
|
| 310 |
+
"grad_norm": 0.14331556856632233,
|
| 311 |
+
"learning_rate": 9.99681412828496e-07,
|
| 312 |
+
"loss": 0.0035,
|
| 313 |
+
"num_tokens": 14260786.0,
|
| 314 |
+
"reward": 0.4140625186264515,
|
| 315 |
+
"reward_std": 0.3847779668867588,
|
| 316 |
+
"rewards/accuracy_reward": 0.3750000037252903,
|
| 317 |
+
"rewards/format_reward": 0.03906250128056854,
|
| 318 |
+
"step": 22
|
| 319 |
+
},
|
| 320 |
+
{
|
| 321 |
+
"clip_ratio": 0.0,
|
| 322 |
+
"completion_length": 633.8303909301758,
|
| 323 |
+
"epoch": 0.2332065906210393,
|
| 324 |
+
"grad_norm": 0.1424940973520279,
|
| 325 |
+
"learning_rate": 9.992832740253644e-07,
|
| 326 |
+
"loss": 0.01,
|
| 327 |
+
"num_tokens": 14966946.0,
|
| 328 |
+
"reward": 0.427455373108387,
|
| 329 |
+
"reward_std": 0.3340139761567116,
|
| 330 |
+
"rewards/accuracy_reward": 0.4073660746216774,
|
| 331 |
+
"rewards/format_reward": 0.02008928614668548,
|
| 332 |
+
"step": 23
|
| 333 |
+
},
|
| 334 |
+
{
|
| 335 |
+
"clip_ratio": 0.0,
|
| 336 |
+
"completion_length": 693.4665451049805,
|
| 337 |
+
"epoch": 0.24334600760456274,
|
| 338 |
+
"grad_norm": 0.1358347088098526,
|
| 339 |
+
"learning_rate": 9.987260573051267e-07,
|
| 340 |
+
"loss": 0.021,
|
| 341 |
+
"num_tokens": 15717836.0,
|
| 342 |
+
"reward": 0.4095982313156128,
|
| 343 |
+
"reward_std": 0.3687121346592903,
|
| 344 |
+
"rewards/accuracy_reward": 0.3805803619325161,
|
| 345 |
+
"rewards/format_reward": 0.029017857974395156,
|
| 346 |
+
"step": 24
|
| 347 |
+
},
|
| 348 |
+
{
|
| 349 |
+
"clip_ratio": 0.0,
|
| 350 |
+
"completion_length": 679.7701263427734,
|
| 351 |
+
"epoch": 0.2534854245880862,
|
| 352 |
+
"grad_norm": 0.12858213484287262,
|
| 353 |
+
"learning_rate": 9.98009940204023e-07,
|
| 354 |
+
"loss": 0.0077,
|
| 355 |
+
"num_tokens": 16458238.0,
|
| 356 |
+
"reward": 0.4821428805589676,
|
| 357 |
+
"reward_std": 0.3479448612779379,
|
| 358 |
+
"rewards/accuracy_reward": 0.4531250074505806,
|
| 359 |
+
"rewards/format_reward": 0.029017857857979834,
|
| 360 |
+
"step": 25
|
| 361 |
+
},
|
| 362 |
+
{
|
| 363 |
+
"clip_ratio": 0.0,
|
| 364 |
+
"completion_length": 643.7042846679688,
|
| 365 |
+
"epoch": 0.26362484157160965,
|
| 366 |
+
"grad_norm": 0.13453267514705658,
|
| 367 |
+
"learning_rate": 9.971351508859486e-07,
|
| 368 |
+
"loss": 0.0002,
|
| 369 |
+
"num_tokens": 17153829.0,
|
| 370 |
+
"reward": 0.4787946566939354,
|
| 371 |
+
"reward_std": 0.33469830453395844,
|
| 372 |
+
"rewards/accuracy_reward": 0.457589291036129,
|
| 373 |
+
"rewards/format_reward": 0.02120535762514919,
|
| 374 |
+
"step": 26
|
| 375 |
+
},
|
| 376 |
+
{
|
| 377 |
+
"clip_ratio": 0.0,
|
| 378 |
+
"completion_length": 647.9174499511719,
|
| 379 |
+
"epoch": 0.2737642585551331,
|
| 380 |
+
"grad_norm": 0.13937269151210785,
|
| 381 |
+
"learning_rate": 9.961019680697591e-07,
|
| 382 |
+
"loss": 0.031,
|
| 383 |
+
"num_tokens": 17862539.0,
|
| 384 |
+
"reward": 0.5033482313156128,
|
| 385 |
+
"reward_std": 0.34470784291625023,
|
| 386 |
+
"rewards/accuracy_reward": 0.4676339291036129,
|
| 387 |
+
"rewards/format_reward": 0.0357142862631008,
|
| 388 |
+
"step": 27
|
| 389 |
+
},
|
| 390 |
+
{
|
| 391 |
+
"clip_ratio": 0.0,
|
| 392 |
+
"completion_length": 729.0391006469727,
|
| 393 |
+
"epoch": 0.28390367553865653,
|
| 394 |
+
"grad_norm": 0.12249790877103806,
|
| 395 |
+
"learning_rate": 9.949107209404663e-07,
|
| 396 |
+
"loss": 0.039,
|
| 397 |
+
"num_tokens": 18649398.0,
|
| 398 |
+
"reward": 0.5368303805589676,
|
| 399 |
+
"reward_std": 0.3495071418583393,
|
| 400 |
+
"rewards/accuracy_reward": 0.5122767835855484,
|
| 401 |
+
"rewards/format_reward": 0.02455357217695564,
|
| 402 |
+
"step": 28
|
| 403 |
+
},
|
| 404 |
+
{
|
| 405 |
+
"clip_ratio": 0.0,
|
| 406 |
+
"completion_length": 696.9040451049805,
|
| 407 |
+
"epoch": 0.29404309252217997,
|
| 408 |
+
"grad_norm": 0.12198278307914734,
|
| 409 |
+
"learning_rate": 9.935617890443554e-07,
|
| 410 |
+
"loss": 0.0252,
|
| 411 |
+
"num_tokens": 19414560.0,
|
| 412 |
+
"reward": 0.541294664144516,
|
| 413 |
+
"reward_std": 0.30416675843298435,
|
| 414 |
+
"rewards/accuracy_reward": 0.5145089328289032,
|
| 415 |
+
"rewards/format_reward": 0.026785714901052415,
|
| 416 |
+
"step": 29
|
| 417 |
+
},
|
| 418 |
+
{
|
| 419 |
+
"clip_ratio": 0.0,
|
| 420 |
+
"completion_length": 760.1384353637695,
|
| 421 |
+
"epoch": 0.3041825095057034,
|
| 422 |
+
"grad_norm": 0.12740154564380646,
|
| 423 |
+
"learning_rate": 9.92055602168058e-07,
|
| 424 |
+
"loss": 0.0327,
|
| 425 |
+
"num_tokens": 20241132.0,
|
| 426 |
+
"reward": 0.5569196678698063,
|
| 427 |
+
"reward_std": 0.30499308556318283,
|
| 428 |
+
"rewards/accuracy_reward": 0.5145089253783226,
|
| 429 |
+
"rewards/format_reward": 0.04241071501746774,
|
| 430 |
+
"step": 30
|
| 431 |
+
},
|
| 432 |
+
{
|
| 433 |
+
"clip_ratio": 0.0,
|
| 434 |
+
"completion_length": 727.1250381469727,
|
| 435 |
+
"epoch": 0.31432192648922685,
|
| 436 |
+
"grad_norm": 0.13364288210868835,
|
| 437 |
+
"learning_rate": 9.90392640201615e-07,
|
| 438 |
+
"loss": 0.0145,
|
| 439 |
+
"num_tokens": 21020244.0,
|
| 440 |
+
"reward": 0.6316964700818062,
|
| 441 |
+
"reward_std": 0.35084592178463936,
|
| 442 |
+
"rewards/accuracy_reward": 0.5758928619325161,
|
| 443 |
+
"rewards/format_reward": 0.05580357206054032,
|
| 444 |
+
"step": 31
|
| 445 |
+
},
|
| 446 |
+
{
|
| 447 |
+
"clip_ratio": 0.0,
|
| 448 |
+
"completion_length": 743.0826187133789,
|
| 449 |
+
"epoch": 0.3244613434727503,
|
| 450 |
+
"grad_norm": 0.1573316901922226,
|
| 451 |
+
"learning_rate": 9.885734329855797e-07,
|
| 452 |
+
"loss": 0.047,
|
| 453 |
+
"num_tokens": 21805790.0,
|
| 454 |
+
"reward": 0.6562500223517418,
|
| 455 |
+
"reward_std": 0.3445834815502167,
|
| 456 |
+
"rewards/accuracy_reward": 0.5792410708963871,
|
| 457 |
+
"rewards/format_reward": 0.0770089291036129,
|
| 458 |
+
"step": 32
|
| 459 |
+
},
|
| 460 |
+
{
|
| 461 |
+
"clip_ratio": 0.0,
|
| 462 |
+
"completion_length": 750.4531631469727,
|
| 463 |
+
"epoch": 0.33460076045627374,
|
| 464 |
+
"grad_norm": 0.15323247015476227,
|
| 465 |
+
"learning_rate": 9.865985601422017e-07,
|
| 466 |
+
"loss": 0.0446,
|
| 467 |
+
"num_tokens": 22612252.0,
|
| 468 |
+
"reward": 0.6238839514553547,
|
| 469 |
+
"reward_std": 0.41414720192551613,
|
| 470 |
+
"rewards/accuracy_reward": 0.49690933898091316,
|
| 471 |
+
"rewards/format_reward": 0.13058035727590322,
|
| 472 |
+
"step": 33
|
| 473 |
+
},
|
| 474 |
+
{
|
| 475 |
+
"clip_ratio": 0.0,
|
| 476 |
+
"completion_length": 765.1116485595703,
|
| 477 |
+
"epoch": 0.34474017743979724,
|
| 478 |
+
"grad_norm": 0.22268784046173096,
|
| 479 |
+
"learning_rate": 9.844686508907537e-07,
|
| 480 |
+
"loss": 0.0329,
|
| 481 |
+
"num_tokens": 23425280.0,
|
| 482 |
+
"reward": 0.7979911044239998,
|
| 483 |
+
"reward_std": 0.4762897975742817,
|
| 484 |
+
"rewards/accuracy_reward": 0.5691964253783226,
|
| 485 |
+
"rewards/format_reward": 0.22879464365541935,
|
| 486 |
+
"step": 34
|
| 487 |
+
},
|
| 488 |
+
{
|
| 489 |
+
"clip_ratio": 0.0,
|
| 490 |
+
"completion_length": 733.958740234375,
|
| 491 |
+
"epoch": 0.3548795944233207,
|
| 492 |
+
"grad_norm": 0.21707987785339355,
|
| 493 |
+
"learning_rate": 9.821843838470534e-07,
|
| 494 |
+
"loss": 0.034,
|
| 495 |
+
"num_tokens": 24225635.0,
|
| 496 |
+
"reward": 0.9587054029107094,
|
| 497 |
+
"reward_std": 0.53031075745821,
|
| 498 |
+
"rewards/accuracy_reward": 0.5680803619325161,
|
| 499 |
+
"rewards/format_reward": 0.390625,
|
| 500 |
+
"step": 35
|
| 501 |
+
},
|
| 502 |
+
{
|
| 503 |
+
"clip_ratio": 0.0,
|
| 504 |
+
"completion_length": 737.1797103881836,
|
| 505 |
+
"epoch": 0.3650190114068441,
|
| 506 |
+
"grad_norm": 0.1912337839603424,
|
| 507 |
+
"learning_rate": 9.797464868072486e-07,
|
| 508 |
+
"loss": 0.033,
|
| 509 |
+
"num_tokens": 25014260.0,
|
| 510 |
+
"reward": 1.112723246216774,
|
| 511 |
+
"reward_std": 0.574417520314455,
|
| 512 |
+
"rewards/accuracy_reward": 0.5881696492433548,
|
| 513 |
+
"rewards/format_reward": 0.5245535708963871,
|
| 514 |
+
"step": 36
|
| 515 |
+
},
|
| 516 |
+
{
|
| 517 |
+
"clip_ratio": 0.0,
|
| 518 |
+
"completion_length": 740.9598541259766,
|
| 519 |
+
"epoch": 0.37515842839036756,
|
| 520 |
+
"grad_norm": 0.22646164894104004,
|
| 521 |
+
"learning_rate": 9.771557365159319e-07,
|
| 522 |
+
"loss": 0.0325,
|
| 523 |
+
"num_tokens": 25803688.0,
|
| 524 |
+
"reward": 1.1941964775323868,
|
| 525 |
+
"reward_std": 0.5038099698722363,
|
| 526 |
+
"rewards/accuracy_reward": 0.5524553544819355,
|
| 527 |
+
"rewards/format_reward": 0.6417410746216774,
|
| 528 |
+
"step": 37
|
| 529 |
+
},
|
| 530 |
+
{
|
| 531 |
+
"clip_ratio": 0.0,
|
| 532 |
+
"completion_length": 741.780158996582,
|
| 533 |
+
"epoch": 0.385297845373891,
|
| 534 |
+
"grad_norm": 0.2826821208000183,
|
| 535 |
+
"learning_rate": 9.744129584186597e-07,
|
| 536 |
+
"loss": 0.0653,
|
| 537 |
+
"num_tokens": 26594931.0,
|
| 538 |
+
"reward": 1.340401828289032,
|
| 539 |
+
"reward_std": 0.47122547030448914,
|
| 540 |
+
"rewards/accuracy_reward": 0.5691964253783226,
|
| 541 |
+
"rewards/format_reward": 0.7712053582072258,
|
| 542 |
+
"step": 38
|
| 543 |
+
},
|
| 544 |
+
{
|
| 545 |
+
"clip_ratio": 0.0,
|
| 546 |
+
"completion_length": 783.4699020385742,
|
| 547 |
+
"epoch": 0.39543726235741444,
|
| 548 |
+
"grad_norm": 0.1316017210483551,
|
| 549 |
+
"learning_rate": 9.71519026398956e-07,
|
| 550 |
+
"loss": 0.0629,
|
| 551 |
+
"num_tokens": 27434704.0,
|
| 552 |
+
"reward": 1.3526786416769028,
|
| 553 |
+
"reward_std": 0.40762051939964294,
|
| 554 |
+
"rewards/accuracy_reward": 0.5200892835855484,
|
| 555 |
+
"rewards/format_reward": 0.832589291036129,
|
| 556 |
+
"step": 39
|
| 557 |
+
},
|
| 558 |
+
{
|
| 559 |
+
"clip_ratio": 0.0,
|
| 560 |
+
"completion_length": 709.8928909301758,
|
| 561 |
+
"epoch": 0.4055766793409379,
|
| 562 |
+
"grad_norm": 0.12696978449821472,
|
| 563 |
+
"learning_rate": 9.68474862499881e-07,
|
| 564 |
+
"loss": 0.0423,
|
| 565 |
+
"num_tokens": 28190032.0,
|
| 566 |
+
"reward": 1.491071492433548,
|
| 567 |
+
"reward_std": 0.35216479003429413,
|
| 568 |
+
"rewards/accuracy_reward": 0.5848214253783226,
|
| 569 |
+
"rewards/format_reward": 0.9062499925494194,
|
| 570 |
+
"step": 40
|
| 571 |
+
},
|
| 572 |
+
{
|
| 573 |
+
"clip_ratio": 0.0,
|
| 574 |
+
"completion_length": 700.1216735839844,
|
| 575 |
+
"epoch": 0.4157160963244613,
|
| 576 |
+
"grad_norm": 0.1216612383723259,
|
| 577 |
+
"learning_rate": 9.652814366302568e-07,
|
| 578 |
+
"loss": 0.0672,
|
| 579 |
+
"num_tokens": 28943317.0,
|
| 580 |
+
"reward": 1.5658482760190964,
|
| 581 |
+
"reward_std": 0.34097408317029476,
|
| 582 |
+
"rewards/accuracy_reward": 0.6484375,
|
| 583 |
+
"rewards/format_reward": 0.917410708963871,
|
| 584 |
+
"step": 41
|
| 585 |
+
},
|
| 586 |
+
{
|
| 587 |
+
"clip_ratio": 0.0,
|
| 588 |
+
"completion_length": 703.3582916259766,
|
| 589 |
+
"epoch": 0.42585551330798477,
|
| 590 |
+
"grad_norm": 0.11864102631807327,
|
| 591 |
+
"learning_rate": 9.619397662556433e-07,
|
| 592 |
+
"loss": 0.0279,
|
| 593 |
+
"num_tokens": 29702494.0,
|
| 594 |
+
"reward": 1.537946492433548,
|
| 595 |
+
"reward_std": 0.2801562137901783,
|
| 596 |
+
"rewards/accuracy_reward": 0.5892857164144516,
|
| 597 |
+
"rewards/format_reward": 0.9486607164144516,
|
| 598 |
+
"step": 42
|
| 599 |
+
},
|
| 600 |
+
{
|
| 601 |
+
"clip_ratio": 0.0,
|
| 602 |
+
"completion_length": 775.4074020385742,
|
| 603 |
+
"epoch": 0.43599493029150826,
|
| 604 |
+
"grad_norm": 0.10538308322429657,
|
| 605 |
+
"learning_rate": 9.5845091607416e-07,
|
| 606 |
+
"loss": 0.0566,
|
| 607 |
+
"num_tokens": 30515963.0,
|
| 608 |
+
"reward": 1.5178572088479996,
|
| 609 |
+
"reward_std": 0.33493360690772533,
|
| 610 |
+
"rewards/accuracy_reward": 0.5758928544819355,
|
| 611 |
+
"rewards/format_reward": 0.9419642984867096,
|
| 612 |
+
"step": 43
|
| 613 |
+
},
|
| 614 |
+
{
|
| 615 |
+
"clip_ratio": 0.0,
|
| 616 |
+
"completion_length": 792.2288360595703,
|
| 617 |
+
"epoch": 0.4461343472750317,
|
| 618 |
+
"grad_norm": 0.10541502386331558,
|
| 619 |
+
"learning_rate": 9.548159976772592e-07,
|
| 620 |
+
"loss": 0.0749,
|
| 621 |
+
"num_tokens": 31355848.0,
|
| 622 |
+
"reward": 1.53683041036129,
|
| 623 |
+
"reward_std": 0.3079599365592003,
|
| 624 |
+
"rewards/accuracy_reward": 0.5747767873108387,
|
| 625 |
+
"rewards/format_reward": 0.9620535746216774,
|
| 626 |
+
"step": 44
|
| 627 |
+
},
|
| 628 |
+
{
|
| 629 |
+
"clip_ratio": 0.0,
|
| 630 |
+
"completion_length": 785.3750228881836,
|
| 631 |
+
"epoch": 0.45627376425855515,
|
| 632 |
+
"grad_norm": 0.11054891347885132,
|
| 633 |
+
"learning_rate": 9.510361691955606e-07,
|
| 634 |
+
"loss": 0.0749,
|
| 635 |
+
"num_tokens": 32181992.0,
|
| 636 |
+
"reward": 1.4888393431901932,
|
| 637 |
+
"reward_std": 0.33042189106345177,
|
| 638 |
+
"rewards/accuracy_reward": 0.5234375037252903,
|
| 639 |
+
"rewards/format_reward": 0.9654017835855484,
|
| 640 |
+
"step": 45
|
| 641 |
+
},
|
| 642 |
+
{
|
| 643 |
+
"clip_ratio": 0.0,
|
| 644 |
+
"completion_length": 743.7846298217773,
|
| 645 |
+
"epoch": 0.4664131812420786,
|
| 646 |
+
"grad_norm": 0.10712749511003494,
|
| 647 |
+
"learning_rate": 9.471126349298556e-07,
|
| 648 |
+
"loss": 0.0636,
|
| 649 |
+
"num_tokens": 32978639.0,
|
| 650 |
+
"reward": 1.5468750447034836,
|
| 651 |
+
"reward_std": 0.34593483060598373,
|
| 652 |
+
"rewards/accuracy_reward": 0.5770089402794838,
|
| 653 |
+
"rewards/format_reward": 0.9698660746216774,
|
| 654 |
+
"step": 46
|
| 655 |
+
},
|
| 656 |
+
{
|
| 657 |
+
"clip_ratio": 0.0,
|
| 658 |
+
"completion_length": 732.7355194091797,
|
| 659 |
+
"epoch": 0.47655259822560203,
|
| 660 |
+
"grad_norm": 0.1037302166223526,
|
| 661 |
+
"learning_rate": 9.430466449674013e-07,
|
| 662 |
+
"loss": 0.0546,
|
| 663 |
+
"num_tokens": 33783522.0,
|
| 664 |
+
"reward": 1.537946492433548,
|
| 665 |
+
"reward_std": 0.2701996862888336,
|
| 666 |
+
"rewards/accuracy_reward": 0.5558035708963871,
|
| 667 |
+
"rewards/format_reward": 0.9821428582072258,
|
| 668 |
+
"step": 47
|
| 669 |
+
},
|
| 670 |
+
{
|
| 671 |
+
"clip_ratio": 0.0,
|
| 672 |
+
"completion_length": 794.5357513427734,
|
| 673 |
+
"epoch": 0.4866920152091255,
|
| 674 |
+
"grad_norm": 0.09865893423557281,
|
| 675 |
+
"learning_rate": 9.388394947836278e-07,
|
| 676 |
+
"loss": 0.0496,
|
| 677 |
+
"num_tokens": 34622162.0,
|
| 678 |
+
"reward": 1.4854911416769028,
|
| 679 |
+
"reward_std": 0.27795105054974556,
|
| 680 |
+
"rewards/accuracy_reward": 0.5133928544819355,
|
| 681 |
+
"rewards/format_reward": 0.972098208963871,
|
| 682 |
+
"step": 48
|
| 683 |
+
},
|
| 684 |
+
{
|
| 685 |
+
"clip_ratio": 0.0,
|
| 686 |
+
"completion_length": 784.8024978637695,
|
| 687 |
+
"epoch": 0.4968314321926489,
|
| 688 |
+
"grad_norm": 0.09806529432535172,
|
| 689 |
+
"learning_rate": 9.344925248293835e-07,
|
| 690 |
+
"loss": 0.0469,
|
| 691 |
+
"num_tokens": 35446353.0,
|
| 692 |
+
"reward": 1.5401786416769028,
|
| 693 |
+
"reward_std": 0.2609728015959263,
|
| 694 |
+
"rewards/accuracy_reward": 0.5658482201397419,
|
| 695 |
+
"rewards/format_reward": 0.9743303582072258,
|
| 696 |
+
"step": 49
|
| 697 |
+
},
|
| 698 |
+
{
|
| 699 |
+
"clip_ratio": 0.0,
|
| 700 |
+
"completion_length": 755.7600708007812,
|
| 701 |
+
"epoch": 0.5069708491761724,
|
| 702 |
+
"grad_norm": 0.09463509172201157,
|
| 703 |
+
"learning_rate": 9.300071201038501e-07,
|
| 704 |
+
"loss": 0.0447,
|
| 705 |
+
"num_tokens": 36249690.0,
|
| 706 |
+
"reward": 1.5591518729925156,
|
| 707 |
+
"reward_std": 0.27370808832347393,
|
| 708 |
+
"rewards/accuracy_reward": 0.5837053693830967,
|
| 709 |
+
"rewards/format_reward": 0.9754464253783226,
|
| 710 |
+
"step": 50
|
| 711 |
+
},
|
| 712 |
+
{
|
| 713 |
+
"clip_ratio": 0.0,
|
| 714 |
+
"completion_length": 683.0134201049805,
|
| 715 |
+
"epoch": 0.5171102661596958,
|
| 716 |
+
"grad_norm": 0.1009419783949852,
|
| 717 |
+
"learning_rate": 9.253847097132655e-07,
|
| 718 |
+
"loss": 0.0259,
|
| 719 |
+
"num_tokens": 36992494.0,
|
| 720 |
+
"reward": 1.6160715073347092,
|
| 721 |
+
"reward_std": 0.2591102607548237,
|
| 722 |
+
"rewards/accuracy_reward": 0.6316964328289032,
|
| 723 |
+
"rewards/format_reward": 0.984375,
|
| 724 |
+
"step": 51
|
| 725 |
+
},
|
| 726 |
+
{
|
| 727 |
+
"clip_ratio": 0.0,
|
| 728 |
+
"completion_length": 744.6250228881836,
|
| 729 |
+
"epoch": 0.5272496831432193,
|
| 730 |
+
"grad_norm": 0.10415738075971603,
|
| 731 |
+
"learning_rate": 9.206267664155906e-07,
|
| 732 |
+
"loss": 0.0411,
|
| 733 |
+
"num_tokens": 37780054.0,
|
| 734 |
+
"reward": 1.561383992433548,
|
| 735 |
+
"reward_std": 0.2772454805672169,
|
| 736 |
+
"rewards/accuracy_reward": 0.5825892947614193,
|
| 737 |
+
"rewards/format_reward": 0.9787946417927742,
|
| 738 |
+
"step": 52
|
| 739 |
+
},
|
| 740 |
+
{
|
| 741 |
+
"clip_ratio": 0.0,
|
| 742 |
+
"completion_length": 768.4509201049805,
|
| 743 |
+
"epoch": 0.5373891001267427,
|
| 744 |
+
"grad_norm": 0.09972584992647171,
|
| 745 |
+
"learning_rate": 9.157348061512726e-07,
|
| 746 |
+
"loss": 0.0551,
|
| 747 |
+
"num_tokens": 38604666.0,
|
| 748 |
+
"reward": 1.5491072237491608,
|
| 749 |
+
"reward_std": 0.27522944658994675,
|
| 750 |
+
"rewards/accuracy_reward": 0.5792410783469677,
|
| 751 |
+
"rewards/format_reward": 0.9698660746216774,
|
| 752 |
+
"step": 53
|
| 753 |
+
},
|
| 754 |
+
{
|
| 755 |
+
"clip_ratio": 0.0,
|
| 756 |
+
"completion_length": 828.3806228637695,
|
| 757 |
+
"epoch": 0.5475285171102662,
|
| 758 |
+
"grad_norm": 0.08794871717691422,
|
| 759 |
+
"learning_rate": 9.107103875602458e-07,
|
| 760 |
+
"loss": 0.0472,
|
| 761 |
+
"num_tokens": 39465447.0,
|
| 762 |
+
"reward": 1.5133929252624512,
|
| 763 |
+
"reward_std": 0.29366694390773773,
|
| 764 |
+
"rewards/accuracy_reward": 0.5345982164144516,
|
| 765 |
+
"rewards/format_reward": 0.9787946417927742,
|
| 766 |
+
"step": 54
|
| 767 |
+
},
|
| 768 |
+
{
|
| 769 |
+
"clip_ratio": 0.0,
|
| 770 |
+
"completion_length": 784.3370895385742,
|
| 771 |
+
"epoch": 0.5576679340937896,
|
| 772 |
+
"grad_norm": 0.10152813047170639,
|
| 773 |
+
"learning_rate": 9.055551114853295e-07,
|
| 774 |
+
"loss": 0.049,
|
| 775 |
+
"num_tokens": 40298605.0,
|
| 776 |
+
"reward": 1.507812574505806,
|
| 777 |
+
"reward_std": 0.2921369504183531,
|
| 778 |
+
"rewards/accuracy_reward": 0.5312499925494194,
|
| 779 |
+
"rewards/format_reward": 0.9765624925494194,
|
| 780 |
+
"step": 55
|
| 781 |
+
},
|
| 782 |
+
{
|
| 783 |
+
"clip_ratio": 0.0,
|
| 784 |
+
"completion_length": 721.1373062133789,
|
| 785 |
+
"epoch": 0.5678073510773131,
|
| 786 |
+
"grad_norm": 0.09925797581672668,
|
| 787 |
+
"learning_rate": 9.002706204621802e-07,
|
| 788 |
+
"loss": 0.0244,
|
| 789 |
+
"num_tokens": 41068424.0,
|
| 790 |
+
"reward": 1.585937574505806,
|
| 791 |
+
"reward_std": 0.26984195969998837,
|
| 792 |
+
"rewards/accuracy_reward": 0.5959821343421936,
|
| 793 |
+
"rewards/format_reward": 0.9899553582072258,
|
| 794 |
+
"step": 56
|
| 795 |
+
},
|
| 796 |
+
{
|
| 797 |
+
"clip_ratio": 0.0,
|
| 798 |
+
"completion_length": 754.9319458007812,
|
| 799 |
+
"epoch": 0.5779467680608364,
|
| 800 |
+
"grad_norm": 0.0985010415315628,
|
| 801 |
+
"learning_rate": 8.948585981959578e-07,
|
| 802 |
+
"loss": 0.055,
|
| 803 |
+
"num_tokens": 41876427.0,
|
| 804 |
+
"reward": 1.5926340222358704,
|
| 805 |
+
"reward_std": 0.2599882632493973,
|
| 806 |
+
"rewards/accuracy_reward": 0.618303582072258,
|
| 807 |
+
"rewards/format_reward": 0.9743303507566452,
|
| 808 |
+
"step": 57
|
| 809 |
+
},
|
| 810 |
+
{
|
| 811 |
+
"clip_ratio": 0.0,
|
| 812 |
+
"completion_length": 735.5156631469727,
|
| 813 |
+
"epoch": 0.5880861850443599,
|
| 814 |
+
"grad_norm": 0.09651587158441544,
|
| 815 |
+
"learning_rate": 8.893207690248775e-07,
|
| 816 |
+
"loss": 0.0342,
|
| 817 |
+
"num_tokens": 42656513.0,
|
| 818 |
+
"reward": 1.5904018431901932,
|
| 819 |
+
"reward_std": 0.2695194110274315,
|
| 820 |
+
"rewards/accuracy_reward": 0.6071428507566452,
|
| 821 |
+
"rewards/format_reward": 0.9832589328289032,
|
| 822 |
+
"step": 58
|
| 823 |
+
},
|
| 824 |
+
{
|
| 825 |
+
"clip_ratio": 0.0,
|
| 826 |
+
"completion_length": 701.5614166259766,
|
| 827 |
+
"epoch": 0.5982256020278834,
|
| 828 |
+
"grad_norm": 0.10193536430597305,
|
| 829 |
+
"learning_rate": 8.836588973708128e-07,
|
| 830 |
+
"loss": 0.0322,
|
| 831 |
+
"num_tokens": 43426704.0,
|
| 832 |
+
"reward": 1.537946492433548,
|
| 833 |
+
"reward_std": 0.2486334890127182,
|
| 834 |
+
"rewards/accuracy_reward": 0.5479910708963871,
|
| 835 |
+
"rewards/format_reward": 0.9899553507566452,
|
| 836 |
+
"step": 59
|
| 837 |
+
},
|
| 838 |
+
{
|
| 839 |
+
"clip_ratio": 0.0,
|
| 840 |
+
"completion_length": 767.4029312133789,
|
| 841 |
+
"epoch": 0.6083650190114068,
|
| 842 |
+
"grad_norm": 0.09067587554454803,
|
| 843 |
+
"learning_rate": 8.778747871771291e-07,
|
| 844 |
+
"loss": 0.0462,
|
| 845 |
+
"num_tokens": 44251545.0,
|
| 846 |
+
"reward": 1.5535715073347092,
|
| 847 |
+
"reward_std": 0.2701433580368757,
|
| 848 |
+
"rewards/accuracy_reward": 0.5714285783469677,
|
| 849 |
+
"rewards/format_reward": 0.9821428582072258,
|
| 850 |
+
"step": 60
|
| 851 |
+
},
|
| 852 |
+
{
|
| 853 |
+
"clip_ratio": 0.0,
|
| 854 |
+
"completion_length": 728.3716888427734,
|
| 855 |
+
"epoch": 0.6185044359949303,
|
| 856 |
+
"grad_norm": 0.10155454277992249,
|
| 857 |
+
"learning_rate": 8.719702813339247e-07,
|
| 858 |
+
"loss": 0.0326,
|
| 859 |
+
"num_tokens": 45026510.0,
|
| 860 |
+
"reward": 1.577008992433548,
|
| 861 |
+
"reward_std": 0.2642171438783407,
|
| 862 |
+
"rewards/accuracy_reward": 0.5904017873108387,
|
| 863 |
+
"rewards/format_reward": 0.9866071343421936,
|
| 864 |
+
"step": 61
|
| 865 |
+
},
|
| 866 |
+
{
|
| 867 |
+
"clip_ratio": 0.0,
|
| 868 |
+
"completion_length": 777.1127548217773,
|
| 869 |
+
"epoch": 0.6286438529784537,
|
| 870 |
+
"grad_norm": 0.09913618117570877,
|
| 871 |
+
"learning_rate": 8.659472610908627e-07,
|
| 872 |
+
"loss": 0.0358,
|
| 873 |
+
"num_tokens": 45852195.0,
|
| 874 |
+
"reward": 1.5312500894069672,
|
| 875 |
+
"reward_std": 0.27686445228755474,
|
| 876 |
+
"rewards/accuracy_reward": 0.5446428544819355,
|
| 877 |
+
"rewards/format_reward": 0.986607126891613,
|
| 878 |
+
"step": 62
|
| 879 |
+
},
|
| 880 |
+
{
|
| 881 |
+
"clip_ratio": 0.0,
|
| 882 |
+
"completion_length": 697.9274826049805,
|
| 883 |
+
"epoch": 0.6387832699619772,
|
| 884 |
+
"grad_norm": 0.10096921026706696,
|
| 885 |
+
"learning_rate": 8.598076454577814e-07,
|
| 886 |
+
"loss": 0.0279,
|
| 887 |
+
"num_tokens": 46601834.0,
|
| 888 |
+
"reward": 1.6305804252624512,
|
| 889 |
+
"reward_std": 0.2701211776584387,
|
| 890 |
+
"rewards/accuracy_reward": 0.6428571492433548,
|
| 891 |
+
"rewards/format_reward": 0.9877232015132904,
|
| 892 |
+
"step": 63
|
| 893 |
+
},
|
| 894 |
+
{
|
| 895 |
+
"clip_ratio": 0.0,
|
| 896 |
+
"completion_length": 724.1774826049805,
|
| 897 |
+
"epoch": 0.6489226869455006,
|
| 898 |
+
"grad_norm": 0.09761769324541092,
|
| 899 |
+
"learning_rate": 8.535533905932737e-07,
|
| 900 |
+
"loss": 0.0398,
|
| 901 |
+
"num_tokens": 47398825.0,
|
| 902 |
+
"reward": 1.601562574505806,
|
| 903 |
+
"reward_std": 0.23534459993243217,
|
| 904 |
+
"rewards/accuracy_reward": 0.6093749962747097,
|
| 905 |
+
"rewards/format_reward": 0.9921875,
|
| 906 |
+
"step": 64
|
| 907 |
+
},
|
| 908 |
+
{
|
| 909 |
+
"clip_ratio": 0.0,
|
| 910 |
+
"completion_length": 717.3169937133789,
|
| 911 |
+
"epoch": 0.6590621039290241,
|
| 912 |
+
"grad_norm": 0.11670506000518799,
|
| 913 |
+
"learning_rate": 8.471864891814304e-07,
|
| 914 |
+
"loss": 0.0516,
|
| 915 |
+
"num_tokens": 48160645.0,
|
| 916 |
+
"reward": 1.6060268580913544,
|
| 917 |
+
"reward_std": 0.26124117337167263,
|
| 918 |
+
"rewards/accuracy_reward": 0.6216517798602581,
|
| 919 |
+
"rewards/format_reward": 0.984375,
|
| 920 |
+
"step": 65
|
| 921 |
+
},
|
| 922 |
+
{
|
| 923 |
+
"clip_ratio": 0.0,
|
| 924 |
+
"completion_length": 766.2712478637695,
|
| 925 |
+
"epoch": 0.6692015209125475,
|
| 926 |
+
"grad_norm": 0.09610045701265335,
|
| 927 |
+
"learning_rate": 8.407089697969456e-07,
|
| 928 |
+
"loss": 0.0429,
|
| 929 |
+
"num_tokens": 48972088.0,
|
| 930 |
+
"reward": 1.5725447088479996,
|
| 931 |
+
"reward_std": 0.25589474849402905,
|
| 932 |
+
"rewards/accuracy_reward": 0.5937499850988388,
|
| 933 |
+
"rewards/format_reward": 0.9787946417927742,
|
| 934 |
+
"step": 66
|
| 935 |
+
},
|
| 936 |
+
{
|
| 937 |
+
"clip_ratio": 0.0,
|
| 938 |
+
"completion_length": 794.2879867553711,
|
| 939 |
+
"epoch": 0.679340937896071,
|
| 940 |
+
"grad_norm": 0.08830226212739944,
|
| 941 |
+
"learning_rate": 8.341228962587881e-07,
|
| 942 |
+
"loss": 0.0485,
|
| 943 |
+
"num_tokens": 49819898.0,
|
| 944 |
+
"reward": 1.545758992433548,
|
| 945 |
+
"reward_std": 0.2661575675010681,
|
| 946 |
+
"rewards/accuracy_reward": 0.5669642947614193,
|
| 947 |
+
"rewards/format_reward": 0.9787946343421936,
|
| 948 |
+
"step": 67
|
| 949 |
+
},
|
| 950 |
+
{
|
| 951 |
+
"clip_ratio": 0.0,
|
| 952 |
+
"completion_length": 702.197582244873,
|
| 953 |
+
"epoch": 0.6894803548795945,
|
| 954 |
+
"grad_norm": 0.08808374404907227,
|
| 955 |
+
"learning_rate": 8.274303669726426e-07,
|
| 956 |
+
"loss": 0.0427,
|
| 957 |
+
"num_tokens": 50570387.0,
|
| 958 |
+
"reward": 1.641741156578064,
|
| 959 |
+
"reward_std": 0.24876852426677942,
|
| 960 |
+
"rewards/accuracy_reward": 0.6595982052385807,
|
| 961 |
+
"rewards/format_reward": 0.9821428582072258,
|
| 962 |
+
"step": 68
|
| 963 |
+
},
|
| 964 |
+
{
|
| 965 |
+
"clip_ratio": 0.0,
|
| 966 |
+
"completion_length": 780.8024978637695,
|
| 967 |
+
"epoch": 0.6996197718631179,
|
| 968 |
+
"grad_norm": 0.0812569409608841,
|
| 969 |
+
"learning_rate": 8.206335142623304e-07,
|
| 970 |
+
"loss": 0.0349,
|
| 971 |
+
"num_tokens": 51399746.0,
|
| 972 |
+
"reward": 1.5446429252624512,
|
| 973 |
+
"reward_std": 0.21720573492348194,
|
| 974 |
+
"rewards/accuracy_reward": 0.5680803582072258,
|
| 975 |
+
"rewards/format_reward": 0.9765624925494194,
|
| 976 |
+
"step": 69
|
| 977 |
+
},
|
| 978 |
+
{
|
| 979 |
+
"clip_ratio": 0.0,
|
| 980 |
+
"completion_length": 757.3337326049805,
|
| 981 |
+
"epoch": 0.7097591888466414,
|
| 982 |
+
"grad_norm": 0.08954399079084396,
|
| 983 |
+
"learning_rate": 8.137345036904259e-07,
|
| 984 |
+
"loss": 0.0439,
|
| 985 |
+
"num_tokens": 52197693.0,
|
| 986 |
+
"reward": 1.5390625596046448,
|
| 987 |
+
"reward_std": 0.25285689160227776,
|
| 988 |
+
"rewards/accuracy_reward": 0.5624999962747097,
|
| 989 |
+
"rewards/format_reward": 0.9765624925494194,
|
| 990 |
+
"step": 70
|
| 991 |
+
},
|
| 992 |
+
{
|
| 993 |
+
"clip_ratio": 0.0,
|
| 994 |
+
"completion_length": 731.8973541259766,
|
| 995 |
+
"epoch": 0.7198986058301647,
|
| 996 |
+
"grad_norm": 0.09004837274551392,
|
| 997 |
+
"learning_rate": 8.067355333682797e-07,
|
| 998 |
+
"loss": 0.0354,
|
| 999 |
+
"num_tokens": 52992241.0,
|
| 1000 |
+
"reward": 1.562500074505806,
|
| 1001 |
+
"reward_std": 0.22563916258513927,
|
| 1002 |
+
"rewards/accuracy_reward": 0.5814732126891613,
|
| 1003 |
+
"rewards/format_reward": 0.9810267761349678,
|
| 1004 |
+
"step": 71
|
| 1005 |
+
},
|
| 1006 |
+
{
|
| 1007 |
+
"clip_ratio": 0.0,
|
| 1008 |
+
"completion_length": 734.0893249511719,
|
| 1009 |
+
"epoch": 0.7300380228136882,
|
| 1010 |
+
"grad_norm": 0.0921608954668045,
|
| 1011 |
+
"learning_rate": 7.996388332556734e-07,
|
| 1012 |
+
"loss": 0.0553,
|
| 1013 |
+
"num_tokens": 53778705.0,
|
| 1014 |
+
"reward": 1.561383992433548,
|
| 1015 |
+
"reward_std": 0.25903933122754097,
|
| 1016 |
+
"rewards/accuracy_reward": 0.5792410857975483,
|
| 1017 |
+
"rewards/format_reward": 0.9821428582072258,
|
| 1018 |
+
"step": 72
|
| 1019 |
+
},
|
| 1020 |
+
{
|
| 1021 |
+
"clip_ratio": 0.0,
|
| 1022 |
+
"completion_length": 783.1897735595703,
|
| 1023 |
+
"epoch": 0.7401774397972116,
|
| 1024 |
+
"grad_norm": 0.09094219654798508,
|
| 1025 |
+
"learning_rate": 7.924466644503264e-07,
|
| 1026 |
+
"loss": 0.0356,
|
| 1027 |
+
"num_tokens": 54612939.0,
|
| 1028 |
+
"reward": 1.5558036416769028,
|
| 1029 |
+
"reward_std": 0.25537824630737305,
|
| 1030 |
+
"rewards/accuracy_reward": 0.5691964365541935,
|
| 1031 |
+
"rewards/format_reward": 0.9866071343421936,
|
| 1032 |
+
"step": 73
|
| 1033 |
+
},
|
| 1034 |
+
{
|
| 1035 |
+
"clip_ratio": 0.0,
|
| 1036 |
+
"completion_length": 725.2801666259766,
|
| 1037 |
+
"epoch": 0.7503168567807351,
|
| 1038 |
+
"grad_norm": 0.09524378925561905,
|
| 1039 |
+
"learning_rate": 7.85161318467482e-07,
|
| 1040 |
+
"loss": 0.0229,
|
| 1041 |
+
"num_tokens": 55398174.0,
|
| 1042 |
+
"reward": 1.6205357909202576,
|
| 1043 |
+
"reward_std": 0.2613362595438957,
|
| 1044 |
+
"rewards/accuracy_reward": 0.6305803582072258,
|
| 1045 |
+
"rewards/format_reward": 0.9899553507566452,
|
| 1046 |
+
"step": 74
|
| 1047 |
+
},
|
| 1048 |
+
{
|
| 1049 |
+
"clip_ratio": 0.0,
|
| 1050 |
+
"completion_length": 710.1853103637695,
|
| 1051 |
+
"epoch": 0.7604562737642585,
|
| 1052 |
+
"grad_norm": 0.0883307084441185,
|
| 1053 |
+
"learning_rate": 7.777851165098011e-07,
|
| 1054 |
+
"loss": 0.0235,
|
| 1055 |
+
"num_tokens": 56165180.0,
|
| 1056 |
+
"reward": 1.6183036416769028,
|
| 1057 |
+
"reward_std": 0.21434970386326313,
|
| 1058 |
+
"rewards/accuracy_reward": 0.6305803619325161,
|
| 1059 |
+
"rewards/format_reward": 0.9877232015132904,
|
| 1060 |
+
"step": 75
|
| 1061 |
+
},
|
| 1062 |
+
{
|
| 1063 |
+
"clip_ratio": 0.0,
|
| 1064 |
+
"completion_length": 690.4687805175781,
|
| 1065 |
+
"epoch": 0.770595690747782,
|
| 1066 |
+
"grad_norm": 0.09709777683019638,
|
| 1067 |
+
"learning_rate": 7.703204087277988e-07,
|
| 1068 |
+
"loss": 0.0316,
|
| 1069 |
+
"num_tokens": 56911080.0,
|
| 1070 |
+
"reward": 1.5267857760190964,
|
| 1071 |
+
"reward_std": 0.23126200772821903,
|
| 1072 |
+
"rewards/accuracy_reward": 0.5401785671710968,
|
| 1073 |
+
"rewards/format_reward": 0.9866071492433548,
|
| 1074 |
+
"step": 76
|
| 1075 |
+
},
|
| 1076 |
+
{
|
| 1077 |
+
"clip_ratio": 0.0,
|
| 1078 |
+
"completion_length": 753.3359756469727,
|
| 1079 |
+
"epoch": 0.7807351077313055,
|
| 1080 |
+
"grad_norm": 0.08559317141771317,
|
| 1081 |
+
"learning_rate": 7.627695734710564e-07,
|
| 1082 |
+
"loss": 0.0362,
|
| 1083 |
+
"num_tokens": 57708293.0,
|
| 1084 |
+
"reward": 1.601562574505806,
|
| 1085 |
+
"reward_std": 0.22745812311768532,
|
| 1086 |
+
"rewards/accuracy_reward": 0.6216517761349678,
|
| 1087 |
+
"rewards/format_reward": 0.979910708963871,
|
| 1088 |
+
"step": 77
|
| 1089 |
+
},
|
| 1090 |
+
{
|
| 1091 |
+
"clip_ratio": 0.0,
|
| 1092 |
+
"completion_length": 771.0870895385742,
|
| 1093 |
+
"epoch": 0.7908745247148289,
|
| 1094 |
+
"grad_norm": 0.08436188846826553,
|
| 1095 |
+
"learning_rate": 7.551350165304499e-07,
|
| 1096 |
+
"loss": 0.0295,
|
| 1097 |
+
"num_tokens": 58516963.0,
|
| 1098 |
+
"reward": 1.6361608058214188,
|
| 1099 |
+
"reward_std": 0.22785574942827225,
|
| 1100 |
+
"rewards/accuracy_reward": 0.6540178693830967,
|
| 1101 |
+
"rewards/format_reward": 0.9821428582072258,
|
| 1102 |
+
"step": 78
|
| 1103 |
+
},
|
| 1104 |
+
{
|
| 1105 |
+
"clip_ratio": 0.0,
|
| 1106 |
+
"completion_length": 738.5379791259766,
|
| 1107 |
+
"epoch": 0.8010139416983524,
|
| 1108 |
+
"grad_norm": 0.08563719689846039,
|
| 1109 |
+
"learning_rate": 7.474191703716338e-07,
|
| 1110 |
+
"loss": 0.0205,
|
| 1111 |
+
"num_tokens": 59309253.0,
|
| 1112 |
+
"reward": 1.6015625894069672,
|
| 1113 |
+
"reward_std": 0.211506649851799,
|
| 1114 |
+
"rewards/accuracy_reward": 0.6071428544819355,
|
| 1115 |
+
"rewards/format_reward": 0.9944196417927742,
|
| 1116 |
+
"step": 79
|
| 1117 |
+
},
|
| 1118 |
+
{
|
| 1119 |
+
"clip_ratio": 0.0,
|
| 1120 |
+
"completion_length": 729.372802734375,
|
| 1121 |
+
"epoch": 0.8111533586818758,
|
| 1122 |
+
"grad_norm": 0.09525574743747711,
|
| 1123 |
+
"learning_rate": 7.396244933600284e-07,
|
| 1124 |
+
"loss": 0.0426,
|
| 1125 |
+
"num_tokens": 60087059.0,
|
| 1126 |
+
"reward": 1.6004464775323868,
|
| 1127 |
+
"reward_std": 0.24874755926430225,
|
| 1128 |
+
"rewards/accuracy_reward": 0.608258917927742,
|
| 1129 |
+
"rewards/format_reward": 0.9921874925494194,
|
| 1130 |
+
"step": 80
|
| 1131 |
+
},
|
| 1132 |
+
{
|
| 1133 |
+
"clip_ratio": 0.0,
|
| 1134 |
+
"completion_length": 742.4721298217773,
|
| 1135 |
+
"epoch": 0.8212927756653993,
|
| 1136 |
+
"grad_norm": 0.0932924821972847,
|
| 1137 |
+
"learning_rate": 7.317534689775527e-07,
|
| 1138 |
+
"loss": 0.035,
|
| 1139 |
+
"num_tokens": 60872018.0,
|
| 1140 |
+
"reward": 1.5714286267757416,
|
| 1141 |
+
"reward_std": 0.25846732780337334,
|
| 1142 |
+
"rewards/accuracy_reward": 0.5881696417927742,
|
| 1143 |
+
"rewards/format_reward": 0.9832589253783226,
|
| 1144 |
+
"step": 81
|
| 1145 |
+
},
|
| 1146 |
+
{
|
| 1147 |
+
"clip_ratio": 0.0,
|
| 1148 |
+
"completion_length": 709.4643173217773,
|
| 1149 |
+
"epoch": 0.8314321926489227,
|
| 1150 |
+
"grad_norm": 0.09472807496786118,
|
| 1151 |
+
"learning_rate": 7.238086050313562e-07,
|
| 1152 |
+
"loss": 0.0341,
|
| 1153 |
+
"num_tokens": 61642954.0,
|
| 1154 |
+
"reward": 1.5870536416769028,
|
| 1155 |
+
"reward_std": 0.23454621247947216,
|
| 1156 |
+
"rewards/accuracy_reward": 0.5993303582072258,
|
| 1157 |
+
"rewards/format_reward": 0.9877232015132904,
|
| 1158 |
+
"step": 82
|
| 1159 |
+
},
|
| 1160 |
+
{
|
| 1161 |
+
"clip_ratio": 0.0,
|
| 1162 |
+
"completion_length": 716.0413208007812,
|
| 1163 |
+
"epoch": 0.8415716096324461,
|
| 1164 |
+
"grad_norm": 0.10518212616443634,
|
| 1165 |
+
"learning_rate": 7.157924328548002e-07,
|
| 1166 |
+
"loss": 0.045,
|
| 1167 |
+
"num_tokens": 62418119.0,
|
| 1168 |
+
"reward": 1.6026786416769028,
|
| 1169 |
+
"reward_std": 0.2639338057488203,
|
| 1170 |
+
"rewards/accuracy_reward": 0.6160714328289032,
|
| 1171 |
+
"rewards/format_reward": 0.9866071343421936,
|
| 1172 |
+
"step": 83
|
| 1173 |
+
},
|
| 1174 |
+
{
|
| 1175 |
+
"clip_ratio": 0.0,
|
| 1176 |
+
"completion_length": 724.6919937133789,
|
| 1177 |
+
"epoch": 0.8517110266159695,
|
| 1178 |
+
"grad_norm": 0.0993557870388031,
|
| 1179 |
+
"learning_rate": 7.077075065009433e-07,
|
| 1180 |
+
"loss": 0.0374,
|
| 1181 |
+
"num_tokens": 63200755.0,
|
| 1182 |
+
"reward": 1.649553656578064,
|
| 1183 |
+
"reward_std": 0.2632820960134268,
|
| 1184 |
+
"rewards/accuracy_reward": 0.659598208963871,
|
| 1185 |
+
"rewards/format_reward": 0.9899553507566452,
|
| 1186 |
+
"step": 84
|
| 1187 |
+
},
|
| 1188 |
+
{
|
| 1189 |
+
"clip_ratio": 0.0,
|
| 1190 |
+
"completion_length": 694.9643096923828,
|
| 1191 |
+
"epoch": 0.861850443599493,
|
| 1192 |
+
"grad_norm": 0.0879211500287056,
|
| 1193 |
+
"learning_rate": 6.995564019287869e-07,
|
| 1194 |
+
"loss": 0.0205,
|
| 1195 |
+
"num_tokens": 63953259.0,
|
| 1196 |
+
"reward": 1.5792411267757416,
|
| 1197 |
+
"reward_std": 0.19979050569236279,
|
| 1198 |
+
"rewards/accuracy_reward": 0.5859375037252903,
|
| 1199 |
+
"rewards/format_reward": 0.9933035671710968,
|
| 1200 |
+
"step": 85
|
| 1201 |
+
},
|
| 1202 |
+
{
|
| 1203 |
+
"clip_ratio": 0.0,
|
| 1204 |
+
"completion_length": 764.1395416259766,
|
| 1205 |
+
"epoch": 0.8719898605830165,
|
| 1206 |
+
"grad_norm": 0.3273586332798004,
|
| 1207 |
+
"learning_rate": 6.913417161825449e-07,
|
| 1208 |
+
"loss": 0.027,
|
| 1209 |
+
"num_tokens": 64772840.0,
|
| 1210 |
+
"reward": 1.5647322237491608,
|
| 1211 |
+
"reward_std": 0.23416299000382423,
|
| 1212 |
+
"rewards/accuracy_reward": 0.5758928582072258,
|
| 1213 |
+
"rewards/format_reward": 0.9888392686843872,
|
| 1214 |
+
"step": 86
|
| 1215 |
+
},
|
| 1216 |
+
{
|
| 1217 |
+
"clip_ratio": 0.0,
|
| 1218 |
+
"completion_length": 723.0658798217773,
|
| 1219 |
+
"epoch": 0.8821292775665399,
|
| 1220 |
+
"grad_norm": 0.231331005692482,
|
| 1221 |
+
"learning_rate": 6.830660665641897e-07,
|
| 1222 |
+
"loss": 0.0197,
|
| 1223 |
+
"num_tokens": 65550427.0,
|
| 1224 |
+
"reward": 1.6183036714792252,
|
| 1225 |
+
"reward_std": 0.2430772576481104,
|
| 1226 |
+
"rewards/accuracy_reward": 0.6272321380674839,
|
| 1227 |
+
"rewards/format_reward": 0.9910714253783226,
|
| 1228 |
+
"step": 87
|
| 1229 |
+
},
|
| 1230 |
+
{
|
| 1231 |
+
"clip_ratio": 0.0,
|
| 1232 |
+
"completion_length": 756.286865234375,
|
| 1233 |
+
"epoch": 0.8922686945500634,
|
| 1234 |
+
"grad_norm": 0.09298628568649292,
|
| 1235 |
+
"learning_rate": 6.747320897995492e-07,
|
| 1236 |
+
"loss": 0.0426,
|
| 1237 |
+
"num_tokens": 66357052.0,
|
| 1238 |
+
"reward": 1.5915179252624512,
|
| 1239 |
+
"reward_std": 0.2366574928164482,
|
| 1240 |
+
"rewards/accuracy_reward": 0.6149553656578064,
|
| 1241 |
+
"rewards/format_reward": 0.9765625,
|
| 1242 |
+
"step": 88
|
| 1243 |
+
},
|
| 1244 |
+
{
|
| 1245 |
+
"clip_ratio": 0.0,
|
| 1246 |
+
"completion_length": 686.9029312133789,
|
| 1247 |
+
"epoch": 0.9024081115335868,
|
| 1248 |
+
"grad_norm": 0.0874035432934761,
|
| 1249 |
+
"learning_rate": 6.66342441198212e-07,
|
| 1250 |
+
"loss": 0.015,
|
| 1251 |
+
"num_tokens": 67096421.0,
|
| 1252 |
+
"reward": 1.6127232909202576,
|
| 1253 |
+
"reward_std": 0.1957838051021099,
|
| 1254 |
+
"rewards/accuracy_reward": 0.6205357126891613,
|
| 1255 |
+
"rewards/format_reward": 0.9921874925494194,
|
| 1256 |
+
"step": 89
|
| 1257 |
+
},
|
| 1258 |
+
{
|
| 1259 |
+
"clip_ratio": 0.0,
|
| 1260 |
+
"completion_length": 701.6573944091797,
|
| 1261 |
+
"epoch": 0.9125475285171103,
|
| 1262 |
+
"grad_norm": 0.09878430515527725,
|
| 1263 |
+
"learning_rate": 6.578997938075125e-07,
|
| 1264 |
+
"loss": 0.0477,
|
| 1265 |
+
"num_tokens": 67849578.0,
|
| 1266 |
+
"reward": 1.5691965073347092,
|
| 1267 |
+
"reward_std": 0.2424814011901617,
|
| 1268 |
+
"rewards/accuracy_reward": 0.5881696492433548,
|
| 1269 |
+
"rewards/format_reward": 0.9810267835855484,
|
| 1270 |
+
"step": 90
|
| 1271 |
+
},
|
| 1272 |
+
{
|
| 1273 |
+
"clip_ratio": 0.0,
|
| 1274 |
+
"completion_length": 717.0368499755859,
|
| 1275 |
+
"epoch": 0.9226869455006337,
|
| 1276 |
+
"grad_norm": 0.0928407534956932,
|
| 1277 |
+
"learning_rate": 6.494068375608646e-07,
|
| 1278 |
+
"loss": 0.0515,
|
| 1279 |
+
"num_tokens": 68626667.0,
|
| 1280 |
+
"reward": 1.5803571939468384,
|
| 1281 |
+
"reward_std": 0.2204410433769226,
|
| 1282 |
+
"rewards/accuracy_reward": 0.5993303582072258,
|
| 1283 |
+
"rewards/format_reward": 0.9810267835855484,
|
| 1284 |
+
"step": 91
|
| 1285 |
+
},
|
| 1286 |
+
{
|
| 1287 |
+
"clip_ratio": 0.0,
|
| 1288 |
+
"completion_length": 750.7065048217773,
|
| 1289 |
+
"epoch": 0.9328263624841572,
|
| 1290 |
+
"grad_norm": 0.09953666478395462,
|
| 1291 |
+
"learning_rate": 6.408662784207149e-07,
|
| 1292 |
+
"loss": 0.0434,
|
| 1293 |
+
"num_tokens": 69438676.0,
|
| 1294 |
+
"reward": 1.6037947237491608,
|
| 1295 |
+
"reward_std": 0.263325035572052,
|
| 1296 |
+
"rewards/accuracy_reward": 0.621651791036129,
|
| 1297 |
+
"rewards/format_reward": 0.9821428507566452,
|
| 1298 |
+
"step": 92
|
| 1299 |
+
},
|
| 1300 |
+
{
|
| 1301 |
+
"clip_ratio": 0.0,
|
| 1302 |
+
"completion_length": 673.391773223877,
|
| 1303 |
+
"epoch": 0.9429657794676806,
|
| 1304 |
+
"grad_norm": 0.09401652216911316,
|
| 1305 |
+
"learning_rate": 6.322808375163895e-07,
|
| 1306 |
+
"loss": 0.0215,
|
| 1307 |
+
"num_tokens": 70168915.0,
|
| 1308 |
+
"reward": 1.6227679252624512,
|
| 1309 |
+
"reward_std": 0.1998270694166422,
|
| 1310 |
+
"rewards/accuracy_reward": 0.629464291036129,
|
| 1311 |
+
"rewards/format_reward": 0.9933035746216774,
|
| 1312 |
+
"step": 93
|
| 1313 |
+
},
|
| 1314 |
+
{
|
| 1315 |
+
"clip_ratio": 0.0,
|
| 1316 |
+
"completion_length": 706.0636520385742,
|
| 1317 |
+
"epoch": 0.9531051964512041,
|
| 1318 |
+
"grad_norm": 0.10031445324420929,
|
| 1319 |
+
"learning_rate": 6.236532502771077e-07,
|
| 1320 |
+
"loss": 0.029,
|
| 1321 |
+
"num_tokens": 70935148.0,
|
| 1322 |
+
"reward": 1.600446492433548,
|
| 1323 |
+
"reward_std": 0.2434455920010805,
|
| 1324 |
+
"rewards/accuracy_reward": 0.6104910634458065,
|
| 1325 |
+
"rewards/format_reward": 0.9899553582072258,
|
| 1326 |
+
"step": 94
|
| 1327 |
+
},
|
| 1328 |
+
{
|
| 1329 |
+
"clip_ratio": 0.0,
|
| 1330 |
+
"completion_length": 702.7511444091797,
|
| 1331 |
+
"epoch": 0.9632446134347274,
|
| 1332 |
+
"grad_norm": 0.09827374666929245,
|
| 1333 |
+
"learning_rate": 6.149862655604403e-07,
|
| 1334 |
+
"loss": 0.0358,
|
| 1335 |
+
"num_tokens": 71690501.0,
|
| 1336 |
+
"reward": 1.5982143431901932,
|
| 1337 |
+
"reward_std": 0.2399199977517128,
|
| 1338 |
+
"rewards/accuracy_reward": 0.6149553507566452,
|
| 1339 |
+
"rewards/format_reward": 0.9832589253783226,
|
| 1340 |
+
"step": 95
|
| 1341 |
+
},
|
| 1342 |
+
{
|
| 1343 |
+
"clip_ratio": 0.0,
|
| 1344 |
+
"completion_length": 713.8114166259766,
|
| 1345 |
+
"epoch": 0.973384030418251,
|
| 1346 |
+
"grad_norm": 0.08831395208835602,
|
| 1347 |
+
"learning_rate": 6.062826447764883e-07,
|
| 1348 |
+
"loss": 0.031,
|
| 1349 |
+
"num_tokens": 72452820.0,
|
| 1350 |
+
"reward": 1.6127232760190964,
|
| 1351 |
+
"reward_std": 0.2217018250375986,
|
| 1352 |
+
"rewards/accuracy_reward": 0.6238839253783226,
|
| 1353 |
+
"rewards/format_reward": 0.9888392761349678,
|
| 1354 |
+
"step": 96
|
| 1355 |
+
},
|
| 1356 |
+
{
|
| 1357 |
+
"clip_ratio": 0.0,
|
| 1358 |
+
"completion_length": 701.4710083007812,
|
| 1359 |
+
"epoch": 0.9835234474017744,
|
| 1360 |
+
"grad_norm": 0.12909561395645142,
|
| 1361 |
+
"learning_rate": 5.975451610080642e-07,
|
| 1362 |
+
"loss": 0.0299,
|
| 1363 |
+
"num_tokens": 73209146.0,
|
| 1364 |
+
"reward": 1.5602679252624512,
|
| 1365 |
+
"reward_std": 0.21769515611231327,
|
| 1366 |
+
"rewards/accuracy_reward": 0.5680803507566452,
|
| 1367 |
+
"rewards/format_reward": 0.9921874925494194,
|
| 1368 |
+
"step": 97
|
| 1369 |
+
},
|
| 1370 |
+
{
|
| 1371 |
+
"clip_ratio": 0.0,
|
| 1372 |
+
"completion_length": 698.564453125,
|
| 1373 |
+
"epoch": 0.9936628643852978,
|
| 1374 |
+
"grad_norm": 0.09522398561239243,
|
| 1375 |
+
"learning_rate": 5.887765981271517e-07,
|
| 1376 |
+
"loss": 0.0337,
|
| 1377 |
+
"num_tokens": 73991697.0,
|
| 1378 |
+
"reward": 1.56808041036129,
|
| 1379 |
+
"reward_std": 0.24109553173184395,
|
| 1380 |
+
"rewards/accuracy_reward": 0.5792410746216774,
|
| 1381 |
+
"rewards/format_reward": 0.9888392761349678,
|
| 1382 |
+
"step": 98
|
| 1383 |
+
},
|
| 1384 |
+
{
|
| 1385 |
+
"clip_ratio": 0.0,
|
| 1386 |
+
"completion_length": 719.4676666259766,
|
| 1387 |
+
"epoch": 1.0101394169835234,
|
| 1388 |
+
"grad_norm": 0.09128111600875854,
|
| 1389 |
+
"learning_rate": 5.7997974990793e-07,
|
| 1390 |
+
"loss": 0.0236,
|
| 1391 |
+
"num_tokens": 74765332.0,
|
| 1392 |
+
"reward": 1.648437574505806,
|
| 1393 |
+
"reward_std": 0.22949284128844738,
|
| 1394 |
+
"rewards/accuracy_reward": 0.65625,
|
| 1395 |
+
"rewards/format_reward": 0.9921874925494194,
|
| 1396 |
+
"step": 99
|
| 1397 |
+
},
|
| 1398 |
+
{
|
| 1399 |
+
"clip_ratio": 0.0,
|
| 1400 |
+
"completion_length": 720.794677734375,
|
| 1401 |
+
"epoch": 1.020278833967047,
|
| 1402 |
+
"grad_norm": 0.09661993384361267,
|
| 1403 |
+
"learning_rate": 5.711574191366427e-07,
|
| 1404 |
+
"loss": 0.0317,
|
| 1405 |
+
"num_tokens": 75538540.0,
|
| 1406 |
+
"reward": 1.617187574505806,
|
| 1407 |
+
"reward_std": 0.21921591460704803,
|
| 1408 |
+
"rewards/accuracy_reward": 0.6316964328289032,
|
| 1409 |
+
"rewards/format_reward": 0.9854910671710968,
|
| 1410 |
+
"step": 100
|
| 1411 |
+
},
|
| 1412 |
+
{
|
| 1413 |
+
"clip_ratio": 0.0,
|
| 1414 |
+
"completion_length": 715.5815048217773,
|
| 1415 |
+
"epoch": 1.0304182509505704,
|
| 1416 |
+
"grad_norm": 0.09119318425655365,
|
| 1417 |
+
"learning_rate": 5.623124167185929e-07,
|
| 1418 |
+
"loss": 0.0238,
|
| 1419 |
+
"num_tokens": 76302229.0,
|
| 1420 |
+
"reward": 1.6149554401636124,
|
| 1421 |
+
"reward_std": 0.23784785345196724,
|
| 1422 |
+
"rewards/accuracy_reward": 0.6328125,
|
| 1423 |
+
"rewards/format_reward": 0.9821428507566452,
|
| 1424 |
+
"step": 101
|
| 1425 |
+
},
|
| 1426 |
+
{
|
| 1427 |
+
"clip_ratio": 0.0,
|
| 1428 |
+
"completion_length": 671.747802734375,
|
| 1429 |
+
"epoch": 1.0405576679340938,
|
| 1430 |
+
"grad_norm": 0.0963052362203598,
|
| 1431 |
+
"learning_rate": 5.534475607825565e-07,
|
| 1432 |
+
"loss": 0.0087,
|
| 1433 |
+
"num_tokens": 77031403.0,
|
| 1434 |
+
"reward": 1.6227679252624512,
|
| 1435 |
+
"reward_std": 0.22612315602600574,
|
| 1436 |
+
"rewards/accuracy_reward": 0.6272321417927742,
|
| 1437 |
+
"rewards/format_reward": 0.9955357015132904,
|
| 1438 |
+
"step": 102
|
| 1439 |
+
},
|
| 1440 |
+
{
|
| 1441 |
+
"clip_ratio": 0.0,
|
| 1442 |
+
"completion_length": 736.8772659301758,
|
| 1443 |
+
"epoch": 1.0506970849176172,
|
| 1444 |
+
"grad_norm": 0.09166496247053146,
|
| 1445 |
+
"learning_rate": 5.445656757828879e-07,
|
| 1446 |
+
"loss": 0.0313,
|
| 1447 |
+
"num_tokens": 77822949.0,
|
| 1448 |
+
"reward": 1.5044643431901932,
|
| 1449 |
+
"reward_std": 0.22669786028563976,
|
| 1450 |
+
"rewards/accuracy_reward": 0.5189732238650322,
|
| 1451 |
+
"rewards/format_reward": 0.9854910746216774,
|
| 1452 |
+
"step": 103
|
| 1453 |
+
},
|
| 1454 |
+
{
|
| 1455 |
+
"clip_ratio": 0.0,
|
| 1456 |
+
"completion_length": 719.7199020385742,
|
| 1457 |
+
"epoch": 1.0608365019011408,
|
| 1458 |
+
"grad_norm": 0.09398168325424194,
|
| 1459 |
+
"learning_rate": 5.356695915996161e-07,
|
| 1460 |
+
"loss": 0.0265,
|
| 1461 |
+
"num_tokens": 78602178.0,
|
| 1462 |
+
"reward": 1.5602679252624512,
|
| 1463 |
+
"reward_std": 0.25012633949518204,
|
| 1464 |
+
"rewards/accuracy_reward": 0.5691964328289032,
|
| 1465 |
+
"rewards/format_reward": 0.9910714253783226,
|
| 1466 |
+
"step": 104
|
| 1467 |
+
},
|
| 1468 |
+
{
|
| 1469 |
+
"clip_ratio": 0.0,
|
| 1470 |
+
"completion_length": 688.8962249755859,
|
| 1471 |
+
"epoch": 1.0709759188846641,
|
| 1472 |
+
"grad_norm": 0.11429024487733841,
|
| 1473 |
+
"learning_rate": 5.267621426368075e-07,
|
| 1474 |
+
"loss": 0.0246,
|
| 1475 |
+
"num_tokens": 79338645.0,
|
| 1476 |
+
"reward": 1.6383929550647736,
|
| 1477 |
+
"reward_std": 0.266002238728106,
|
| 1478 |
+
"rewards/accuracy_reward": 0.6495535746216774,
|
| 1479 |
+
"rewards/format_reward": 0.9888392761349678,
|
| 1480 |
+
"step": 105
|
| 1481 |
+
},
|
| 1482 |
+
{
|
| 1483 |
+
"clip_ratio": 0.0,
|
| 1484 |
+
"completion_length": 645.3884124755859,
|
| 1485 |
+
"epoch": 1.0811153358681875,
|
| 1486 |
+
"grad_norm": 0.0960325375199318,
|
| 1487 |
+
"learning_rate": 5.178461669194903e-07,
|
| 1488 |
+
"loss": 0.0291,
|
| 1489 |
+
"num_tokens": 80044121.0,
|
| 1490 |
+
"reward": 1.670758992433548,
|
| 1491 |
+
"reward_std": 0.21290560998022556,
|
| 1492 |
+
"rewards/accuracy_reward": 0.6808035671710968,
|
| 1493 |
+
"rewards/format_reward": 0.9899553582072258,
|
| 1494 |
+
"step": 106
|
| 1495 |
+
},
|
| 1496 |
+
{
|
| 1497 |
+
"clip_ratio": 0.0,
|
| 1498 |
+
"completion_length": 682.0167694091797,
|
| 1499 |
+
"epoch": 1.091254752851711,
|
| 1500 |
+
"grad_norm": 0.08784578740596771,
|
| 1501 |
+
"learning_rate": 5.08924505189423e-07,
|
| 1502 |
+
"loss": 0.0275,
|
| 1503 |
+
"num_tokens": 80781168.0,
|
| 1504 |
+
"reward": 1.649553656578064,
|
| 1505 |
+
"reward_std": 0.1967218518257141,
|
| 1506 |
+
"rewards/accuracy_reward": 0.65625,
|
| 1507 |
+
"rewards/format_reward": 0.9933035597205162,
|
| 1508 |
+
"step": 107
|
| 1509 |
+
},
|
| 1510 |
+
{
|
| 1511 |
+
"clip_ratio": 0.0,
|
| 1512 |
+
"completion_length": 718.1272659301758,
|
| 1513 |
+
"epoch": 1.1013941698352345,
|
| 1514 |
+
"grad_norm": 0.08991685509681702,
|
| 1515 |
+
"learning_rate": 5e-07,
|
| 1516 |
+
"loss": 0.0394,
|
| 1517 |
+
"num_tokens": 81548730.0,
|
| 1518 |
+
"reward": 1.626116156578064,
|
| 1519 |
+
"reward_std": 0.24366096779704094,
|
| 1520 |
+
"rewards/accuracy_reward": 0.6450892835855484,
|
| 1521 |
+
"rewards/format_reward": 0.981026791036129,
|
| 1522 |
+
"step": 108
|
| 1523 |
+
},
|
| 1524 |
+
{
|
| 1525 |
+
"clip_ratio": 0.0,
|
| 1526 |
+
"completion_length": 746.9297256469727,
|
| 1527 |
+
"epoch": 1.111533586818758,
|
| 1528 |
+
"grad_norm": 0.10698525607585907,
|
| 1529 |
+
"learning_rate": 4.91075494810577e-07,
|
| 1530 |
+
"loss": 0.0338,
|
| 1531 |
+
"num_tokens": 82335971.0,
|
| 1532 |
+
"reward": 1.6015625596046448,
|
| 1533 |
+
"reward_std": 0.2378039713948965,
|
| 1534 |
+
"rewards/accuracy_reward": 0.6160714253783226,
|
| 1535 |
+
"rewards/format_reward": 0.9854910671710968,
|
| 1536 |
+
"step": 109
|
| 1537 |
+
},
|
| 1538 |
+
{
|
| 1539 |
+
"clip_ratio": 0.0,
|
| 1540 |
+
"completion_length": 744.1752624511719,
|
| 1541 |
+
"epoch": 1.1216730038022813,
|
| 1542 |
+
"grad_norm": 0.0870419293642044,
|
| 1543 |
+
"learning_rate": 4.821538330805098e-07,
|
| 1544 |
+
"loss": 0.0489,
|
| 1545 |
+
"num_tokens": 83129272.0,
|
| 1546 |
+
"reward": 1.5982143580913544,
|
| 1547 |
+
"reward_std": 0.22117138653993607,
|
| 1548 |
+
"rewards/accuracy_reward": 0.6183035746216774,
|
| 1549 |
+
"rewards/format_reward": 0.979910708963871,
|
| 1550 |
+
"step": 110
|
| 1551 |
+
},
|
| 1552 |
+
{
|
| 1553 |
+
"clip_ratio": 0.0,
|
| 1554 |
+
"completion_length": 729.8170013427734,
|
| 1555 |
+
"epoch": 1.131812420785805,
|
| 1556 |
+
"grad_norm": 0.09355349838733673,
|
| 1557 |
+
"learning_rate": 4.732378573631924e-07,
|
| 1558 |
+
"loss": 0.0372,
|
| 1559 |
+
"num_tokens": 83905916.0,
|
| 1560 |
+
"reward": 1.6004465073347092,
|
| 1561 |
+
"reward_std": 0.24993346445262432,
|
| 1562 |
+
"rewards/accuracy_reward": 0.609375,
|
| 1563 |
+
"rewards/format_reward": 0.9910714253783226,
|
| 1564 |
+
"step": 111
|
| 1565 |
+
},
|
| 1566 |
+
{
|
| 1567 |
+
"clip_ratio": 0.0,
|
| 1568 |
+
"completion_length": 730.1663208007812,
|
| 1569 |
+
"epoch": 1.1419518377693283,
|
| 1570 |
+
"grad_norm": 0.2215433269739151,
|
| 1571 |
+
"learning_rate": 4.643304084003838e-07,
|
| 1572 |
+
"loss": 0.025,
|
| 1573 |
+
"num_tokens": 84687593.0,
|
| 1574 |
+
"reward": 1.5837054252624512,
|
| 1575 |
+
"reward_std": 0.23716635070741177,
|
| 1576 |
+
"rewards/accuracy_reward": 0.594866082072258,
|
| 1577 |
+
"rewards/format_reward": 0.9888392686843872,
|
| 1578 |
+
"step": 112
|
| 1579 |
+
},
|
| 1580 |
+
{
|
| 1581 |
+
"clip_ratio": 0.0,
|
| 1582 |
+
"completion_length": 704.9989242553711,
|
| 1583 |
+
"epoch": 1.1520912547528517,
|
| 1584 |
+
"grad_norm": 0.10191125422716141,
|
| 1585 |
+
"learning_rate": 4.55434324217112e-07,
|
| 1586 |
+
"loss": 0.0477,
|
| 1587 |
+
"num_tokens": 85448512.0,
|
| 1588 |
+
"reward": 1.5513393580913544,
|
| 1589 |
+
"reward_std": 0.24850638210773468,
|
| 1590 |
+
"rewards/accuracy_reward": 0.5691964291036129,
|
| 1591 |
+
"rewards/format_reward": 0.9821428433060646,
|
| 1592 |
+
"step": 113
|
| 1593 |
+
},
|
| 1594 |
+
{
|
| 1595 |
+
"clip_ratio": 0.0,
|
| 1596 |
+
"completion_length": 696.9486846923828,
|
| 1597 |
+
"epoch": 1.162230671736375,
|
| 1598 |
+
"grad_norm": 0.08355645090341568,
|
| 1599 |
+
"learning_rate": 4.4655243921744367e-07,
|
| 1600 |
+
"loss": 0.0175,
|
| 1601 |
+
"num_tokens": 86197938.0,
|
| 1602 |
+
"reward": 1.5982143729925156,
|
| 1603 |
+
"reward_std": 0.1789538525044918,
|
| 1604 |
+
"rewards/accuracy_reward": 0.6049107164144516,
|
| 1605 |
+
"rewards/format_reward": 0.9933035671710968,
|
| 1606 |
+
"step": 114
|
| 1607 |
+
},
|
| 1608 |
+
{
|
| 1609 |
+
"clip_ratio": 0.0,
|
| 1610 |
+
"completion_length": 697.7143096923828,
|
| 1611 |
+
"epoch": 1.1723700887198987,
|
| 1612 |
+
"grad_norm": 0.08945267647504807,
|
| 1613 |
+
"learning_rate": 4.37687583281407e-07,
|
| 1614 |
+
"loss": 0.0261,
|
| 1615 |
+
"num_tokens": 86950306.0,
|
| 1616 |
+
"reward": 1.5982143580913544,
|
| 1617 |
+
"reward_std": 0.19760245829820633,
|
| 1618 |
+
"rewards/accuracy_reward": 0.6071428619325161,
|
| 1619 |
+
"rewards/format_reward": 0.9910714253783226,
|
| 1620 |
+
"step": 115
|
| 1621 |
+
},
|
| 1622 |
+
{
|
| 1623 |
+
"clip_ratio": 0.0,
|
| 1624 |
+
"completion_length": 681.9843978881836,
|
| 1625 |
+
"epoch": 1.182509505703422,
|
| 1626 |
+
"grad_norm": 0.10237658023834229,
|
| 1627 |
+
"learning_rate": 4.2884258086335745e-07,
|
| 1628 |
+
"loss": 0.017,
|
| 1629 |
+
"num_tokens": 87700604.0,
|
| 1630 |
+
"reward": 1.625000074505806,
|
| 1631 |
+
"reward_std": 0.23482281155884266,
|
| 1632 |
+
"rewards/accuracy_reward": 0.6328124962747097,
|
| 1633 |
+
"rewards/format_reward": 0.9921875,
|
| 1634 |
+
"step": 116
|
| 1635 |
+
},
|
| 1636 |
+
{
|
| 1637 |
+
"clip_ratio": 0.0,
|
| 1638 |
+
"completion_length": 679.974365234375,
|
| 1639 |
+
"epoch": 1.1926489226869454,
|
| 1640 |
+
"grad_norm": 0.09299944341182709,
|
| 1641 |
+
"learning_rate": 4.2002025009206987e-07,
|
| 1642 |
+
"loss": 0.0178,
|
| 1643 |
+
"num_tokens": 88441813.0,
|
| 1644 |
+
"reward": 1.6104911416769028,
|
| 1645 |
+
"reward_std": 0.19839700311422348,
|
| 1646 |
+
"rewards/accuracy_reward": 0.6171875037252903,
|
| 1647 |
+
"rewards/format_reward": 0.9933035597205162,
|
| 1648 |
+
"step": 117
|
| 1649 |
+
},
|
| 1650 |
+
{
|
| 1651 |
+
"clip_ratio": 0.0,
|
| 1652 |
+
"completion_length": 686.677490234375,
|
| 1653 |
+
"epoch": 1.202788339670469,
|
| 1654 |
+
"grad_norm": 0.16542084515094757,
|
| 1655 |
+
"learning_rate": 4.1122340187284845e-07,
|
| 1656 |
+
"loss": 0.0392,
|
| 1657 |
+
"num_tokens": 89179364.0,
|
| 1658 |
+
"reward": 1.609375074505806,
|
| 1659 |
+
"reward_std": 0.2622632719576359,
|
| 1660 |
+
"rewards/accuracy_reward": 0.6183035597205162,
|
| 1661 |
+
"rewards/format_reward": 0.991071417927742,
|
| 1662 |
+
"step": 118
|
| 1663 |
+
},
|
| 1664 |
+
{
|
| 1665 |
+
"clip_ratio": 0.0,
|
| 1666 |
+
"completion_length": 694.8125228881836,
|
| 1667 |
+
"epoch": 1.2129277566539924,
|
| 1668 |
+
"grad_norm": 0.11430428177118301,
|
| 1669 |
+
"learning_rate": 4.0245483899193586e-07,
|
| 1670 |
+
"loss": 0.0139,
|
| 1671 |
+
"num_tokens": 89929292.0,
|
| 1672 |
+
"reward": 1.6049107760190964,
|
| 1673 |
+
"reward_std": 0.22252802923321724,
|
| 1674 |
+
"rewards/accuracy_reward": 0.610491082072258,
|
| 1675 |
+
"rewards/format_reward": 0.9944196417927742,
|
| 1676 |
+
"step": 119
|
| 1677 |
+
},
|
| 1678 |
+
{
|
| 1679 |
+
"clip_ratio": 0.0,
|
| 1680 |
+
"completion_length": 684.7466812133789,
|
| 1681 |
+
"epoch": 1.2230671736375158,
|
| 1682 |
+
"grad_norm": 0.08012707531452179,
|
| 1683 |
+
"learning_rate": 3.937173552235116e-07,
|
| 1684 |
+
"loss": 0.0213,
|
| 1685 |
+
"num_tokens": 90675633.0,
|
| 1686 |
+
"reward": 1.5915179401636124,
|
| 1687 |
+
"reward_std": 0.16910855192691088,
|
| 1688 |
+
"rewards/accuracy_reward": 0.6037946492433548,
|
| 1689 |
+
"rewards/format_reward": 0.9877232164144516,
|
| 1690 |
+
"step": 120
|
| 1691 |
+
},
|
| 1692 |
+
{
|
| 1693 |
+
"clip_ratio": 0.0,
|
| 1694 |
+
"completion_length": 690.8917694091797,
|
| 1695 |
+
"epoch": 1.2332065906210392,
|
| 1696 |
+
"grad_norm": 0.10144107788801193,
|
| 1697 |
+
"learning_rate": 3.850137344395598e-07,
|
| 1698 |
+
"loss": 0.0521,
|
| 1699 |
+
"num_tokens": 91417888.0,
|
| 1700 |
+
"reward": 1.6227679550647736,
|
| 1701 |
+
"reward_std": 0.22748247906565666,
|
| 1702 |
+
"rewards/accuracy_reward": 0.6350446417927742,
|
| 1703 |
+
"rewards/format_reward": 0.9877232164144516,
|
| 1704 |
+
"step": 121
|
| 1705 |
+
},
|
| 1706 |
+
{
|
| 1707 |
+
"clip_ratio": 0.0,
|
| 1708 |
+
"completion_length": 754.5803909301758,
|
| 1709 |
+
"epoch": 1.2433460076045628,
|
| 1710 |
+
"grad_norm": 0.09234697371721268,
|
| 1711 |
+
"learning_rate": 3.763467497228922e-07,
|
| 1712 |
+
"loss": 0.0348,
|
| 1713 |
+
"num_tokens": 92232024.0,
|
| 1714 |
+
"reward": 1.5513393580913544,
|
| 1715 |
+
"reward_std": 0.22778411209583282,
|
| 1716 |
+
"rewards/accuracy_reward": 0.5725446529686451,
|
| 1717 |
+
"rewards/format_reward": 0.9787946343421936,
|
| 1718 |
+
"step": 122
|
| 1719 |
+
},
|
| 1720 |
+
{
|
| 1721 |
+
"clip_ratio": 0.0,
|
| 1722 |
+
"completion_length": 718.1328506469727,
|
| 1723 |
+
"epoch": 1.2534854245880862,
|
| 1724 |
+
"grad_norm": 0.09252262860536575,
|
| 1725 |
+
"learning_rate": 3.677191624836106e-07,
|
| 1726 |
+
"loss": 0.0375,
|
| 1727 |
+
"num_tokens": 92997183.0,
|
| 1728 |
+
"reward": 1.6417411267757416,
|
| 1729 |
+
"reward_std": 0.2288501374423504,
|
| 1730 |
+
"rewards/accuracy_reward": 0.651785708963871,
|
| 1731 |
+
"rewards/format_reward": 0.9899553433060646,
|
| 1732 |
+
"step": 123
|
| 1733 |
+
},
|
| 1734 |
+
{
|
| 1735 |
+
"clip_ratio": 0.0,
|
| 1736 |
+
"completion_length": 738.9018173217773,
|
| 1737 |
+
"epoch": 1.2636248415716096,
|
| 1738 |
+
"grad_norm": 0.09162338823080063,
|
| 1739 |
+
"learning_rate": 3.591337215792851e-07,
|
| 1740 |
+
"loss": 0.0228,
|
| 1741 |
+
"num_tokens": 93787407.0,
|
| 1742 |
+
"reward": 1.6037947237491608,
|
| 1743 |
+
"reward_std": 0.22520777396857738,
|
| 1744 |
+
"rewards/accuracy_reward": 0.6149553656578064,
|
| 1745 |
+
"rewards/format_reward": 0.9888392686843872,
|
| 1746 |
+
"step": 124
|
| 1747 |
+
},
|
| 1748 |
+
{
|
| 1749 |
+
"clip_ratio": 0.0,
|
| 1750 |
+
"completion_length": 709.0446853637695,
|
| 1751 |
+
"epoch": 1.2737642585551332,
|
| 1752 |
+
"grad_norm": 0.09890223294496536,
|
| 1753 |
+
"learning_rate": 3.505931624391355e-07,
|
| 1754 |
+
"loss": 0.0222,
|
| 1755 |
+
"num_tokens": 94540975.0,
|
| 1756 |
+
"reward": 1.6272322088479996,
|
| 1757 |
+
"reward_std": 0.25062085315585136,
|
| 1758 |
+
"rewards/accuracy_reward": 0.6316964328289032,
|
| 1759 |
+
"rewards/format_reward": 0.995535708963871,
|
| 1760 |
+
"step": 125
|
| 1761 |
+
},
|
| 1762 |
+
{
|
| 1763 |
+
"clip_ratio": 0.0,
|
| 1764 |
+
"completion_length": 695.0937805175781,
|
| 1765 |
+
"epoch": 1.2839036755386566,
|
| 1766 |
+
"grad_norm": 0.09244479238986969,
|
| 1767 |
+
"learning_rate": 3.421002061924876e-07,
|
| 1768 |
+
"loss": 0.0234,
|
| 1769 |
+
"num_tokens": 95290643.0,
|
| 1770 |
+
"reward": 1.6171875894069672,
|
| 1771 |
+
"reward_std": 0.23021192848682404,
|
| 1772 |
+
"rewards/accuracy_reward": 0.6350446417927742,
|
| 1773 |
+
"rewards/format_reward": 0.9821428507566452,
|
| 1774 |
+
"step": 126
|
| 1775 |
+
},
|
| 1776 |
+
{
|
| 1777 |
+
"clip_ratio": 0.0,
|
| 1778 |
+
"completion_length": 705.2265853881836,
|
| 1779 |
+
"epoch": 1.29404309252218,
|
| 1780 |
+
"grad_norm": 0.08838233351707458,
|
| 1781 |
+
"learning_rate": 3.3365755880178807e-07,
|
| 1782 |
+
"loss": 0.0274,
|
| 1783 |
+
"num_tokens": 96044230.0,
|
| 1784 |
+
"reward": 1.6517858058214188,
|
| 1785 |
+
"reward_std": 0.19274440594017506,
|
| 1786 |
+
"rewards/accuracy_reward": 0.6629464291036129,
|
| 1787 |
+
"rewards/format_reward": 0.9888392761349678,
|
| 1788 |
+
"step": 127
|
| 1789 |
+
},
|
| 1790 |
+
{
|
| 1791 |
+
"clip_ratio": 0.0,
|
| 1792 |
+
"completion_length": 730.522346496582,
|
| 1793 |
+
"epoch": 1.3041825095057034,
|
| 1794 |
+
"grad_norm": 0.10294985771179199,
|
| 1795 |
+
"learning_rate": 3.2526791020045087e-07,
|
| 1796 |
+
"loss": 0.0338,
|
| 1797 |
+
"num_tokens": 96831378.0,
|
| 1798 |
+
"reward": 1.5580357909202576,
|
| 1799 |
+
"reward_std": 0.2531749904155731,
|
| 1800 |
+
"rewards/accuracy_reward": 0.5758928544819355,
|
| 1801 |
+
"rewards/format_reward": 0.9821428507566452,
|
| 1802 |
+
"step": 128
|
| 1803 |
+
},
|
| 1804 |
+
{
|
| 1805 |
+
"clip_ratio": 0.0,
|
| 1806 |
+
"completion_length": 683.2656555175781,
|
| 1807 |
+
"epoch": 1.3143219264892267,
|
| 1808 |
+
"grad_norm": 0.09631213545799255,
|
| 1809 |
+
"learning_rate": 3.169339334358104e-07,
|
| 1810 |
+
"loss": 0.019,
|
| 1811 |
+
"num_tokens": 97576928.0,
|
| 1812 |
+
"reward": 1.6283482909202576,
|
| 1813 |
+
"reward_std": 0.22065912000834942,
|
| 1814 |
+
"rewards/accuracy_reward": 0.6372767835855484,
|
| 1815 |
+
"rewards/format_reward": 0.9910714328289032,
|
| 1816 |
+
"step": 129
|
| 1817 |
+
},
|
| 1818 |
+
{
|
| 1819 |
+
"clip_ratio": 0.0,
|
| 1820 |
+
"completion_length": 700.1573944091797,
|
| 1821 |
+
"epoch": 1.3244613434727504,
|
| 1822 |
+
"grad_norm": 0.09407830238342285,
|
| 1823 |
+
"learning_rate": 3.086582838174551e-07,
|
| 1824 |
+
"loss": 0.0196,
|
| 1825 |
+
"num_tokens": 98336757.0,
|
| 1826 |
+
"reward": 1.6116072237491608,
|
| 1827 |
+
"reward_std": 0.20910983718931675,
|
| 1828 |
+
"rewards/accuracy_reward": 0.6183035671710968,
|
| 1829 |
+
"rewards/format_reward": 0.9933035597205162,
|
| 1830 |
+
"step": 130
|
| 1831 |
+
},
|
| 1832 |
+
{
|
| 1833 |
+
"clip_ratio": 0.0,
|
| 1834 |
+
"completion_length": 694.3426666259766,
|
| 1835 |
+
"epoch": 1.3346007604562737,
|
| 1836 |
+
"grad_norm": 0.10380922257900238,
|
| 1837 |
+
"learning_rate": 3.004435980712129e-07,
|
| 1838 |
+
"loss": 0.0372,
|
| 1839 |
+
"num_tokens": 99082352.0,
|
| 1840 |
+
"reward": 1.5892857611179352,
|
| 1841 |
+
"reward_std": 0.23904000967741013,
|
| 1842 |
+
"rewards/accuracy_reward": 0.6026785783469677,
|
| 1843 |
+
"rewards/format_reward": 0.9866071343421936,
|
| 1844 |
+
"step": 131
|
| 1845 |
+
},
|
| 1846 |
+
{
|
| 1847 |
+
"clip_ratio": 0.0,
|
| 1848 |
+
"completion_length": 689.7589569091797,
|
| 1849 |
+
"epoch": 1.3447401774397973,
|
| 1850 |
+
"grad_norm": 0.08147790282964706,
|
| 1851 |
+
"learning_rate": 2.922924934990568e-07,
|
| 1852 |
+
"loss": 0.0384,
|
| 1853 |
+
"num_tokens": 99825528.0,
|
| 1854 |
+
"reward": 1.6372768580913544,
|
| 1855 |
+
"reward_std": 0.19569051638245583,
|
| 1856 |
+
"rewards/accuracy_reward": 0.645089291036129,
|
| 1857 |
+
"rewards/format_reward": 0.9921874925494194,
|
| 1858 |
+
"step": 132
|
| 1859 |
+
},
|
| 1860 |
+
{
|
| 1861 |
+
"clip_ratio": 0.0,
|
| 1862 |
+
"completion_length": 652.4018096923828,
|
| 1863 |
+
"epoch": 1.3548795944233207,
|
| 1864 |
+
"grad_norm": 0.10540962219238281,
|
| 1865 |
+
"learning_rate": 2.8420756714519954e-07,
|
| 1866 |
+
"loss": 0.0049,
|
| 1867 |
+
"num_tokens": 100554384.0,
|
| 1868 |
+
"reward": 1.6138393431901932,
|
| 1869 |
+
"reward_std": 0.2248824257403612,
|
| 1870 |
+
"rewards/accuracy_reward": 0.6183035708963871,
|
| 1871 |
+
"rewards/format_reward": 0.995535708963871,
|
| 1872 |
+
"step": 133
|
| 1873 |
+
},
|
| 1874 |
+
{
|
| 1875 |
+
"clip_ratio": 0.0,
|
| 1876 |
+
"completion_length": 691.8192138671875,
|
| 1877 |
+
"epoch": 1.3650190114068441,
|
| 1878 |
+
"grad_norm": 0.10231253504753113,
|
| 1879 |
+
"learning_rate": 2.7619139496864376e-07,
|
| 1880 |
+
"loss": 0.0299,
|
| 1881 |
+
"num_tokens": 101310238.0,
|
| 1882 |
+
"reward": 1.5345982760190964,
|
| 1883 |
+
"reward_std": 0.22884078323841095,
|
| 1884 |
+
"rewards/accuracy_reward": 0.543526791036129,
|
| 1885 |
+
"rewards/format_reward": 0.991071417927742,
|
| 1886 |
+
"step": 134
|
| 1887 |
+
},
|
| 1888 |
+
{
|
| 1889 |
+
"clip_ratio": 0.0,
|
| 1890 |
+
"completion_length": 654.686408996582,
|
| 1891 |
+
"epoch": 1.3751584283903675,
|
| 1892 |
+
"grad_norm": 0.10083113610744476,
|
| 1893 |
+
"learning_rate": 2.6824653102244727e-07,
|
| 1894 |
+
"loss": 0.0201,
|
| 1895 |
+
"num_tokens": 102030909.0,
|
| 1896 |
+
"reward": 1.6171875596046448,
|
| 1897 |
+
"reward_std": 0.22183530405163765,
|
| 1898 |
+
"rewards/accuracy_reward": 0.6227678582072258,
|
| 1899 |
+
"rewards/format_reward": 0.9944196343421936,
|
| 1900 |
+
"step": 135
|
| 1901 |
+
},
|
| 1902 |
+
{
|
| 1903 |
+
"clip_ratio": 0.0,
|
| 1904 |
+
"completion_length": 673.8951187133789,
|
| 1905 |
+
"epoch": 1.385297845373891,
|
| 1906 |
+
"grad_norm": 0.10060420632362366,
|
| 1907 |
+
"learning_rate": 2.603755066399718e-07,
|
| 1908 |
+
"loss": 0.0253,
|
| 1909 |
+
"num_tokens": 102764799.0,
|
| 1910 |
+
"reward": 1.6138393431901932,
|
| 1911 |
+
"reward_std": 0.2160534653812647,
|
| 1912 |
+
"rewards/accuracy_reward": 0.620535708963871,
|
| 1913 |
+
"rewards/format_reward": 0.9933035671710968,
|
| 1914 |
+
"step": 136
|
| 1915 |
+
},
|
| 1916 |
+
{
|
| 1917 |
+
"clip_ratio": 0.0,
|
| 1918 |
+
"completion_length": 719.1886596679688,
|
| 1919 |
+
"epoch": 1.3954372623574145,
|
| 1920 |
+
"grad_norm": 0.08712632954120636,
|
| 1921 |
+
"learning_rate": 2.5258082962836614e-07,
|
| 1922 |
+
"loss": 0.0291,
|
| 1923 |
+
"num_tokens": 103543752.0,
|
| 1924 |
+
"reward": 1.502232238650322,
|
| 1925 |
+
"reward_std": 0.17017750442028046,
|
| 1926 |
+
"rewards/accuracy_reward": 0.5100446492433548,
|
| 1927 |
+
"rewards/format_reward": 0.9921874850988388,
|
| 1928 |
+
"step": 137
|
| 1929 |
+
},
|
| 1930 |
+
{
|
| 1931 |
+
"clip_ratio": 0.0,
|
| 1932 |
+
"completion_length": 696.5636520385742,
|
| 1933 |
+
"epoch": 1.4055766793409379,
|
| 1934 |
+
"grad_norm": 0.20818105340003967,
|
| 1935 |
+
"learning_rate": 2.4486498346955023e-07,
|
| 1936 |
+
"loss": 0.0298,
|
| 1937 |
+
"num_tokens": 104290705.0,
|
| 1938 |
+
"reward": 1.6026786416769028,
|
| 1939 |
+
"reward_std": 0.24658331461250782,
|
| 1940 |
+
"rewards/accuracy_reward": 0.6104910746216774,
|
| 1941 |
+
"rewards/format_reward": 0.9921874925494194,
|
| 1942 |
+
"step": 138
|
| 1943 |
+
},
|
| 1944 |
+
{
|
| 1945 |
+
"clip_ratio": 0.0,
|
| 1946 |
+
"completion_length": 707.4475784301758,
|
| 1947 |
+
"epoch": 1.4157160963244613,
|
| 1948 |
+
"grad_norm": 0.09881385415792465,
|
| 1949 |
+
"learning_rate": 2.372304265289436e-07,
|
| 1950 |
+
"loss": 0.0172,
|
| 1951 |
+
"num_tokens": 105048874.0,
|
| 1952 |
+
"reward": 1.56026791036129,
|
| 1953 |
+
"reward_std": 0.24689678102731705,
|
| 1954 |
+
"rewards/accuracy_reward": 0.5703125037252903,
|
| 1955 |
+
"rewards/format_reward": 0.9899553582072258,
|
| 1956 |
+
"step": 139
|
| 1957 |
+
},
|
| 1958 |
+
{
|
| 1959 |
+
"clip_ratio": 0.0,
|
| 1960 |
+
"completion_length": 685.3381958007812,
|
| 1961 |
+
"epoch": 1.4258555133079849,
|
| 1962 |
+
"grad_norm": 0.0895194411277771,
|
| 1963 |
+
"learning_rate": 2.2967959127220137e-07,
|
| 1964 |
+
"loss": 0.0223,
|
| 1965 |
+
"num_tokens": 105784897.0,
|
| 1966 |
+
"reward": 1.6450893580913544,
|
| 1967 |
+
"reward_std": 0.20549658872187138,
|
| 1968 |
+
"rewards/accuracy_reward": 0.6551339253783226,
|
| 1969 |
+
"rewards/format_reward": 0.9899553582072258,
|
| 1970 |
+
"step": 140
|
| 1971 |
+
},
|
| 1972 |
+
{
|
| 1973 |
+
"clip_ratio": 0.0,
|
| 1974 |
+
"completion_length": 710.3538131713867,
|
| 1975 |
+
"epoch": 1.4359949302915083,
|
| 1976 |
+
"grad_norm": 0.10867197811603546,
|
| 1977 |
+
"learning_rate": 2.2221488349019902e-07,
|
| 1978 |
+
"loss": 0.036,
|
| 1979 |
+
"num_tokens": 106569734.0,
|
| 1980 |
+
"reward": 1.5904018580913544,
|
| 1981 |
+
"reward_std": 0.25699869357049465,
|
| 1982 |
+
"rewards/accuracy_reward": 0.5993303544819355,
|
| 1983 |
+
"rewards/format_reward": 0.9910714253783226,
|
| 1984 |
+
"step": 141
|
| 1985 |
+
},
|
| 1986 |
+
{
|
| 1987 |
+
"clip_ratio": 0.0,
|
| 1988 |
+
"completion_length": 672.2600936889648,
|
| 1989 |
+
"epoch": 1.4461343472750317,
|
| 1990 |
+
"grad_norm": 0.09679386019706726,
|
| 1991 |
+
"learning_rate": 2.1483868153251788e-07,
|
| 1992 |
+
"loss": 0.0437,
|
| 1993 |
+
"num_tokens": 107303623.0,
|
| 1994 |
+
"reward": 1.6272322237491608,
|
| 1995 |
+
"reward_std": 0.22507693991065025,
|
| 1996 |
+
"rewards/accuracy_reward": 0.6484374925494194,
|
| 1997 |
+
"rewards/format_reward": 0.9787946343421936,
|
| 1998 |
+
"step": 142
|
| 1999 |
+
},
|
| 2000 |
+
{
|
| 2001 |
+
"clip_ratio": 0.0,
|
| 2002 |
+
"completion_length": 721.3147583007812,
|
| 2003 |
+
"epoch": 1.456273764258555,
|
| 2004 |
+
"grad_norm": 0.09544187784194946,
|
| 2005 |
+
"learning_rate": 2.0755333554967346e-07,
|
| 2006 |
+
"loss": 0.0345,
|
| 2007 |
+
"num_tokens": 108087121.0,
|
| 2008 |
+
"reward": 1.5524554252624512,
|
| 2009 |
+
"reward_std": 0.2506922725588083,
|
| 2010 |
+
"rewards/accuracy_reward": 0.5658482275903225,
|
| 2011 |
+
"rewards/format_reward": 0.9866071417927742,
|
| 2012 |
+
"step": 143
|
| 2013 |
+
},
|
| 2014 |
+
{
|
| 2015 |
+
"clip_ratio": 0.0,
|
| 2016 |
+
"completion_length": 728.9341812133789,
|
| 2017 |
+
"epoch": 1.4664131812420786,
|
| 2018 |
+
"grad_norm": 0.10036015510559082,
|
| 2019 |
+
"learning_rate": 2.0036116674432652e-07,
|
| 2020 |
+
"loss": 0.0161,
|
| 2021 |
+
"num_tokens": 108876694.0,
|
| 2022 |
+
"reward": 1.5703125447034836,
|
| 2023 |
+
"reward_std": 0.23400406539440155,
|
| 2024 |
+
"rewards/accuracy_reward": 0.5803571455180645,
|
| 2025 |
+
"rewards/format_reward": 0.9899553507566452,
|
| 2026 |
+
"step": 144
|
| 2027 |
+
},
|
| 2028 |
+
{
|
| 2029 |
+
"clip_ratio": 0.0,
|
| 2030 |
+
"completion_length": 633.6451225280762,
|
| 2031 |
+
"epoch": 1.476552598225602,
|
| 2032 |
+
"grad_norm": 0.10225114226341248,
|
| 2033 |
+
"learning_rate": 1.9326446663172035e-07,
|
| 2034 |
+
"loss": 0.0215,
|
| 2035 |
+
"num_tokens": 109575920.0,
|
| 2036 |
+
"reward": 1.6830357909202576,
|
| 2037 |
+
"reward_std": 0.23387996666133404,
|
| 2038 |
+
"rewards/accuracy_reward": 0.6863839328289032,
|
| 2039 |
+
"rewards/format_reward": 0.9966517761349678,
|
| 2040 |
+
"step": 145
|
| 2041 |
+
},
|
| 2042 |
+
{
|
| 2043 |
+
"clip_ratio": 0.0,
|
| 2044 |
+
"completion_length": 653.1841735839844,
|
| 2045 |
+
"epoch": 1.4866920152091254,
|
| 2046 |
+
"grad_norm": 1.6936026811599731,
|
| 2047 |
+
"learning_rate": 1.8626549630957395e-07,
|
| 2048 |
+
"loss": 0.028,
|
| 2049 |
+
"num_tokens": 110289317.0,
|
| 2050 |
+
"reward": 1.6517858058214188,
|
| 2051 |
+
"reward_std": 0.24972818791866302,
|
| 2052 |
+
"rewards/accuracy_reward": 0.6607142947614193,
|
| 2053 |
+
"rewards/format_reward": 0.9910714253783226,
|
| 2054 |
+
"step": 146
|
| 2055 |
+
},
|
| 2056 |
+
{
|
| 2057 |
+
"clip_ratio": 0.0,
|
| 2058 |
+
"completion_length": 688.7846298217773,
|
| 2059 |
+
"epoch": 1.496831432192649,
|
| 2060 |
+
"grad_norm": 0.10226480662822723,
|
| 2061 |
+
"learning_rate": 1.7936648573766954e-07,
|
| 2062 |
+
"loss": 0.0531,
|
| 2063 |
+
"num_tokens": 111043180.0,
|
| 2064 |
+
"reward": 1.5881697088479996,
|
| 2065 |
+
"reward_std": 0.2411833181977272,
|
| 2066 |
+
"rewards/accuracy_reward": 0.5993303693830967,
|
| 2067 |
+
"rewards/format_reward": 0.9888392835855484,
|
| 2068 |
+
"step": 147
|
| 2069 |
+
},
|
| 2070 |
+
{
|
| 2071 |
+
"clip_ratio": 0.0,
|
| 2072 |
+
"completion_length": 730.7690048217773,
|
| 2073 |
+
"epoch": 1.5069708491761724,
|
| 2074 |
+
"grad_norm": 0.09249524772167206,
|
| 2075 |
+
"learning_rate": 1.725696330273575e-07,
|
| 2076 |
+
"loss": 0.0272,
|
| 2077 |
+
"num_tokens": 111828293.0,
|
| 2078 |
+
"reward": 1.5714286267757416,
|
| 2079 |
+
"reward_std": 0.2146799899637699,
|
| 2080 |
+
"rewards/accuracy_reward": 0.5803571343421936,
|
| 2081 |
+
"rewards/format_reward": 0.9910714328289032,
|
| 2082 |
+
"step": 148
|
| 2083 |
+
},
|
| 2084 |
+
{
|
| 2085 |
+
"clip_ratio": 0.0,
|
| 2086 |
+
"completion_length": 666.1719055175781,
|
| 2087 |
+
"epoch": 1.5171102661596958,
|
| 2088 |
+
"grad_norm": 0.09008808434009552,
|
| 2089 |
+
"learning_rate": 1.65877103741212e-07,
|
| 2090 |
+
"loss": 0.018,
|
| 2091 |
+
"num_tokens": 112549911.0,
|
| 2092 |
+
"reward": 1.7042411416769028,
|
| 2093 |
+
"reward_std": 0.18795600719749928,
|
| 2094 |
+
"rewards/accuracy_reward": 0.706473208963871,
|
| 2095 |
+
"rewards/format_reward": 0.9977678507566452,
|
| 2096 |
+
"step": 149
|
| 2097 |
+
},
|
| 2098 |
+
{
|
| 2099 |
+
"clip_ratio": 0.0,
|
| 2100 |
+
"completion_length": 704.3828430175781,
|
| 2101 |
+
"epoch": 1.5272496831432192,
|
| 2102 |
+
"grad_norm": 0.09881480038166046,
|
| 2103 |
+
"learning_rate": 1.592910302030544e-07,
|
| 2104 |
+
"loss": 0.0242,
|
| 2105 |
+
"num_tokens": 113300342.0,
|
| 2106 |
+
"reward": 1.5926340222358704,
|
| 2107 |
+
"reward_std": 0.2511101048439741,
|
| 2108 |
+
"rewards/accuracy_reward": 0.6015624888241291,
|
| 2109 |
+
"rewards/format_reward": 0.9910714253783226,
|
| 2110 |
+
"step": 150
|
| 2111 |
+
},
|
| 2112 |
+
{
|
| 2113 |
+
"clip_ratio": 0.0,
|
| 2114 |
+
"completion_length": 697.3582992553711,
|
| 2115 |
+
"epoch": 1.5373891001267426,
|
| 2116 |
+
"grad_norm": 0.11632478982210159,
|
| 2117 |
+
"learning_rate": 1.5281351081856976e-07,
|
| 2118 |
+
"loss": 0.0248,
|
| 2119 |
+
"num_tokens": 114056431.0,
|
| 2120 |
+
"reward": 1.6015625596046448,
|
| 2121 |
+
"reward_std": 0.20664218626916409,
|
| 2122 |
+
"rewards/accuracy_reward": 0.6127232201397419,
|
| 2123 |
+
"rewards/format_reward": 0.9888392835855484,
|
| 2124 |
+
"step": 151
|
| 2125 |
+
},
|
| 2126 |
+
{
|
| 2127 |
+
"clip_ratio": 0.0,
|
| 2128 |
+
"completion_length": 680.4286041259766,
|
| 2129 |
+
"epoch": 1.5475285171102662,
|
| 2130 |
+
"grad_norm": 0.09516316652297974,
|
| 2131 |
+
"learning_rate": 1.4644660940672627e-07,
|
| 2132 |
+
"loss": 0.0338,
|
| 2133 |
+
"num_tokens": 114791687.0,
|
| 2134 |
+
"reward": 1.648437574505806,
|
| 2135 |
+
"reward_std": 0.19783248752355576,
|
| 2136 |
+
"rewards/accuracy_reward": 0.6584821343421936,
|
| 2137 |
+
"rewards/format_reward": 0.9899553507566452,
|
| 2138 |
+
"step": 152
|
| 2139 |
+
},
|
| 2140 |
+
{
|
| 2141 |
+
"clip_ratio": 0.0,
|
| 2142 |
+
"completion_length": 669.5949096679688,
|
| 2143 |
+
"epoch": 1.5576679340937896,
|
| 2144 |
+
"grad_norm": 0.09097380936145782,
|
| 2145 |
+
"learning_rate": 1.4019235454221856e-07,
|
| 2146 |
+
"loss": 0.0149,
|
| 2147 |
+
"num_tokens": 115525716.0,
|
| 2148 |
+
"reward": 1.6104911714792252,
|
| 2149 |
+
"reward_std": 0.20410908479243517,
|
| 2150 |
+
"rewards/accuracy_reward": 0.6171875,
|
| 2151 |
+
"rewards/format_reward": 0.9933035746216774,
|
| 2152 |
+
"step": 153
|
| 2153 |
+
},
|
| 2154 |
+
{
|
| 2155 |
+
"clip_ratio": 0.0,
|
| 2156 |
+
"completion_length": 677.7366333007812,
|
| 2157 |
+
"epoch": 1.5678073510773132,
|
| 2158 |
+
"grad_norm": 0.09991227835416794,
|
| 2159 |
+
"learning_rate": 1.3405273890913737e-07,
|
| 2160 |
+
"loss": 0.0334,
|
| 2161 |
+
"num_tokens": 116254984.0,
|
| 2162 |
+
"reward": 1.5892857760190964,
|
| 2163 |
+
"reward_std": 0.2337410505861044,
|
| 2164 |
+
"rewards/accuracy_reward": 0.6004464365541935,
|
| 2165 |
+
"rewards/format_reward": 0.9888392761349678,
|
| 2166 |
+
"step": 154
|
| 2167 |
+
},
|
| 2168 |
+
{
|
| 2169 |
+
"clip_ratio": 0.0,
|
| 2170 |
+
"completion_length": 691.3571853637695,
|
| 2171 |
+
"epoch": 1.5779467680608366,
|
| 2172 |
+
"grad_norm": 0.18712472915649414,
|
| 2173 |
+
"learning_rate": 1.280297186660752e-07,
|
| 2174 |
+
"loss": 0.0335,
|
| 2175 |
+
"num_tokens": 116992816.0,
|
| 2176 |
+
"reward": 1.6595982760190964,
|
| 2177 |
+
"reward_std": 0.2452637990936637,
|
| 2178 |
+
"rewards/accuracy_reward": 0.6707589328289032,
|
| 2179 |
+
"rewards/format_reward": 0.9888392835855484,
|
| 2180 |
+
"step": 155
|
| 2181 |
+
},
|
| 2182 |
+
{
|
| 2183 |
+
"clip_ratio": 0.0,
|
| 2184 |
+
"completion_length": 758.325927734375,
|
| 2185 |
+
"epoch": 1.58808618504436,
|
| 2186 |
+
"grad_norm": 0.09767128527164459,
|
| 2187 |
+
"learning_rate": 1.2212521282287093e-07,
|
| 2188 |
+
"loss": 0.0328,
|
| 2189 |
+
"num_tokens": 117792204.0,
|
| 2190 |
+
"reward": 1.5558036416769028,
|
| 2191 |
+
"reward_std": 0.23506544344127178,
|
| 2192 |
+
"rewards/accuracy_reward": 0.5647321492433548,
|
| 2193 |
+
"rewards/format_reward": 0.9910714104771614,
|
| 2194 |
+
"step": 156
|
| 2195 |
+
},
|
| 2196 |
+
{
|
| 2197 |
+
"clip_ratio": 0.0,
|
| 2198 |
+
"completion_length": 704.4464721679688,
|
| 2199 |
+
"epoch": 1.5982256020278833,
|
| 2200 |
+
"grad_norm": 0.08936415612697601,
|
| 2201 |
+
"learning_rate": 1.1634110262918717e-07,
|
| 2202 |
+
"loss": 0.0365,
|
| 2203 |
+
"num_tokens": 118567396.0,
|
| 2204 |
+
"reward": 1.5758929252624512,
|
| 2205 |
+
"reward_std": 0.21123607829213142,
|
| 2206 |
+
"rewards/accuracy_reward": 0.5970982126891613,
|
| 2207 |
+
"rewards/format_reward": 0.9787946417927742,
|
| 2208 |
+
"step": 157
|
| 2209 |
+
},
|
| 2210 |
+
{
|
| 2211 |
+
"clip_ratio": 0.0,
|
| 2212 |
+
"completion_length": 721.3761444091797,
|
| 2213 |
+
"epoch": 1.6083650190114067,
|
| 2214 |
+
"grad_norm": 0.9475556015968323,
|
| 2215 |
+
"learning_rate": 1.1067923097512255e-07,
|
| 2216 |
+
"loss": 0.051,
|
| 2217 |
+
"num_tokens": 119344461.0,
|
| 2218 |
+
"reward": 1.5680804252624512,
|
| 2219 |
+
"reward_std": 0.24346480891108513,
|
| 2220 |
+
"rewards/accuracy_reward": 0.5848214253783226,
|
| 2221 |
+
"rewards/format_reward": 0.9832589253783226,
|
| 2222 |
+
"step": 158
|
| 2223 |
+
},
|
| 2224 |
+
{
|
| 2225 |
+
"clip_ratio": 0.0,
|
| 2226 |
+
"completion_length": 720.2455749511719,
|
| 2227 |
+
"epoch": 1.6185044359949303,
|
| 2228 |
+
"grad_norm": 0.10232140868902206,
|
| 2229 |
+
"learning_rate": 1.0514140180404202e-07,
|
| 2230 |
+
"loss": 0.0421,
|
| 2231 |
+
"num_tokens": 120118857.0,
|
| 2232 |
+
"reward": 1.5736607909202576,
|
| 2233 |
+
"reward_std": 0.2383702825754881,
|
| 2234 |
+
"rewards/accuracy_reward": 0.5859375,
|
| 2235 |
+
"rewards/format_reward": 0.987723208963871,
|
| 2236 |
+
"step": 159
|
| 2237 |
+
},
|
| 2238 |
+
{
|
| 2239 |
+
"clip_ratio": 0.0,
|
| 2240 |
+
"completion_length": 724.2288284301758,
|
| 2241 |
+
"epoch": 1.6286438529784537,
|
| 2242 |
+
"grad_norm": 0.17790253460407257,
|
| 2243 |
+
"learning_rate": 9.972937953781984e-08,
|
| 2244 |
+
"loss": 0.0345,
|
| 2245 |
+
"num_tokens": 120914734.0,
|
| 2246 |
+
"reward": 1.5167411416769028,
|
| 2247 |
+
"reward_std": 0.22692781873047352,
|
| 2248 |
+
"rewards/accuracy_reward": 0.527901791036129,
|
| 2249 |
+
"rewards/format_reward": 0.9888392835855484,
|
| 2250 |
+
"step": 160
|
| 2251 |
+
},
|
| 2252 |
+
{
|
| 2253 |
+
"clip_ratio": 0.0,
|
| 2254 |
+
"completion_length": 671.4531478881836,
|
| 2255 |
+
"epoch": 1.6387832699619773,
|
| 2256 |
+
"grad_norm": 0.08403746038675308,
|
| 2257 |
+
"learning_rate": 9.444488851467041e-08,
|
| 2258 |
+
"loss": 0.0196,
|
| 2259 |
+
"num_tokens": 121634028.0,
|
| 2260 |
+
"reward": 1.6439733058214188,
|
| 2261 |
+
"reward_std": 0.18759915605187416,
|
| 2262 |
+
"rewards/accuracy_reward": 0.6517857015132904,
|
| 2263 |
+
"rewards/format_reward": 0.9921874925494194,
|
| 2264 |
+
"step": 161
|
| 2265 |
+
},
|
| 2266 |
+
{
|
| 2267 |
+
"clip_ratio": 0.0,
|
| 2268 |
+
"completion_length": 686.6283798217773,
|
| 2269 |
+
"epoch": 1.6489226869455007,
|
| 2270 |
+
"grad_norm": 0.11062318086624146,
|
| 2271 |
+
"learning_rate": 8.928961243975436e-08,
|
| 2272 |
+
"loss": 0.0421,
|
| 2273 |
+
"num_tokens": 122382103.0,
|
| 2274 |
+
"reward": 1.5970982909202576,
|
| 2275 |
+
"reward_std": 0.22882544435560703,
|
| 2276 |
+
"rewards/accuracy_reward": 0.6119505539536476,
|
| 2277 |
+
"rewards/format_reward": 0.991071417927742,
|
| 2278 |
+
"step": 162
|
| 2279 |
+
},
|
| 2280 |
+
{
|
| 2281 |
+
"clip_ratio": 0.0,
|
| 2282 |
+
"completion_length": 666.947566986084,
|
| 2283 |
+
"epoch": 1.659062103929024,
|
| 2284 |
+
"grad_norm": 0.09584940969944,
|
| 2285 |
+
"learning_rate": 8.426519384872732e-08,
|
| 2286 |
+
"loss": 0.0095,
|
| 2287 |
+
"num_tokens": 123101824.0,
|
| 2288 |
+
"reward": 1.6573661416769028,
|
| 2289 |
+
"reward_std": 0.2075268942862749,
|
| 2290 |
+
"rewards/accuracy_reward": 0.6607142835855484,
|
| 2291 |
+
"rewards/format_reward": 0.9966517835855484,
|
| 2292 |
+
"step": 163
|
| 2293 |
+
},
|
| 2294 |
+
{
|
| 2295 |
+
"clip_ratio": 0.0,
|
| 2296 |
+
"completion_length": 697.6964645385742,
|
| 2297 |
+
"epoch": 1.6692015209125475,
|
| 2298 |
+
"grad_norm": 0.10291486978530884,
|
| 2299 |
+
"learning_rate": 7.937323358440934e-08,
|
| 2300 |
+
"loss": 0.0226,
|
| 2301 |
+
"num_tokens": 123857992.0,
|
| 2302 |
+
"reward": 1.6450893431901932,
|
| 2303 |
+
"reward_std": 0.25325831212103367,
|
| 2304 |
+
"rewards/accuracy_reward": 0.6506696417927742,
|
| 2305 |
+
"rewards/format_reward": 0.9944196343421936,
|
| 2306 |
+
"step": 164
|
| 2307 |
+
},
|
| 2308 |
+
{
|
| 2309 |
+
"clip_ratio": 0.0,
|
| 2310 |
+
"completion_length": 691.475471496582,
|
| 2311 |
+
"epoch": 1.6793409378960709,
|
| 2312 |
+
"grad_norm": 0.09860701858997345,
|
| 2313 |
+
"learning_rate": 7.461529028673463e-08,
|
| 2314 |
+
"loss": 0.0197,
|
| 2315 |
+
"num_tokens": 124610362.0,
|
| 2316 |
+
"reward": 1.6473215073347092,
|
| 2317 |
+
"reward_std": 0.2233567014336586,
|
| 2318 |
+
"rewards/accuracy_reward": 0.652901791036129,
|
| 2319 |
+
"rewards/format_reward": 0.9944196343421936,
|
| 2320 |
+
"step": 165
|
| 2321 |
+
},
|
| 2322 |
+
{
|
| 2323 |
+
"clip_ratio": 0.0,
|
| 2324 |
+
"completion_length": 712.0446701049805,
|
| 2325 |
+
"epoch": 1.6894803548795945,
|
| 2326 |
+
"grad_norm": 0.10766546428203583,
|
| 2327 |
+
"learning_rate": 6.999287989614971e-08,
|
| 2328 |
+
"loss": 0.0403,
|
| 2329 |
+
"num_tokens": 125380794.0,
|
| 2330 |
+
"reward": 1.5881696939468384,
|
| 2331 |
+
"reward_std": 0.2603952419012785,
|
| 2332 |
+
"rewards/accuracy_reward": 0.5970982126891613,
|
| 2333 |
+
"rewards/format_reward": 0.9910714253783226,
|
| 2334 |
+
"step": 166
|
| 2335 |
+
},
|
| 2336 |
+
{
|
| 2337 |
+
"clip_ratio": 0.0,
|
| 2338 |
+
"completion_length": 671.4163131713867,
|
| 2339 |
+
"epoch": 1.6996197718631179,
|
| 2340 |
+
"grad_norm": 0.15764014422893524,
|
| 2341 |
+
"learning_rate": 6.550747517061656e-08,
|
| 2342 |
+
"loss": 0.0282,
|
| 2343 |
+
"num_tokens": 126108927.0,
|
| 2344 |
+
"reward": 1.6595982909202576,
|
| 2345 |
+
"reward_std": 0.20449624955654144,
|
| 2346 |
+
"rewards/accuracy_reward": 0.6662946343421936,
|
| 2347 |
+
"rewards/format_reward": 0.9933035671710968,
|
| 2348 |
+
"step": 167
|
| 2349 |
+
},
|
| 2350 |
+
{
|
| 2351 |
+
"clip_ratio": 0.0,
|
| 2352 |
+
"completion_length": 703.9542617797852,
|
| 2353 |
+
"epoch": 1.7097591888466415,
|
| 2354 |
+
"grad_norm": 0.09230643510818481,
|
| 2355 |
+
"learning_rate": 6.116050521637218e-08,
|
| 2356 |
+
"loss": 0.0282,
|
| 2357 |
+
"num_tokens": 126867998.0,
|
| 2358 |
+
"reward": 1.6395090073347092,
|
| 2359 |
+
"reward_std": 0.2097290549427271,
|
| 2360 |
+
"rewards/accuracy_reward": 0.6450892835855484,
|
| 2361 |
+
"rewards/format_reward": 0.9944196343421936,
|
| 2362 |
+
"step": 168
|
| 2363 |
+
},
|
| 2364 |
+
{
|
| 2365 |
+
"clip_ratio": 0.0,
|
| 2366 |
+
"completion_length": 672.6339569091797,
|
| 2367 |
+
"epoch": 1.7198986058301649,
|
| 2368 |
+
"grad_norm": 0.11055152863264084,
|
| 2369 |
+
"learning_rate": 5.6953355032598795e-08,
|
| 2370 |
+
"loss": 0.0277,
|
| 2371 |
+
"num_tokens": 127607750.0,
|
| 2372 |
+
"reward": 1.5736607760190964,
|
| 2373 |
+
"reward_std": 0.2487525064498186,
|
| 2374 |
+
"rewards/accuracy_reward": 0.5881696380674839,
|
| 2375 |
+
"rewards/format_reward": 0.9854910746216774,
|
| 2376 |
+
"step": 169
|
| 2377 |
+
},
|
| 2378 |
+
{
|
| 2379 |
+
"clip_ratio": 0.0,
|
| 2380 |
+
"completion_length": 696.6395416259766,
|
| 2381 |
+
"epoch": 1.7300380228136882,
|
| 2382 |
+
"grad_norm": 0.09632628411054611,
|
| 2383 |
+
"learning_rate": 5.288736507014435e-08,
|
| 2384 |
+
"loss": 0.0203,
|
| 2385 |
+
"num_tokens": 128355939.0,
|
| 2386 |
+
"reward": 1.640625074505806,
|
| 2387 |
+
"reward_std": 0.22130538150668144,
|
| 2388 |
+
"rewards/accuracy_reward": 0.6517857238650322,
|
| 2389 |
+
"rewards/format_reward": 0.988839291036129,
|
| 2390 |
+
"step": 170
|
| 2391 |
+
},
|
| 2392 |
+
{
|
| 2393 |
+
"clip_ratio": 0.0,
|
| 2394 |
+
"completion_length": 665.4888763427734,
|
| 2395 |
+
"epoch": 1.7401774397972116,
|
| 2396 |
+
"grad_norm": 0.185902938246727,
|
| 2397 |
+
"learning_rate": 4.896383080443933e-08,
|
| 2398 |
+
"loss": 0.014,
|
| 2399 |
+
"num_tokens": 129076969.0,
|
| 2400 |
+
"reward": 1.59933041036129,
|
| 2401 |
+
"reward_std": 0.22374757565557957,
|
| 2402 |
+
"rewards/accuracy_reward": 0.6037946306169033,
|
| 2403 |
+
"rewards/format_reward": 0.995535708963871,
|
| 2404 |
+
"step": 171
|
| 2405 |
+
},
|
| 2406 |
+
{
|
| 2407 |
+
"clip_ratio": 0.0,
|
| 2408 |
+
"completion_length": 655.0580596923828,
|
| 2409 |
+
"epoch": 1.750316856780735,
|
| 2410 |
+
"grad_norm": 0.0977133959531784,
|
| 2411 |
+
"learning_rate": 4.518400232274078e-08,
|
| 2412 |
+
"loss": 0.0151,
|
| 2413 |
+
"num_tokens": 129794405.0,
|
| 2414 |
+
"reward": 1.5870536416769028,
|
| 2415 |
+
"reward_std": 0.21878547966480255,
|
| 2416 |
+
"rewards/accuracy_reward": 0.599759615957737,
|
| 2417 |
+
"rewards/format_reward": 0.9933035597205162,
|
| 2418 |
+
"step": 172
|
| 2419 |
+
},
|
| 2420 |
+
{
|
| 2421 |
+
"clip_ratio": 0.0,
|
| 2422 |
+
"completion_length": 660.8594131469727,
|
| 2423 |
+
"epoch": 1.7604562737642584,
|
| 2424 |
+
"grad_norm": 0.10144926607608795,
|
| 2425 |
+
"learning_rate": 4.1549083925840165e-08,
|
| 2426 |
+
"loss": 0.0387,
|
| 2427 |
+
"num_tokens": 130503983.0,
|
| 2428 |
+
"reward": 1.6529018580913544,
|
| 2429 |
+
"reward_std": 0.22397751361131668,
|
| 2430 |
+
"rewards/accuracy_reward": 0.6662946417927742,
|
| 2431 |
+
"rewards/format_reward": 0.9866071417927742,
|
| 2432 |
+
"step": 173
|
| 2433 |
+
},
|
| 2434 |
+
{
|
| 2435 |
+
"clip_ratio": 0.0,
|
| 2436 |
+
"completion_length": 672.0837326049805,
|
| 2437 |
+
"epoch": 1.770595690747782,
|
| 2438 |
+
"grad_norm": 0.09892982989549637,
|
| 2439 |
+
"learning_rate": 3.806023374435663e-08,
|
| 2440 |
+
"loss": 0.0305,
|
| 2441 |
+
"num_tokens": 131234698.0,
|
| 2442 |
+
"reward": 1.6082590222358704,
|
| 2443 |
+
"reward_std": 0.21078570373356342,
|
| 2444 |
+
"rewards/accuracy_reward": 0.6149553582072258,
|
| 2445 |
+
"rewards/format_reward": 0.9933035597205162,
|
| 2446 |
+
"step": 174
|
| 2447 |
+
},
|
| 2448 |
+
{
|
| 2449 |
+
"clip_ratio": 0.0,
|
| 2450 |
+
"completion_length": 700.0781631469727,
|
| 2451 |
+
"epoch": 1.7807351077313056,
|
| 2452 |
+
"grad_norm": 0.08607691526412964,
|
| 2453 |
+
"learning_rate": 3.4718563369743213e-08,
|
| 2454 |
+
"loss": 0.0126,
|
| 2455 |
+
"num_tokens": 131995296.0,
|
| 2456 |
+
"reward": 1.5825893580913544,
|
| 2457 |
+
"reward_std": 0.19891513139009476,
|
| 2458 |
+
"rewards/accuracy_reward": 0.5870535746216774,
|
| 2459 |
+
"rewards/format_reward": 0.9955357164144516,
|
| 2460 |
+
"step": 175
|
| 2461 |
+
},
|
| 2462 |
+
{
|
| 2463 |
+
"clip_ratio": 0.0,
|
| 2464 |
+
"completion_length": 683.560302734375,
|
| 2465 |
+
"epoch": 1.790874524714829,
|
| 2466 |
+
"grad_norm": 0.0931280255317688,
|
| 2467 |
+
"learning_rate": 3.15251375001192e-08,
|
| 2468 |
+
"loss": 0.0266,
|
| 2469 |
+
"num_tokens": 132747870.0,
|
| 2470 |
+
"reward": 1.6540179401636124,
|
| 2471 |
+
"reward_std": 0.2136262021958828,
|
| 2472 |
+
"rewards/accuracy_reward": 0.6662946417927742,
|
| 2473 |
+
"rewards/format_reward": 0.987723208963871,
|
| 2474 |
+
"step": 176
|
| 2475 |
+
},
|
| 2476 |
+
{
|
| 2477 |
+
"clip_ratio": 0.0,
|
| 2478 |
+
"completion_length": 673.4799385070801,
|
| 2479 |
+
"epoch": 1.8010139416983524,
|
| 2480 |
+
"grad_norm": 0.09533898532390594,
|
| 2481 |
+
"learning_rate": 2.8480973601043955e-08,
|
| 2482 |
+
"loss": 0.0372,
|
| 2483 |
+
"num_tokens": 133486116.0,
|
| 2484 |
+
"reward": 1.6417411416769028,
|
| 2485 |
+
"reward_std": 0.18038041703402996,
|
| 2486 |
+
"rewards/accuracy_reward": 0.6529017798602581,
|
| 2487 |
+
"rewards/format_reward": 0.9888392761349678,
|
| 2488 |
+
"step": 177
|
| 2489 |
+
},
|
| 2490 |
+
{
|
| 2491 |
+
"clip_ratio": 0.0,
|
| 2492 |
+
"completion_length": 709.9821853637695,
|
| 2493 |
+
"epoch": 1.8111533586818758,
|
| 2494 |
+
"grad_norm": 0.1904437392950058,
|
| 2495 |
+
"learning_rate": 2.558704158134023e-08,
|
| 2496 |
+
"loss": 0.0261,
|
| 2497 |
+
"num_tokens": 134262284.0,
|
| 2498 |
+
"reward": 1.5825893431901932,
|
| 2499 |
+
"reward_std": 0.22126466780900955,
|
| 2500 |
+
"rewards/accuracy_reward": 0.5937500111758709,
|
| 2501 |
+
"rewards/format_reward": 0.9888392761349678,
|
| 2502 |
+
"step": 178
|
| 2503 |
+
},
|
| 2504 |
+
{
|
| 2505 |
+
"clip_ratio": 0.0,
|
| 2506 |
+
"completion_length": 719.8861846923828,
|
| 2507 |
+
"epoch": 1.8212927756653992,
|
| 2508 |
+
"grad_norm": 0.0858464315533638,
|
| 2509 |
+
"learning_rate": 2.2844263484068093e-08,
|
| 2510 |
+
"loss": 0.0161,
|
| 2511 |
+
"num_tokens": 135034302.0,
|
| 2512 |
+
"reward": 1.5937500447034836,
|
| 2513 |
+
"reward_std": 0.18957781419157982,
|
| 2514 |
+
"rewards/accuracy_reward": 0.6004464216530323,
|
| 2515 |
+
"rewards/format_reward": 0.9933035671710968,
|
| 2516 |
+
"step": 179
|
| 2517 |
+
},
|
| 2518 |
+
{
|
| 2519 |
+
"clip_ratio": 0.0,
|
| 2520 |
+
"completion_length": 670.2042694091797,
|
| 2521 |
+
"epoch": 1.8314321926489225,
|
| 2522 |
+
"grad_norm": 0.09248825907707214,
|
| 2523 |
+
"learning_rate": 2.025351319275137e-08,
|
| 2524 |
+
"loss": 0.0237,
|
| 2525 |
+
"num_tokens": 135765773.0,
|
| 2526 |
+
"reward": 1.6517857909202576,
|
| 2527 |
+
"reward_std": 0.20931300334632397,
|
| 2528 |
+
"rewards/accuracy_reward": 0.6607142835855484,
|
| 2529 |
+
"rewards/format_reward": 0.991071417927742,
|
| 2530 |
+
"step": 180
|
| 2531 |
+
},
|
| 2532 |
+
{
|
| 2533 |
+
"clip_ratio": 0.0,
|
| 2534 |
+
"completion_length": 697.1328506469727,
|
| 2535 |
+
"epoch": 1.8415716096324461,
|
| 2536 |
+
"grad_norm": 0.09593623131513596,
|
| 2537 |
+
"learning_rate": 1.781561615294652e-08,
|
| 2538 |
+
"loss": 0.0136,
|
| 2539 |
+
"num_tokens": 136524868.0,
|
| 2540 |
+
"reward": 1.5881697237491608,
|
| 2541 |
+
"reward_std": 0.23646764643490314,
|
| 2542 |
+
"rewards/accuracy_reward": 0.597098208963871,
|
| 2543 |
+
"rewards/format_reward": 0.9910714253783226,
|
| 2544 |
+
"step": 181
|
| 2545 |
+
},
|
| 2546 |
+
{
|
| 2547 |
+
"clip_ratio": 0.0,
|
| 2548 |
+
"completion_length": 709.1886520385742,
|
| 2549 |
+
"epoch": 1.8517110266159695,
|
| 2550 |
+
"grad_norm": 0.10065972059965134,
|
| 2551 |
+
"learning_rate": 1.553134910924636e-08,
|
| 2552 |
+
"loss": 0.0373,
|
| 2553 |
+
"num_tokens": 137296405.0,
|
| 2554 |
+
"reward": 1.539062574505806,
|
| 2555 |
+
"reward_std": 0.22722396813333035,
|
| 2556 |
+
"rewards/accuracy_reward": 0.5479910708963871,
|
| 2557 |
+
"rewards/format_reward": 0.991071417927742,
|
| 2558 |
+
"step": 182
|
| 2559 |
+
},
|
| 2560 |
+
{
|
| 2561 |
+
"clip_ratio": 0.0,
|
| 2562 |
+
"completion_length": 661.6339569091797,
|
| 2563 |
+
"epoch": 1.8618504435994931,
|
| 2564 |
+
"grad_norm": 0.09599631279706955,
|
| 2565 |
+
"learning_rate": 1.340143985779829e-08,
|
| 2566 |
+
"loss": 0.0283,
|
| 2567 |
+
"num_tokens": 138015989.0,
|
| 2568 |
+
"reward": 1.6540179550647736,
|
| 2569 |
+
"reward_std": 0.18772431463003159,
|
| 2570 |
+
"rewards/accuracy_reward": 0.661830373108387,
|
| 2571 |
+
"rewards/format_reward": 0.9921874925494194,
|
| 2572 |
+
"step": 183
|
| 2573 |
+
},
|
| 2574 |
+
{
|
| 2575 |
+
"clip_ratio": 0.0,
|
| 2576 |
+
"completion_length": 697.3761444091797,
|
| 2577 |
+
"epoch": 1.8719898605830165,
|
| 2578 |
+
"grad_norm": 0.09907021373510361,
|
| 2579 |
+
"learning_rate": 1.1426567014420297e-08,
|
| 2580 |
+
"loss": 0.047,
|
| 2581 |
+
"num_tokens": 138767542.0,
|
| 2582 |
+
"reward": 1.6383929550647736,
|
| 2583 |
+
"reward_std": 0.23151767998933792,
|
| 2584 |
+
"rewards/accuracy_reward": 0.6551339402794838,
|
| 2585 |
+
"rewards/format_reward": 0.9832589104771614,
|
| 2586 |
+
"step": 184
|
| 2587 |
+
},
|
| 2588 |
+
{
|
| 2589 |
+
"clip_ratio": 0.0,
|
| 2590 |
+
"completion_length": 706.0480117797852,
|
| 2591 |
+
"epoch": 1.88212927756654,
|
| 2592 |
+
"grad_norm": 0.09566432982683182,
|
| 2593 |
+
"learning_rate": 9.607359798384784e-09,
|
| 2594 |
+
"loss": 0.0372,
|
| 2595 |
+
"num_tokens": 139515641.0,
|
| 2596 |
+
"reward": 1.592633992433548,
|
| 2597 |
+
"reward_std": 0.2507201302796602,
|
| 2598 |
+
"rewards/accuracy_reward": 0.6015625074505806,
|
| 2599 |
+
"rewards/format_reward": 0.991071417927742,
|
| 2600 |
+
"step": 185
|
| 2601 |
+
},
|
| 2602 |
+
{
|
| 2603 |
+
"clip_ratio": 0.0,
|
| 2604 |
+
"completion_length": 675.1484756469727,
|
| 2605 |
+
"epoch": 1.8922686945500633,
|
| 2606 |
+
"grad_norm": 0.09894827008247375,
|
| 2607 |
+
"learning_rate": 7.944397831941951e-09,
|
| 2608 |
+
"loss": 0.031,
|
| 2609 |
+
"num_tokens": 140267070.0,
|
| 2610 |
+
"reward": 1.5837054252624512,
|
| 2611 |
+
"reward_std": 0.21183509565889835,
|
| 2612 |
+
"rewards/accuracy_reward": 0.5937499962747097,
|
| 2613 |
+
"rewards/format_reward": 0.9899553507566452,
|
| 2614 |
+
"step": 186
|
| 2615 |
+
},
|
| 2616 |
+
{
|
| 2617 |
+
"clip_ratio": 0.0,
|
| 2618 |
+
"completion_length": 740.9620971679688,
|
| 2619 |
+
"epoch": 1.9024081115335867,
|
| 2620 |
+
"grad_norm": 0.1215052530169487,
|
| 2621 |
+
"learning_rate": 6.438210955644452e-09,
|
| 2622 |
+
"loss": 0.0443,
|
| 2623 |
+
"num_tokens": 141073396.0,
|
| 2624 |
+
"reward": 1.5044643580913544,
|
| 2625 |
+
"reward_std": 0.2504011485725641,
|
| 2626 |
+
"rewards/accuracy_reward": 0.5256696455180645,
|
| 2627 |
+
"rewards/format_reward": 0.9787946343421936,
|
| 2628 |
+
"step": 187
|
| 2629 |
+
},
|
| 2630 |
+
{
|
| 2631 |
+
"clip_ratio": 0.0,
|
| 2632 |
+
"completion_length": 691.3471298217773,
|
| 2633 |
+
"epoch": 1.9125475285171103,
|
| 2634 |
+
"grad_norm": 0.08761589974164963,
|
| 2635 |
+
"learning_rate": 5.0892790595336575e-09,
|
| 2636 |
+
"loss": 0.0307,
|
| 2637 |
+
"num_tokens": 141829859.0,
|
| 2638 |
+
"reward": 1.6305804550647736,
|
| 2639 |
+
"reward_std": 0.220117699354887,
|
| 2640 |
+
"rewards/accuracy_reward": 0.6406250037252903,
|
| 2641 |
+
"rewards/format_reward": 0.9899553433060646,
|
| 2642 |
+
"step": 188
|
| 2643 |
+
},
|
| 2644 |
+
{
|
| 2645 |
+
"clip_ratio": 0.0,
|
| 2646 |
+
"completion_length": 687.7377548217773,
|
| 2647 |
+
"epoch": 1.9226869455006337,
|
| 2648 |
+
"grad_norm": 0.10144967585802078,
|
| 2649 |
+
"learning_rate": 3.898031930240797e-09,
|
| 2650 |
+
"loss": 0.03,
|
| 2651 |
+
"num_tokens": 142571072.0,
|
| 2652 |
+
"reward": 1.6037946790456772,
|
| 2653 |
+
"reward_std": 0.22627083584666252,
|
| 2654 |
+
"rewards/accuracy_reward": 0.6138392947614193,
|
| 2655 |
+
"rewards/format_reward": 0.9899553582072258,
|
| 2656 |
+
"step": 189
|
| 2657 |
+
},
|
| 2658 |
+
{
|
| 2659 |
+
"clip_ratio": 0.0,
|
| 2660 |
+
"completion_length": 743.5949020385742,
|
| 2661 |
+
"epoch": 1.9328263624841573,
|
| 2662 |
+
"grad_norm": 0.08726272732019424,
|
| 2663 |
+
"learning_rate": 2.8648491140513264e-09,
|
| 2664 |
+
"loss": 0.0307,
|
| 2665 |
+
"num_tokens": 143363781.0,
|
| 2666 |
+
"reward": 1.5948661267757416,
|
| 2667 |
+
"reward_std": 0.2067482192069292,
|
| 2668 |
+
"rewards/accuracy_reward": 0.6004464328289032,
|
| 2669 |
+
"rewards/format_reward": 0.9944196417927742,
|
| 2670 |
+
"step": 190
|
| 2671 |
+
},
|
| 2672 |
+
{
|
| 2673 |
+
"clip_ratio": 0.0,
|
| 2674 |
+
"completion_length": 631.5368728637695,
|
| 2675 |
+
"epoch": 1.9429657794676807,
|
| 2676 |
+
"grad_norm": 0.09442251920700073,
|
| 2677 |
+
"learning_rate": 1.9900597959770505e-09,
|
| 2678 |
+
"loss": 0.0309,
|
| 2679 |
+
"num_tokens": 144054766.0,
|
| 2680 |
+
"reward": 1.7064732909202576,
|
| 2681 |
+
"reward_std": 0.1611343901604414,
|
| 2682 |
+
"rewards/accuracy_reward": 0.7142857164144516,
|
| 2683 |
+
"rewards/format_reward": 0.9921874925494194,
|
| 2684 |
+
"step": 191
|
| 2685 |
+
},
|
| 2686 |
+
{
|
| 2687 |
+
"clip_ratio": 0.0,
|
| 2688 |
+
"completion_length": 702.8839721679688,
|
| 2689 |
+
"epoch": 1.953105196451204,
|
| 2690 |
+
"grad_norm": 0.09530234336853027,
|
| 2691 |
+
"learning_rate": 1.2739426948732424e-09,
|
| 2692 |
+
"loss": 0.0236,
|
| 2693 |
+
"num_tokens": 144818518.0,
|
| 2694 |
+
"reward": 1.6216518580913544,
|
| 2695 |
+
"reward_std": 0.22148274257779121,
|
| 2696 |
+
"rewards/accuracy_reward": 0.6283482164144516,
|
| 2697 |
+
"rewards/format_reward": 0.9933035597205162,
|
| 2698 |
+
"step": 192
|
| 2699 |
+
},
|
| 2700 |
+
{
|
| 2701 |
+
"clip_ratio": 0.0,
|
| 2702 |
+
"completion_length": 705.2678833007812,
|
| 2703 |
+
"epoch": 1.9632446134347274,
|
| 2704 |
+
"grad_norm": 0.11064296960830688,
|
| 2705 |
+
"learning_rate": 7.16725974635568e-10,
|
| 2706 |
+
"loss": 0.0467,
|
| 2707 |
+
"num_tokens": 145576278.0,
|
| 2708 |
+
"reward": 1.6149554401636124,
|
| 2709 |
+
"reward_std": 0.2686098553240299,
|
| 2710 |
+
"rewards/accuracy_reward": 0.6305803507566452,
|
| 2711 |
+
"rewards/format_reward": 0.9843749850988388,
|
| 2712 |
+
"step": 193
|
| 2713 |
+
},
|
| 2714 |
+
{
|
| 2715 |
+
"clip_ratio": 0.0,
|
| 2716 |
+
"completion_length": 709.090446472168,
|
| 2717 |
+
"epoch": 1.9733840304182508,
|
| 2718 |
+
"grad_norm": 0.1301695555448532,
|
| 2719 |
+
"learning_rate": 3.185871715041255e-10,
|
| 2720 |
+
"loss": 0.0256,
|
| 2721 |
+
"num_tokens": 146344471.0,
|
| 2722 |
+
"reward": 1.5904018580913544,
|
| 2723 |
+
"reward_std": 0.20222523249685764,
|
| 2724 |
+
"rewards/accuracy_reward": 0.5959821455180645,
|
| 2725 |
+
"rewards/format_reward": 0.9944196343421936,
|
| 2726 |
+
"step": 194
|
| 2727 |
+
},
|
| 2728 |
+
{
|
| 2729 |
+
"clip_ratio": 0.0,
|
| 2730 |
+
"completion_length": 649.944221496582,
|
| 2731 |
+
"epoch": 1.9835234474017744,
|
| 2732 |
+
"grad_norm": 0.10688284784555435,
|
| 2733 |
+
"learning_rate": 7.96531374983589e-11,
|
| 2734 |
+
"loss": 0.0264,
|
| 2735 |
+
"num_tokens": 147047301.0,
|
| 2736 |
+
"reward": 1.6361607760190964,
|
| 2737 |
+
"reward_std": 0.23337006010115147,
|
| 2738 |
+
"rewards/accuracy_reward": 0.6417410708963871,
|
| 2739 |
+
"rewards/format_reward": 0.9944196343421936,
|
| 2740 |
+
"step": 195
|
| 2741 |
+
},
|
| 2742 |
+
{
|
| 2743 |
+
"clip_ratio": 0.0,
|
| 2744 |
+
"completion_length": 693.314453125,
|
| 2745 |
+
"epoch": 1.9936628643852978,
|
| 2746 |
+
"grad_norm": 0.09134732186794281,
|
| 2747 |
+
"learning_rate": 0.0,
|
| 2748 |
+
"loss": 0.0173,
|
| 2749 |
+
"num_tokens": 147801661.0,
|
| 2750 |
+
"reward": 1.56026791036129,
|
| 2751 |
+
"reward_std": 0.21255235373973846,
|
| 2752 |
+
"rewards/accuracy_reward": 0.565848208963871,
|
| 2753 |
+
"rewards/format_reward": 0.9944196343421936,
|
| 2754 |
+
"step": 196
|
| 2755 |
+
},
|
| 2756 |
+
{
|
| 2757 |
+
"epoch": 1.9936628643852978,
|
| 2758 |
+
"step": 196,
|
| 2759 |
+
"total_flos": 0.0,
|
| 2760 |
+
"train_loss": 0.02676454978019156,
|
| 2761 |
+
"train_runtime": 31339.2496,
|
| 2762 |
+
"train_samples_per_second": 0.705,
|
| 2763 |
+
"train_steps_per_second": 0.006
|
| 2764 |
+
}
|
| 2765 |
+
],
|
| 2766 |
+
"logging_steps": 1,
|
| 2767 |
+
"max_steps": 196,
|
| 2768 |
+
"num_input_tokens_seen": 0,
|
| 2769 |
+
"num_train_epochs": 2,
|
| 2770 |
+
"save_steps": 500,
|
| 2771 |
+
"stateful_callbacks": {
|
| 2772 |
+
"TrainerControl": {
|
| 2773 |
+
"args": {
|
| 2774 |
+
"should_epoch_stop": false,
|
| 2775 |
+
"should_evaluate": false,
|
| 2776 |
+
"should_log": false,
|
| 2777 |
+
"should_save": true,
|
| 2778 |
+
"should_training_stop": true
|
| 2779 |
+
},
|
| 2780 |
+
"attributes": {}
|
| 2781 |
+
}
|
| 2782 |
+
},
|
| 2783 |
+
"total_flos": 0.0,
|
| 2784 |
+
"train_batch_size": 16,
|
| 2785 |
+
"trial_name": null,
|
| 2786 |
+
"trial_params": null
|
| 2787 |
+
}
|