Model save
Browse files- README.md +68 -0
- all_results.json +8 -0
- generation_config.json +9 -0
- train_results.json +8 -0
- trainer_state.json +2143 -0
README.md
ADDED
|
@@ -0,0 +1,68 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
base_model: deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B
|
| 3 |
+
library_name: transformers
|
| 4 |
+
model_name: GRPO
|
| 5 |
+
tags:
|
| 6 |
+
- generated_from_trainer
|
| 7 |
+
- trl
|
| 8 |
+
- grpo
|
| 9 |
+
licence: license
|
| 10 |
+
---
|
| 11 |
+
|
| 12 |
+
# Model Card for GRPO
|
| 13 |
+
|
| 14 |
+
This model is a fine-tuned version of [deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B](https://huggingface.co/deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B).
|
| 15 |
+
It has been trained using [TRL](https://github.com/huggingface/trl).
|
| 16 |
+
|
| 17 |
+
## Quick start
|
| 18 |
+
|
| 19 |
+
```python
|
| 20 |
+
from transformers import pipeline
|
| 21 |
+
|
| 22 |
+
question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
|
| 23 |
+
generator = pipeline("text-generation", model="LLucass/GRPO", device="cuda")
|
| 24 |
+
output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
|
| 25 |
+
print(output["generated_text"])
|
| 26 |
+
```
|
| 27 |
+
|
| 28 |
+
## Training procedure
|
| 29 |
+
|
| 30 |
+
[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/lavatorywang-nus/openr1/runs/h7vq56gq)
|
| 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: 0.18.0
|
| 38 |
+
- Transformers: 4.50.0
|
| 39 |
+
- Pytorch: 2.5.1
|
| 40 |
+
- Datasets: 3.6.0
|
| 41 |
+
- Tokenizers: 0.21.1
|
| 42 |
+
|
| 43 |
+
## Citations
|
| 44 |
+
|
| 45 |
+
Cite GRPO as:
|
| 46 |
+
|
| 47 |
+
```bibtex
|
| 48 |
+
@article{zhihong2024deepseekmath,
|
| 49 |
+
title = {{DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models}},
|
| 50 |
+
author = {Zhihong Shao and Peiyi Wang and Qihao Zhu and Runxin Xu and Junxiao Song and Mingchuan Zhang and Y. K. Li and Y. Wu and Daya Guo},
|
| 51 |
+
year = 2024,
|
| 52 |
+
eprint = {arXiv:2402.03300},
|
| 53 |
+
}
|
| 54 |
+
|
| 55 |
+
```
|
| 56 |
+
|
| 57 |
+
Cite TRL as:
|
| 58 |
+
|
| 59 |
+
```bibtex
|
| 60 |
+
@misc{vonwerra2022trl,
|
| 61 |
+
title = {{TRL: Transformer Reinforcement Learning}},
|
| 62 |
+
author = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang and Kashif Rasul and Quentin Gallouédec},
|
| 63 |
+
year = 2020,
|
| 64 |
+
journal = {GitHub repository},
|
| 65 |
+
publisher = {GitHub},
|
| 66 |
+
howpublished = {\url{https://github.com/huggingface/trl}}
|
| 67 |
+
}
|
| 68 |
+
```
|
all_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"total_flos": 0.0,
|
| 3 |
+
"train_loss": 0.00011944215420650531,
|
| 4 |
+
"train_runtime": 12092.6435,
|
| 5 |
+
"train_samples": 7000,
|
| 6 |
+
"train_samples_per_second": 1.191,
|
| 7 |
+
"train_steps_per_second": 0.012
|
| 8 |
+
}
|
generation_config.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"bos_token_id": 151646,
|
| 4 |
+
"do_sample": true,
|
| 5 |
+
"eos_token_id": 151643,
|
| 6 |
+
"temperature": 0.6,
|
| 7 |
+
"top_p": 0.95,
|
| 8 |
+
"transformers_version": "4.50.0"
|
| 9 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"total_flos": 0.0,
|
| 3 |
+
"train_loss": 0.00011944215420650531,
|
| 4 |
+
"train_runtime": 12092.6435,
|
| 5 |
+
"train_samples": 7000,
|
| 6 |
+
"train_samples_per_second": 1.191,
|
| 7 |
+
"train_steps_per_second": 0.012
|
| 8 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,2143 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_global_step": null,
|
| 3 |
+
"best_metric": null,
|
| 4 |
+
"best_model_checkpoint": null,
|
| 5 |
+
"epoch": 0.17142857142857143,
|
| 6 |
+
"eval_steps": 500,
|
| 7 |
+
"global_step": 150,
|
| 8 |
+
"is_hyper_param_search": false,
|
| 9 |
+
"is_local_process_zero": true,
|
| 10 |
+
"is_world_process_zero": true,
|
| 11 |
+
"log_history": [
|
| 12 |
+
{
|
| 13 |
+
"completion_length": 2693.6875610351562,
|
| 14 |
+
"entropy": 0.3662109375,
|
| 15 |
+
"epoch": 0.001142857142857143,
|
| 16 |
+
"grad_norm": 0.12395373731851578,
|
| 17 |
+
"kl": 0.0,
|
| 18 |
+
"learning_rate": 6.666666666666667e-08,
|
| 19 |
+
"loss": 0.0,
|
| 20 |
+
"reward": 0.7708333535119891,
|
| 21 |
+
"reward_std": 0.4629540964961052,
|
| 22 |
+
"rewards/accuracy_reward": 0.25000001303851604,
|
| 23 |
+
"rewards/format_reward": 0.5208333386108279,
|
| 24 |
+
"step": 1
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"completion_length": 3127.3958435058594,
|
| 28 |
+
"entropy": 0.353515625,
|
| 29 |
+
"epoch": 0.002285714285714286,
|
| 30 |
+
"grad_norm": 0.14846429228782654,
|
| 31 |
+
"kl": 0.0,
|
| 32 |
+
"learning_rate": 1.3333333333333334e-07,
|
| 33 |
+
"loss": 0.0,
|
| 34 |
+
"reward": 0.6458333637565374,
|
| 35 |
+
"reward_std": 0.4249730706214905,
|
| 36 |
+
"rewards/accuracy_reward": 0.2812500102445483,
|
| 37 |
+
"rewards/format_reward": 0.3645833386108279,
|
| 38 |
+
"step": 2
|
| 39 |
+
},
|
| 40 |
+
{
|
| 41 |
+
"completion_length": 3685.041748046875,
|
| 42 |
+
"entropy": 0.4443359375,
|
| 43 |
+
"epoch": 0.0034285714285714284,
|
| 44 |
+
"grad_norm": 0.10399040579795837,
|
| 45 |
+
"kl": 4.1425228118896484e-05,
|
| 46 |
+
"learning_rate": 2e-07,
|
| 47 |
+
"loss": 0.0,
|
| 48 |
+
"reward": 0.23958333674818277,
|
| 49 |
+
"reward_std": 0.3668827787041664,
|
| 50 |
+
"rewards/accuracy_reward": 0.0729166679084301,
|
| 51 |
+
"rewards/format_reward": 0.16666667256504297,
|
| 52 |
+
"step": 3
|
| 53 |
+
},
|
| 54 |
+
{
|
| 55 |
+
"completion_length": 2380.291778564453,
|
| 56 |
+
"entropy": 0.40478515625,
|
| 57 |
+
"epoch": 0.004571428571428572,
|
| 58 |
+
"grad_norm": 0.16352659463882446,
|
| 59 |
+
"kl": 3.409385681152344e-05,
|
| 60 |
+
"learning_rate": 2.6666666666666667e-07,
|
| 61 |
+
"loss": 0.0,
|
| 62 |
+
"reward": 0.8229166865348816,
|
| 63 |
+
"reward_std": 0.507609948515892,
|
| 64 |
+
"rewards/accuracy_reward": 0.19791667722165585,
|
| 65 |
+
"rewards/format_reward": 0.6250000223517418,
|
| 66 |
+
"step": 4
|
| 67 |
+
},
|
| 68 |
+
{
|
| 69 |
+
"completion_length": 3441.2188720703125,
|
| 70 |
+
"entropy": 0.45458984375,
|
| 71 |
+
"epoch": 0.005714285714285714,
|
| 72 |
+
"grad_norm": 0.15812984108924866,
|
| 73 |
+
"kl": 4.1961669921875e-05,
|
| 74 |
+
"learning_rate": 3.333333333333333e-07,
|
| 75 |
+
"loss": 0.0,
|
| 76 |
+
"reward": 0.42708334885537624,
|
| 77 |
+
"reward_std": 0.5058739930391312,
|
| 78 |
+
"rewards/accuracy_reward": 0.07291666697710752,
|
| 79 |
+
"rewards/format_reward": 0.35416667722165585,
|
| 80 |
+
"step": 5
|
| 81 |
+
},
|
| 82 |
+
{
|
| 83 |
+
"completion_length": 3382.3438110351562,
|
| 84 |
+
"entropy": 0.45166015625,
|
| 85 |
+
"epoch": 0.006857142857142857,
|
| 86 |
+
"grad_norm": 0.15454305708408356,
|
| 87 |
+
"kl": 4.26173210144043e-05,
|
| 88 |
+
"learning_rate": 4e-07,
|
| 89 |
+
"loss": 0.0,
|
| 90 |
+
"reward": 0.40625000558793545,
|
| 91 |
+
"reward_std": 0.5202516540884972,
|
| 92 |
+
"rewards/accuracy_reward": 0.0833333358168602,
|
| 93 |
+
"rewards/format_reward": 0.3229166744276881,
|
| 94 |
+
"step": 6
|
| 95 |
+
},
|
| 96 |
+
{
|
| 97 |
+
"completion_length": 3277.291748046875,
|
| 98 |
+
"entropy": 0.39404296875,
|
| 99 |
+
"epoch": 0.008,
|
| 100 |
+
"grad_norm": 0.13690507411956787,
|
| 101 |
+
"kl": 2.562999725341797e-05,
|
| 102 |
+
"learning_rate": 4.6666666666666666e-07,
|
| 103 |
+
"loss": 0.0,
|
| 104 |
+
"reward": 0.8854166865348816,
|
| 105 |
+
"reward_std": 0.6845719665288925,
|
| 106 |
+
"rewards/accuracy_reward": 0.2708333432674408,
|
| 107 |
+
"rewards/format_reward": 0.6145833432674408,
|
| 108 |
+
"step": 7
|
| 109 |
+
},
|
| 110 |
+
{
|
| 111 |
+
"completion_length": 2841.916748046875,
|
| 112 |
+
"entropy": 0.36083984375,
|
| 113 |
+
"epoch": 0.009142857142857144,
|
| 114 |
+
"grad_norm": 0.1767321527004242,
|
| 115 |
+
"kl": 2.4050474166870117e-05,
|
| 116 |
+
"learning_rate": 5.333333333333333e-07,
|
| 117 |
+
"loss": 0.0,
|
| 118 |
+
"reward": 0.8854166967794299,
|
| 119 |
+
"reward_std": 0.3672378845512867,
|
| 120 |
+
"rewards/accuracy_reward": 0.3958333535119891,
|
| 121 |
+
"rewards/format_reward": 0.4895833460614085,
|
| 122 |
+
"step": 8
|
| 123 |
+
},
|
| 124 |
+
{
|
| 125 |
+
"completion_length": 3480.6563110351562,
|
| 126 |
+
"entropy": 0.4384765625,
|
| 127 |
+
"epoch": 0.010285714285714285,
|
| 128 |
+
"grad_norm": 0.15406936407089233,
|
| 129 |
+
"kl": 3.796815872192383e-05,
|
| 130 |
+
"learning_rate": 6e-07,
|
| 131 |
+
"loss": 0.0,
|
| 132 |
+
"reward": 0.5520833432674408,
|
| 133 |
+
"reward_std": 0.6496799141168594,
|
| 134 |
+
"rewards/accuracy_reward": 0.17708333488553762,
|
| 135 |
+
"rewards/format_reward": 0.3750000074505806,
|
| 136 |
+
"step": 9
|
| 137 |
+
},
|
| 138 |
+
{
|
| 139 |
+
"completion_length": 2963.572967529297,
|
| 140 |
+
"entropy": 0.3544921875,
|
| 141 |
+
"epoch": 0.011428571428571429,
|
| 142 |
+
"grad_norm": 0.15688633918762207,
|
| 143 |
+
"kl": 2.5287270545959473e-05,
|
| 144 |
+
"learning_rate": 6.666666666666666e-07,
|
| 145 |
+
"loss": 0.0,
|
| 146 |
+
"reward": 0.5937500223517418,
|
| 147 |
+
"reward_std": 0.5099271312355995,
|
| 148 |
+
"rewards/accuracy_reward": 0.17708333861082792,
|
| 149 |
+
"rewards/format_reward": 0.4166666753590107,
|
| 150 |
+
"step": 10
|
| 151 |
+
},
|
| 152 |
+
{
|
| 153 |
+
"completion_length": 3573.7500610351562,
|
| 154 |
+
"entropy": 0.37890625,
|
| 155 |
+
"epoch": 0.012571428571428572,
|
| 156 |
+
"grad_norm": 0.12983083724975586,
|
| 157 |
+
"kl": 2.5391578674316406e-05,
|
| 158 |
+
"learning_rate": 7.333333333333332e-07,
|
| 159 |
+
"loss": 0.0,
|
| 160 |
+
"reward": 0.3125000111758709,
|
| 161 |
+
"reward_std": 0.5802810192108154,
|
| 162 |
+
"rewards/accuracy_reward": 0.1041666679084301,
|
| 163 |
+
"rewards/format_reward": 0.20833334140479565,
|
| 164 |
+
"step": 11
|
| 165 |
+
},
|
| 166 |
+
{
|
| 167 |
+
"completion_length": 2520.8958740234375,
|
| 168 |
+
"entropy": 0.39111328125,
|
| 169 |
+
"epoch": 0.013714285714285714,
|
| 170 |
+
"grad_norm": 0.20449091494083405,
|
| 171 |
+
"kl": 3.743171691894531e-05,
|
| 172 |
+
"learning_rate": 8e-07,
|
| 173 |
+
"loss": 0.0,
|
| 174 |
+
"reward": 0.8020833656191826,
|
| 175 |
+
"reward_std": 0.4411254972219467,
|
| 176 |
+
"rewards/accuracy_reward": 0.14583333395421505,
|
| 177 |
+
"rewards/format_reward": 0.6562500223517418,
|
| 178 |
+
"step": 12
|
| 179 |
+
},
|
| 180 |
+
{
|
| 181 |
+
"completion_length": 3038.041748046875,
|
| 182 |
+
"entropy": 0.3828125,
|
| 183 |
+
"epoch": 0.014857142857142857,
|
| 184 |
+
"grad_norm": 0.14574433863162994,
|
| 185 |
+
"kl": 2.5153160095214844e-05,
|
| 186 |
+
"learning_rate": 8.666666666666667e-07,
|
| 187 |
+
"loss": 0.0,
|
| 188 |
+
"reward": 0.6875000298023224,
|
| 189 |
+
"reward_std": 0.3254704251885414,
|
| 190 |
+
"rewards/accuracy_reward": 0.22916666697710752,
|
| 191 |
+
"rewards/format_reward": 0.4583333432674408,
|
| 192 |
+
"step": 13
|
| 193 |
+
},
|
| 194 |
+
{
|
| 195 |
+
"completion_length": 3116.3125610351562,
|
| 196 |
+
"entropy": 0.37109375,
|
| 197 |
+
"epoch": 0.016,
|
| 198 |
+
"grad_norm": 0.202586367726326,
|
| 199 |
+
"kl": 1.9026920199394226e-05,
|
| 200 |
+
"learning_rate": 9.333333333333333e-07,
|
| 201 |
+
"loss": 0.0,
|
| 202 |
+
"reward": 0.5833333507180214,
|
| 203 |
+
"reward_std": 0.4630111753940582,
|
| 204 |
+
"rewards/accuracy_reward": 0.21875001024454832,
|
| 205 |
+
"rewards/format_reward": 0.3645833395421505,
|
| 206 |
+
"step": 14
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"completion_length": 2924.0521240234375,
|
| 210 |
+
"entropy": 0.36328125,
|
| 211 |
+
"epoch": 0.017142857142857144,
|
| 212 |
+
"grad_norm": 0.09130721539258957,
|
| 213 |
+
"kl": 1.4469027519226074e-05,
|
| 214 |
+
"learning_rate": 1e-06,
|
| 215 |
+
"loss": 0.0,
|
| 216 |
+
"reward": 0.604166679084301,
|
| 217 |
+
"reward_std": 0.22134994342923164,
|
| 218 |
+
"rewards/accuracy_reward": 0.1979166716337204,
|
| 219 |
+
"rewards/format_reward": 0.4062500074505806,
|
| 220 |
+
"step": 15
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"completion_length": 3887.5521850585938,
|
| 224 |
+
"entropy": 0.4755859375,
|
| 225 |
+
"epoch": 0.018285714285714287,
|
| 226 |
+
"grad_norm": 0.11806491017341614,
|
| 227 |
+
"kl": 2.9832124710083008e-05,
|
| 228 |
+
"learning_rate": 9.998781585307575e-07,
|
| 229 |
+
"loss": 0.0,
|
| 230 |
+
"reward": 0.11458333674818277,
|
| 231 |
+
"reward_std": 0.26997610181570053,
|
| 232 |
+
"rewards/accuracy_reward": 0.041666666977107525,
|
| 233 |
+
"rewards/format_reward": 0.07291666977107525,
|
| 234 |
+
"step": 16
|
| 235 |
+
},
|
| 236 |
+
{
|
| 237 |
+
"completion_length": 2579.625030517578,
|
| 238 |
+
"entropy": 0.44091796875,
|
| 239 |
+
"epoch": 0.019428571428571427,
|
| 240 |
+
"grad_norm": 0.2032601535320282,
|
| 241 |
+
"kl": 3.93986701965332e-05,
|
| 242 |
+
"learning_rate": 9.99512700102336e-07,
|
| 243 |
+
"loss": 0.0,
|
| 244 |
+
"reward": 0.7083333507180214,
|
| 245 |
+
"reward_std": 0.39187028259038925,
|
| 246 |
+
"rewards/accuracy_reward": 0.19791667442768812,
|
| 247 |
+
"rewards/format_reward": 0.5104166753590107,
|
| 248 |
+
"step": 17
|
| 249 |
+
},
|
| 250 |
+
{
|
| 251 |
+
"completion_length": 3089.104248046875,
|
| 252 |
+
"entropy": 0.3671875,
|
| 253 |
+
"epoch": 0.02057142857142857,
|
| 254 |
+
"grad_norm": 0.11376938223838806,
|
| 255 |
+
"kl": 1.2531876564025879e-05,
|
| 256 |
+
"learning_rate": 9.989038226169207e-07,
|
| 257 |
+
"loss": 0.0,
|
| 258 |
+
"reward": 0.5625000251457095,
|
| 259 |
+
"reward_std": 0.35285963863134384,
|
| 260 |
+
"rewards/accuracy_reward": 0.1666666679084301,
|
| 261 |
+
"rewards/format_reward": 0.3958333386108279,
|
| 262 |
+
"step": 18
|
| 263 |
+
},
|
| 264 |
+
{
|
| 265 |
+
"completion_length": 3130.760498046875,
|
| 266 |
+
"entropy": 0.39111328125,
|
| 267 |
+
"epoch": 0.021714285714285714,
|
| 268 |
+
"grad_norm": 0.09636794775724411,
|
| 269 |
+
"kl": 2.8267502784729004e-05,
|
| 270 |
+
"learning_rate": 9.98051855792412e-07,
|
| 271 |
+
"loss": 0.0,
|
| 272 |
+
"reward": 0.8125000111758709,
|
| 273 |
+
"reward_std": 0.3496965616941452,
|
| 274 |
+
"rewards/accuracy_reward": 0.36458333395421505,
|
| 275 |
+
"rewards/format_reward": 0.44791667722165585,
|
| 276 |
+
"step": 19
|
| 277 |
+
},
|
| 278 |
+
{
|
| 279 |
+
"completion_length": 2585.9896545410156,
|
| 280 |
+
"entropy": 0.329833984375,
|
| 281 |
+
"epoch": 0.022857142857142857,
|
| 282 |
+
"grad_norm": 0.15105831623077393,
|
| 283 |
+
"kl": 6.628036499023438e-05,
|
| 284 |
+
"learning_rate": 9.969572609838744e-07,
|
| 285 |
+
"loss": 0.0,
|
| 286 |
+
"reward": 0.9791666716337204,
|
| 287 |
+
"reward_std": 0.3452813923358917,
|
| 288 |
+
"rewards/accuracy_reward": 0.2812500037252903,
|
| 289 |
+
"rewards/format_reward": 0.6979166716337204,
|
| 290 |
+
"step": 20
|
| 291 |
+
},
|
| 292 |
+
{
|
| 293 |
+
"completion_length": 2804.229248046875,
|
| 294 |
+
"entropy": 0.42578125,
|
| 295 |
+
"epoch": 0.024,
|
| 296 |
+
"grad_norm": 0.2109123021364212,
|
| 297 |
+
"kl": 0.00016552209854125977,
|
| 298 |
+
"learning_rate": 9.956206309337066e-07,
|
| 299 |
+
"loss": 0.0,
|
| 300 |
+
"reward": 0.6145833432674408,
|
| 301 |
+
"reward_std": 0.4177238382399082,
|
| 302 |
+
"rewards/accuracy_reward": 0.1458333395421505,
|
| 303 |
+
"rewards/format_reward": 0.46875002048909664,
|
| 304 |
+
"step": 21
|
| 305 |
+
},
|
| 306 |
+
{
|
| 307 |
+
"completion_length": 1903.1459045410156,
|
| 308 |
+
"entropy": 0.419921875,
|
| 309 |
+
"epoch": 0.025142857142857144,
|
| 310 |
+
"grad_norm": 0.20996998250484467,
|
| 311 |
+
"kl": 0.00026351213455200195,
|
| 312 |
+
"learning_rate": 9.940426894506606e-07,
|
| 313 |
+
"loss": 0.0,
|
| 314 |
+
"reward": 1.1041667014360428,
|
| 315 |
+
"reward_std": 0.4033822976052761,
|
| 316 |
+
"rewards/accuracy_reward": 0.29166667722165585,
|
| 317 |
+
"rewards/format_reward": 0.8125000149011612,
|
| 318 |
+
"step": 22
|
| 319 |
+
},
|
| 320 |
+
{
|
| 321 |
+
"completion_length": 2714.0729370117188,
|
| 322 |
+
"entropy": 0.36865234375,
|
| 323 |
+
"epoch": 0.026285714285714287,
|
| 324 |
+
"grad_norm": 0.16544093191623688,
|
| 325 |
+
"kl": 0.00011658668518066406,
|
| 326 |
+
"learning_rate": 9.922242910178859e-07,
|
| 327 |
+
"loss": 0.0,
|
| 328 |
+
"reward": 0.6770833507180214,
|
| 329 |
+
"reward_std": 0.6271640285849571,
|
| 330 |
+
"rewards/accuracy_reward": 0.1770833432674408,
|
| 331 |
+
"rewards/format_reward": 0.5000000223517418,
|
| 332 |
+
"step": 23
|
| 333 |
+
},
|
| 334 |
+
{
|
| 335 |
+
"completion_length": 2834.2396850585938,
|
| 336 |
+
"entropy": 0.373046875,
|
| 337 |
+
"epoch": 0.027428571428571427,
|
| 338 |
+
"grad_norm": 0.10939397662878036,
|
| 339 |
+
"kl": 0.0001084059476852417,
|
| 340 |
+
"learning_rate": 9.901664203302124e-07,
|
| 341 |
+
"loss": 0.0,
|
| 342 |
+
"reward": 0.7916666865348816,
|
| 343 |
+
"reward_std": 0.5711240321397781,
|
| 344 |
+
"rewards/accuracy_reward": 0.2187500074505806,
|
| 345 |
+
"rewards/format_reward": 0.572916679084301,
|
| 346 |
+
"step": 24
|
| 347 |
+
},
|
| 348 |
+
{
|
| 349 |
+
"completion_length": 2877.1354370117188,
|
| 350 |
+
"entropy": 0.4296875,
|
| 351 |
+
"epoch": 0.02857142857142857,
|
| 352 |
+
"grad_norm": 0.10193013399839401,
|
| 353 |
+
"kl": 0.00018364191055297852,
|
| 354 |
+
"learning_rate": 9.878701917609207e-07,
|
| 355 |
+
"loss": 0.0,
|
| 356 |
+
"reward": 0.677083358168602,
|
| 357 |
+
"reward_std": 0.2898401468992233,
|
| 358 |
+
"rewards/accuracy_reward": 0.2395833432674408,
|
| 359 |
+
"rewards/format_reward": 0.4375,
|
| 360 |
+
"step": 25
|
| 361 |
+
},
|
| 362 |
+
{
|
| 363 |
+
"completion_length": 3221.2396850585938,
|
| 364 |
+
"entropy": 0.4248046875,
|
| 365 |
+
"epoch": 0.029714285714285714,
|
| 366 |
+
"grad_norm": 0.07458896934986115,
|
| 367 |
+
"kl": 3.0487775802612305e-05,
|
| 368 |
+
"learning_rate": 9.853368487582886e-07,
|
| 369 |
+
"loss": 0.0,
|
| 370 |
+
"reward": 0.6562500149011612,
|
| 371 |
+
"reward_std": 0.25371449440717697,
|
| 372 |
+
"rewards/accuracy_reward": 0.19791666977107525,
|
| 373 |
+
"rewards/format_reward": 0.4583333358168602,
|
| 374 |
+
"step": 26
|
| 375 |
+
},
|
| 376 |
+
{
|
| 377 |
+
"completion_length": 3297.3959350585938,
|
| 378 |
+
"entropy": 0.45703125,
|
| 379 |
+
"epoch": 0.030857142857142857,
|
| 380 |
+
"grad_norm": 0.0925775095820427,
|
| 381 |
+
"kl": 0.00012201815843582153,
|
| 382 |
+
"learning_rate": 9.825677631722435e-07,
|
| 383 |
+
"loss": 0.0,
|
| 384 |
+
"reward": 0.541666679084301,
|
| 385 |
+
"reward_std": 0.4426998719573021,
|
| 386 |
+
"rewards/accuracy_reward": 0.15625000279396772,
|
| 387 |
+
"rewards/format_reward": 0.385416679084301,
|
| 388 |
+
"step": 27
|
| 389 |
+
},
|
| 390 |
+
{
|
| 391 |
+
"completion_length": 2984.2188110351562,
|
| 392 |
+
"entropy": 0.3994140625,
|
| 393 |
+
"epoch": 0.032,
|
| 394 |
+
"grad_norm": 0.12723609805107117,
|
| 395 |
+
"kl": 0.00015980005264282227,
|
| 396 |
+
"learning_rate": 9.795644345114794e-07,
|
| 397 |
+
"loss": 0.0,
|
| 398 |
+
"reward": 0.8437500447034836,
|
| 399 |
+
"reward_std": 0.48607436567544937,
|
| 400 |
+
"rewards/accuracy_reward": 0.3333333460614085,
|
| 401 |
+
"rewards/format_reward": 0.5104166865348816,
|
| 402 |
+
"step": 28
|
| 403 |
+
},
|
| 404 |
+
{
|
| 405 |
+
"completion_length": 3707.2501220703125,
|
| 406 |
+
"entropy": 0.43408203125,
|
| 407 |
+
"epoch": 0.03314285714285714,
|
| 408 |
+
"grad_norm": 0.20130394399166107,
|
| 409 |
+
"kl": 0.0003858804702758789,
|
| 410 |
+
"learning_rate": 9.76328489131448e-07,
|
| 411 |
+
"loss": 0.0,
|
| 412 |
+
"reward": 0.2500000027939677,
|
| 413 |
+
"reward_std": 0.37919554859399796,
|
| 414 |
+
"rewards/accuracy_reward": 0.06250000186264515,
|
| 415 |
+
"rewards/format_reward": 0.18750001024454832,
|
| 416 |
+
"step": 29
|
| 417 |
+
},
|
| 418 |
+
{
|
| 419 |
+
"completion_length": 3099.9063110351562,
|
| 420 |
+
"entropy": 0.384765625,
|
| 421 |
+
"epoch": 0.03428571428571429,
|
| 422 |
+
"grad_norm": 0.13564985990524292,
|
| 423 |
+
"kl": 0.0005750656127929688,
|
| 424 |
+
"learning_rate": 9.728616793536587e-07,
|
| 425 |
+
"loss": 0.0,
|
| 426 |
+
"reward": 0.8854166828095913,
|
| 427 |
+
"reward_std": 0.5532158613204956,
|
| 428 |
+
"rewards/accuracy_reward": 0.3125000009313226,
|
| 429 |
+
"rewards/format_reward": 0.572916679084301,
|
| 430 |
+
"step": 30
|
| 431 |
+
},
|
| 432 |
+
{
|
| 433 |
+
"completion_length": 3310.8125610351562,
|
| 434 |
+
"entropy": 0.40087890625,
|
| 435 |
+
"epoch": 0.03542857142857143,
|
| 436 |
+
"grad_norm": 0.14703762531280518,
|
| 437 |
+
"kl": 0.0007152557373046875,
|
| 438 |
+
"learning_rate": 9.69165882516764e-07,
|
| 439 |
+
"loss": 0.0,
|
| 440 |
+
"reward": 0.4583333469927311,
|
| 441 |
+
"reward_std": 0.4937985762953758,
|
| 442 |
+
"rewards/accuracy_reward": 0.16666667070239782,
|
| 443 |
+
"rewards/format_reward": 0.2916666716337204,
|
| 444 |
+
"step": 31
|
| 445 |
+
},
|
| 446 |
+
{
|
| 447 |
+
"completion_length": 3543.666748046875,
|
| 448 |
+
"entropy": 0.4521484375,
|
| 449 |
+
"epoch": 0.036571428571428574,
|
| 450 |
+
"grad_norm": 0.11301636695861816,
|
| 451 |
+
"kl": 0.00032842159271240234,
|
| 452 |
+
"learning_rate": 9.65243099959949e-07,
|
| 453 |
+
"loss": 0.0,
|
| 454 |
+
"reward": 0.6250000223517418,
|
| 455 |
+
"reward_std": 0.46596524864435196,
|
| 456 |
+
"rewards/accuracy_reward": 0.28125000558793545,
|
| 457 |
+
"rewards/format_reward": 0.34375,
|
| 458 |
+
"step": 32
|
| 459 |
+
},
|
| 460 |
+
{
|
| 461 |
+
"completion_length": 3395.947998046875,
|
| 462 |
+
"entropy": 0.384765625,
|
| 463 |
+
"epoch": 0.037714285714285714,
|
| 464 |
+
"grad_norm": 0.12107253074645996,
|
| 465 |
+
"kl": 0.00042450428009033203,
|
| 466 |
+
"learning_rate": 9.610954559391704e-07,
|
| 467 |
+
"loss": 0.0,
|
| 468 |
+
"reward": 0.604166679084301,
|
| 469 |
+
"reward_std": 0.5497709587216377,
|
| 470 |
+
"rewards/accuracy_reward": 0.18750000093132257,
|
| 471 |
+
"rewards/format_reward": 0.41666667722165585,
|
| 472 |
+
"step": 33
|
| 473 |
+
},
|
| 474 |
+
{
|
| 475 |
+
"completion_length": 2621.218780517578,
|
| 476 |
+
"entropy": 0.45263671875,
|
| 477 |
+
"epoch": 0.038857142857142854,
|
| 478 |
+
"grad_norm": 0.15317150950431824,
|
| 479 |
+
"kl": 0.0013637542724609375,
|
| 480 |
+
"learning_rate": 9.567251964768342e-07,
|
| 481 |
+
"loss": 0.0001,
|
| 482 |
+
"reward": 0.8541666865348816,
|
| 483 |
+
"reward_std": 0.4670567326247692,
|
| 484 |
+
"rewards/accuracy_reward": 0.31250001303851604,
|
| 485 |
+
"rewards/format_reward": 0.5416666828095913,
|
| 486 |
+
"step": 34
|
| 487 |
+
},
|
| 488 |
+
{
|
| 489 |
+
"completion_length": 3166.3958740234375,
|
| 490 |
+
"entropy": 0.43115234375,
|
| 491 |
+
"epoch": 0.04,
|
| 492 |
+
"grad_norm": 0.1469903290271759,
|
| 493 |
+
"kl": 0.0011509060859680176,
|
| 494 |
+
"learning_rate": 9.521346881455354e-07,
|
| 495 |
+
"loss": 0.0,
|
| 496 |
+
"reward": 0.6458333656191826,
|
| 497 |
+
"reward_std": 0.6130613833665848,
|
| 498 |
+
"rewards/accuracy_reward": 0.23958333767950535,
|
| 499 |
+
"rewards/format_reward": 0.4062500149011612,
|
| 500 |
+
"step": 35
|
| 501 |
+
},
|
| 502 |
+
{
|
| 503 |
+
"completion_length": 3509.697998046875,
|
| 504 |
+
"entropy": 0.513671875,
|
| 505 |
+
"epoch": 0.04114285714285714,
|
| 506 |
+
"grad_norm": 0.11033376306295395,
|
| 507 |
+
"kl": 0.0011191368103027344,
|
| 508 |
+
"learning_rate": 9.473264167865171e-07,
|
| 509 |
+
"loss": 0.0,
|
| 510 |
+
"reward": 0.23958333395421505,
|
| 511 |
+
"reward_std": 0.24118434637784958,
|
| 512 |
+
"rewards/accuracy_reward": 0.031250000931322575,
|
| 513 |
+
"rewards/format_reward": 0.20833334140479565,
|
| 514 |
+
"step": 36
|
| 515 |
+
},
|
| 516 |
+
{
|
| 517 |
+
"completion_length": 3363.8333740234375,
|
| 518 |
+
"entropy": 0.42138671875,
|
| 519 |
+
"epoch": 0.04228571428571429,
|
| 520 |
+
"grad_norm": 0.11778294295072556,
|
| 521 |
+
"kl": 0.0008115768432617188,
|
| 522 |
+
"learning_rate": 9.42302986163543e-07,
|
| 523 |
+
"loss": 0.0,
|
| 524 |
+
"reward": 0.2812500149011612,
|
| 525 |
+
"reward_std": 0.13804075866937637,
|
| 526 |
+
"rewards/accuracy_reward": 0.031250000931322575,
|
| 527 |
+
"rewards/format_reward": 0.25,
|
| 528 |
+
"step": 37
|
| 529 |
+
},
|
| 530 |
+
{
|
| 531 |
+
"completion_length": 3610.8438110351562,
|
| 532 |
+
"entropy": 0.44677734375,
|
| 533 |
+
"epoch": 0.04342857142857143,
|
| 534 |
+
"grad_norm": 0.061884235590696335,
|
| 535 |
+
"kl": 0.0005736351013183594,
|
| 536 |
+
"learning_rate": 9.370671165529144e-07,
|
| 537 |
+
"loss": 0.0,
|
| 538 |
+
"reward": 0.21875000558793545,
|
| 539 |
+
"reward_std": 0.17128896713256836,
|
| 540 |
+
"rewards/accuracy_reward": 0.10416666697710752,
|
| 541 |
+
"rewards/format_reward": 0.11458333861082792,
|
| 542 |
+
"step": 38
|
| 543 |
+
},
|
| 544 |
+
{
|
| 545 |
+
"completion_length": 2926.4063110351562,
|
| 546 |
+
"entropy": 0.36669921875,
|
| 547 |
+
"epoch": 0.044571428571428574,
|
| 548 |
+
"grad_norm": 0.1068028062582016,
|
| 549 |
+
"kl": 0.0011527538299560547,
|
| 550 |
+
"learning_rate": 9.316216432703916e-07,
|
| 551 |
+
"loss": 0.0,
|
| 552 |
+
"reward": 0.7708333656191826,
|
| 553 |
+
"reward_std": 0.1930682435631752,
|
| 554 |
+
"rewards/accuracy_reward": 0.25,
|
| 555 |
+
"rewards/format_reward": 0.5208333507180214,
|
| 556 |
+
"step": 39
|
| 557 |
+
},
|
| 558 |
+
{
|
| 559 |
+
"completion_length": 2785.1146545410156,
|
| 560 |
+
"entropy": 0.388671875,
|
| 561 |
+
"epoch": 0.045714285714285714,
|
| 562 |
+
"grad_norm": 0.17572174966335297,
|
| 563 |
+
"kl": 0.0032024383544921875,
|
| 564 |
+
"learning_rate": 9.259695151358214e-07,
|
| 565 |
+
"loss": 0.0001,
|
| 566 |
+
"reward": 0.7291666902601719,
|
| 567 |
+
"reward_std": 0.3721684589982033,
|
| 568 |
+
"rewards/accuracy_reward": 0.1979166716337204,
|
| 569 |
+
"rewards/format_reward": 0.5312500186264515,
|
| 570 |
+
"step": 40
|
| 571 |
+
},
|
| 572 |
+
{
|
| 573 |
+
"completion_length": 3123.947998046875,
|
| 574 |
+
"entropy": 0.35791015625,
|
| 575 |
+
"epoch": 0.046857142857142854,
|
| 576 |
+
"grad_norm": 0.144905224442482,
|
| 577 |
+
"kl": 0.0007574558258056641,
|
| 578 |
+
"learning_rate": 9.20113792876298e-07,
|
| 579 |
+
"loss": 0.0,
|
| 580 |
+
"reward": 0.5416666865348816,
|
| 581 |
+
"reward_std": 0.4893290549516678,
|
| 582 |
+
"rewards/accuracy_reward": 0.12500000651925802,
|
| 583 |
+
"rewards/format_reward": 0.416666679084301,
|
| 584 |
+
"step": 41
|
| 585 |
+
},
|
| 586 |
+
{
|
| 587 |
+
"completion_length": 3056.197998046875,
|
| 588 |
+
"entropy": 0.48193359375,
|
| 589 |
+
"epoch": 0.048,
|
| 590 |
+
"grad_norm": 0.07001210004091263,
|
| 591 |
+
"kl": 0.000598907470703125,
|
| 592 |
+
"learning_rate": 9.140576474687263e-07,
|
| 593 |
+
"loss": 0.0,
|
| 594 |
+
"reward": 0.30208333395421505,
|
| 595 |
+
"reward_std": 0.15690935403108597,
|
| 596 |
+
"rewards/accuracy_reward": 0.02083333395421505,
|
| 597 |
+
"rewards/format_reward": 0.2812500009313226,
|
| 598 |
+
"step": 42
|
| 599 |
+
},
|
| 600 |
+
{
|
| 601 |
+
"completion_length": 3097.479248046875,
|
| 602 |
+
"entropy": 0.4033203125,
|
| 603 |
+
"epoch": 0.04914285714285714,
|
| 604 |
+
"grad_norm": 0.09348543733358383,
|
| 605 |
+
"kl": 0.001148223876953125,
|
| 606 |
+
"learning_rate": 9.078043584226815e-07,
|
| 607 |
+
"loss": 0.0,
|
| 608 |
+
"reward": 0.4895833358168602,
|
| 609 |
+
"reward_std": 0.3020758181810379,
|
| 610 |
+
"rewards/accuracy_reward": 0.1666666679084301,
|
| 611 |
+
"rewards/format_reward": 0.32291666977107525,
|
| 612 |
+
"step": 43
|
| 613 |
+
},
|
| 614 |
+
{
|
| 615 |
+
"completion_length": 2797.7084197998047,
|
| 616 |
+
"entropy": 0.39013671875,
|
| 617 |
+
"epoch": 0.05028571428571429,
|
| 618 |
+
"grad_norm": 0.15556485950946808,
|
| 619 |
+
"kl": 0.0014767646789550781,
|
| 620 |
+
"learning_rate": 9.013573120044966e-07,
|
| 621 |
+
"loss": 0.0001,
|
| 622 |
+
"reward": 0.8020833386108279,
|
| 623 |
+
"reward_std": 0.3607782945036888,
|
| 624 |
+
"rewards/accuracy_reward": 0.2708333395421505,
|
| 625 |
+
"rewards/format_reward": 0.5312500102445483,
|
| 626 |
+
"step": 44
|
| 627 |
+
},
|
| 628 |
+
{
|
| 629 |
+
"completion_length": 3618.791748046875,
|
| 630 |
+
"entropy": 0.423828125,
|
| 631 |
+
"epoch": 0.05142857142857143,
|
| 632 |
+
"grad_norm": 0.09923144429922104,
|
| 633 |
+
"kl": 0.0026645660400390625,
|
| 634 |
+
"learning_rate": 8.9471999940354e-07,
|
| 635 |
+
"loss": 0.0001,
|
| 636 |
+
"reward": 0.5833333348855376,
|
| 637 |
+
"reward_std": 0.4189528524875641,
|
| 638 |
+
"rewards/accuracy_reward": 0.2604166716337204,
|
| 639 |
+
"rewards/format_reward": 0.3229166781529784,
|
| 640 |
+
"step": 45
|
| 641 |
+
},
|
| 642 |
+
{
|
| 643 |
+
"completion_length": 3482.479248046875,
|
| 644 |
+
"entropy": 0.50634765625,
|
| 645 |
+
"epoch": 0.052571428571428575,
|
| 646 |
+
"grad_norm": 0.12269324064254761,
|
| 647 |
+
"kl": 0.0013875961303710938,
|
| 648 |
+
"learning_rate": 8.878960148416747e-07,
|
| 649 |
+
"loss": 0.0001,
|
| 650 |
+
"reward": 0.22916667722165585,
|
| 651 |
+
"reward_std": 0.26679350435733795,
|
| 652 |
+
"rewards/accuracy_reward": 0.0416666679084301,
|
| 653 |
+
"rewards/format_reward": 0.18750000186264515,
|
| 654 |
+
"step": 46
|
| 655 |
+
},
|
| 656 |
+
{
|
| 657 |
+
"completion_length": 2958.291748046875,
|
| 658 |
+
"entropy": 0.390625,
|
| 659 |
+
"epoch": 0.053714285714285714,
|
| 660 |
+
"grad_norm": 0.16963143646717072,
|
| 661 |
+
"kl": 0.0011917352676391602,
|
| 662 |
+
"learning_rate": 8.808890536269229e-07,
|
| 663 |
+
"loss": 0.0,
|
| 664 |
+
"reward": 0.8854166967794299,
|
| 665 |
+
"reward_std": 0.5451135858893394,
|
| 666 |
+
"rewards/accuracy_reward": 0.3437500149011612,
|
| 667 |
+
"rewards/format_reward": 0.5416666669771075,
|
| 668 |
+
"step": 47
|
| 669 |
+
},
|
| 670 |
+
{
|
| 671 |
+
"completion_length": 2956.416717529297,
|
| 672 |
+
"entropy": 0.396484375,
|
| 673 |
+
"epoch": 0.054857142857142854,
|
| 674 |
+
"grad_norm": 0.14105089008808136,
|
| 675 |
+
"kl": 0.0033426284790039062,
|
| 676 |
+
"learning_rate": 8.737029101523929e-07,
|
| 677 |
+
"loss": 0.0001,
|
| 678 |
+
"reward": 0.7395833507180214,
|
| 679 |
+
"reward_std": 0.5457281768321991,
|
| 680 |
+
"rewards/accuracy_reward": 0.29166666977107525,
|
| 681 |
+
"rewards/format_reward": 0.4479166679084301,
|
| 682 |
+
"step": 48
|
| 683 |
+
},
|
| 684 |
+
{
|
| 685 |
+
"completion_length": 2448.1146850585938,
|
| 686 |
+
"entropy": 0.36865234375,
|
| 687 |
+
"epoch": 0.056,
|
| 688 |
+
"grad_norm": 0.15970121324062347,
|
| 689 |
+
"kl": 0.006764888763427734,
|
| 690 |
+
"learning_rate": 8.663414758415478e-07,
|
| 691 |
+
"loss": 0.0003,
|
| 692 |
+
"reward": 0.895833395421505,
|
| 693 |
+
"reward_std": 0.464010052382946,
|
| 694 |
+
"rewards/accuracy_reward": 0.25000000558793545,
|
| 695 |
+
"rewards/format_reward": 0.6458333507180214,
|
| 696 |
+
"step": 49
|
| 697 |
+
},
|
| 698 |
+
{
|
| 699 |
+
"completion_length": 3050.1041870117188,
|
| 700 |
+
"entropy": 0.34521484375,
|
| 701 |
+
"epoch": 0.05714285714285714,
|
| 702 |
+
"grad_norm": 0.12386268377304077,
|
| 703 |
+
"kl": 0.0011911392211914062,
|
| 704 |
+
"learning_rate": 8.588087370409302e-07,
|
| 705 |
+
"loss": 0.0,
|
| 706 |
+
"reward": 0.6562500325962901,
|
| 707 |
+
"reward_std": 0.4570996016263962,
|
| 708 |
+
"rewards/accuracy_reward": 0.2916666818782687,
|
| 709 |
+
"rewards/format_reward": 0.3645833497866988,
|
| 710 |
+
"step": 50
|
| 711 |
+
},
|
| 712 |
+
{
|
| 713 |
+
"completion_length": 2495.2708740234375,
|
| 714 |
+
"entropy": 0.44873046875,
|
| 715 |
+
"epoch": 0.05828571428571429,
|
| 716 |
+
"grad_norm": 0.12384030222892761,
|
| 717 |
+
"kl": 0.005596160888671875,
|
| 718 |
+
"learning_rate": 8.511087728614862e-07,
|
| 719 |
+
"loss": 0.0002,
|
| 720 |
+
"reward": 0.6875000149011612,
|
| 721 |
+
"reward_std": 0.3033446706831455,
|
| 722 |
+
"rewards/accuracy_reward": 0.1458333395421505,
|
| 723 |
+
"rewards/format_reward": 0.5416666716337204,
|
| 724 |
+
"step": 51
|
| 725 |
+
},
|
| 726 |
+
{
|
| 727 |
+
"completion_length": 3027.406280517578,
|
| 728 |
+
"entropy": 0.384765625,
|
| 729 |
+
"epoch": 0.05942857142857143,
|
| 730 |
+
"grad_norm": 0.0901699811220169,
|
| 731 |
+
"kl": 0.0021982192993164062,
|
| 732 |
+
"learning_rate": 8.432457529696548e-07,
|
| 733 |
+
"loss": 0.0001,
|
| 734 |
+
"reward": 0.8750000298023224,
|
| 735 |
+
"reward_std": 0.5351639539003372,
|
| 736 |
+
"rewards/accuracy_reward": 0.3958333507180214,
|
| 737 |
+
"rewards/format_reward": 0.4791666865348816,
|
| 738 |
+
"step": 52
|
| 739 |
+
},
|
| 740 |
+
{
|
| 741 |
+
"completion_length": 2952.8646850585938,
|
| 742 |
+
"entropy": 0.41845703125,
|
| 743 |
+
"epoch": 0.060571428571428575,
|
| 744 |
+
"grad_norm": 0.09236861765384674,
|
| 745 |
+
"kl": 0.0012669563293457031,
|
| 746 |
+
"learning_rate": 8.352239353294194e-07,
|
| 747 |
+
"loss": 0.0001,
|
| 748 |
+
"reward": 0.8541666865348816,
|
| 749 |
+
"reward_std": 0.5214647725224495,
|
| 750 |
+
"rewards/accuracy_reward": 0.260416679084301,
|
| 751 |
+
"rewards/format_reward": 0.5937500074505806,
|
| 752 |
+
"step": 53
|
| 753 |
+
},
|
| 754 |
+
{
|
| 755 |
+
"completion_length": 2996.1250610351562,
|
| 756 |
+
"entropy": 0.3837890625,
|
| 757 |
+
"epoch": 0.061714285714285715,
|
| 758 |
+
"grad_norm": 0.15135987102985382,
|
| 759 |
+
"kl": 0.0015659332275390625,
|
| 760 |
+
"learning_rate": 8.270476638965461e-07,
|
| 761 |
+
"loss": 0.0001,
|
| 762 |
+
"reward": 0.9479166939854622,
|
| 763 |
+
"reward_std": 0.7639089524745941,
|
| 764 |
+
"rewards/accuracy_reward": 0.4062500111758709,
|
| 765 |
+
"rewards/format_reward": 0.5416666828095913,
|
| 766 |
+
"step": 54
|
| 767 |
+
},
|
| 768 |
+
{
|
| 769 |
+
"completion_length": 3076.2500610351562,
|
| 770 |
+
"entropy": 0.4130859375,
|
| 771 |
+
"epoch": 0.06285714285714286,
|
| 772 |
+
"grad_norm": 0.12680813670158386,
|
| 773 |
+
"kl": 0.0023276805877685547,
|
| 774 |
+
"learning_rate": 8.187213662662538e-07,
|
| 775 |
+
"loss": 0.0001,
|
| 776 |
+
"reward": 0.6979166865348816,
|
| 777 |
+
"reward_std": 0.5675121322274208,
|
| 778 |
+
"rewards/accuracy_reward": 0.23958333861082792,
|
| 779 |
+
"rewards/format_reward": 0.458333358168602,
|
| 780 |
+
"step": 55
|
| 781 |
+
},
|
| 782 |
+
{
|
| 783 |
+
"completion_length": 3058.104248046875,
|
| 784 |
+
"entropy": 0.4072265625,
|
| 785 |
+
"epoch": 0.064,
|
| 786 |
+
"grad_norm": 0.10628776252269745,
|
| 787 |
+
"kl": 0.0009794235229492188,
|
| 788 |
+
"learning_rate": 8.102495512755938e-07,
|
| 789 |
+
"loss": 0.0,
|
| 790 |
+
"reward": 0.6562500298023224,
|
| 791 |
+
"reward_std": 0.3362164571881294,
|
| 792 |
+
"rewards/accuracy_reward": 0.19791666697710752,
|
| 793 |
+
"rewards/format_reward": 0.4583333469927311,
|
| 794 |
+
"step": 56
|
| 795 |
+
},
|
| 796 |
+
{
|
| 797 |
+
"completion_length": 3532.2813110351562,
|
| 798 |
+
"entropy": 0.3369140625,
|
| 799 |
+
"epoch": 0.06514285714285714,
|
| 800 |
+
"grad_norm": 0.09391733258962631,
|
| 801 |
+
"kl": 0.0005993843078613281,
|
| 802 |
+
"learning_rate": 8.01636806561836e-07,
|
| 803 |
+
"loss": 0.0,
|
| 804 |
+
"reward": 0.3854166865348816,
|
| 805 |
+
"reward_std": 0.3325711265206337,
|
| 806 |
+
"rewards/accuracy_reward": 0.08333333395421505,
|
| 807 |
+
"rewards/format_reward": 0.3020833432674408,
|
| 808 |
+
"step": 57
|
| 809 |
+
},
|
| 810 |
+
{
|
| 811 |
+
"completion_length": 2239.1145935058594,
|
| 812 |
+
"entropy": 0.322998046875,
|
| 813 |
+
"epoch": 0.06628571428571428,
|
| 814 |
+
"grad_norm": 0.11698172241449356,
|
| 815 |
+
"kl": 0.0037631988525390625,
|
| 816 |
+
"learning_rate": 7.928877960781808e-07,
|
| 817 |
+
"loss": 0.0002,
|
| 818 |
+
"reward": 1.0000000223517418,
|
| 819 |
+
"reward_std": 0.39449498802423477,
|
| 820 |
+
"rewards/accuracy_reward": 0.2604166669771075,
|
| 821 |
+
"rewards/format_reward": 0.7395833358168602,
|
| 822 |
+
"step": 58
|
| 823 |
+
},
|
| 824 |
+
{
|
| 825 |
+
"completion_length": 3092.3438110351562,
|
| 826 |
+
"entropy": 0.3662109375,
|
| 827 |
+
"epoch": 0.06742857142857143,
|
| 828 |
+
"grad_norm": 0.10882271081209183,
|
| 829 |
+
"kl": 0.001026153564453125,
|
| 830 |
+
"learning_rate": 7.840072575681468e-07,
|
| 831 |
+
"loss": 0.0,
|
| 832 |
+
"reward": 0.5625000298023224,
|
| 833 |
+
"reward_std": 0.39667778089642525,
|
| 834 |
+
"rewards/accuracy_reward": 0.19791667442768812,
|
| 835 |
+
"rewards/format_reward": 0.36458333395421505,
|
| 836 |
+
"step": 59
|
| 837 |
+
},
|
| 838 |
+
{
|
| 839 |
+
"completion_length": 3120.5209350585938,
|
| 840 |
+
"entropy": 0.38037109375,
|
| 841 |
+
"epoch": 0.06857142857142857,
|
| 842 |
+
"grad_norm": 0.1319083720445633,
|
| 843 |
+
"kl": 0.0019273757934570312,
|
| 844 |
+
"learning_rate": 7.75e-07,
|
| 845 |
+
"loss": 0.0001,
|
| 846 |
+
"reward": 0.583333358168602,
|
| 847 |
+
"reward_std": 0.49578939378261566,
|
| 848 |
+
"rewards/accuracy_reward": 0.13541666977107525,
|
| 849 |
+
"rewards/format_reward": 0.4479166902601719,
|
| 850 |
+
"step": 60
|
| 851 |
+
},
|
| 852 |
+
{
|
| 853 |
+
"completion_length": 2971.9166870117188,
|
| 854 |
+
"entropy": 0.36669921875,
|
| 855 |
+
"epoch": 0.06971428571428571,
|
| 856 |
+
"grad_norm": 0.17734739184379578,
|
| 857 |
+
"kl": 0.0010924339294433594,
|
| 858 |
+
"learning_rate": 7.658709009626109e-07,
|
| 859 |
+
"loss": 0.0,
|
| 860 |
+
"reward": 0.8020833730697632,
|
| 861 |
+
"reward_std": 0.5149242952466011,
|
| 862 |
+
"rewards/accuracy_reward": 0.2083333358168602,
|
| 863 |
+
"rewards/format_reward": 0.5937500149011612,
|
| 864 |
+
"step": 61
|
| 865 |
+
},
|
| 866 |
+
{
|
| 867 |
+
"completion_length": 2529.7188720703125,
|
| 868 |
+
"entropy": 0.329833984375,
|
| 869 |
+
"epoch": 0.07085714285714285,
|
| 870 |
+
"grad_norm": 0.26185768842697144,
|
| 871 |
+
"kl": 0.016622543334960938,
|
| 872 |
+
"learning_rate": 7.566249040241553e-07,
|
| 873 |
+
"loss": 0.0007,
|
| 874 |
+
"reward": 0.9270833656191826,
|
| 875 |
+
"reward_std": 0.4443442225456238,
|
| 876 |
+
"rewards/accuracy_reward": 0.27083334885537624,
|
| 877 |
+
"rewards/format_reward": 0.6562500149011612,
|
| 878 |
+
"step": 62
|
| 879 |
+
},
|
| 880 |
+
{
|
| 881 |
+
"completion_length": 2196.1250610351562,
|
| 882 |
+
"entropy": 0.36279296875,
|
| 883 |
+
"epoch": 0.072,
|
| 884 |
+
"grad_norm": 0.1202726885676384,
|
| 885 |
+
"kl": 0.0028791427612304688,
|
| 886 |
+
"learning_rate": 7.472670160550848e-07,
|
| 887 |
+
"loss": 0.0001,
|
| 888 |
+
"reward": 1.1562500596046448,
|
| 889 |
+
"reward_std": 0.43789636343717575,
|
| 890 |
+
"rewards/accuracy_reward": 0.385416679084301,
|
| 891 |
+
"rewards/format_reward": 0.7708333432674408,
|
| 892 |
+
"step": 63
|
| 893 |
+
},
|
| 894 |
+
{
|
| 895 |
+
"completion_length": 3074.041717529297,
|
| 896 |
+
"entropy": 0.42041015625,
|
| 897 |
+
"epoch": 0.07314285714285715,
|
| 898 |
+
"grad_norm": 0.10591074079275131,
|
| 899 |
+
"kl": 0.0019989013671875,
|
| 900 |
+
"learning_rate": 7.37802304516818e-07,
|
| 901 |
+
"loss": 0.0001,
|
| 902 |
+
"reward": 0.6250000149011612,
|
| 903 |
+
"reward_std": 0.4529266282916069,
|
| 904 |
+
"rewards/accuracy_reward": 0.18750000651925802,
|
| 905 |
+
"rewards/format_reward": 0.4375000074505806,
|
| 906 |
+
"step": 64
|
| 907 |
+
},
|
| 908 |
+
{
|
| 909 |
+
"completion_length": 2871.416748046875,
|
| 910 |
+
"entropy": 0.365478515625,
|
| 911 |
+
"epoch": 0.07428571428571429,
|
| 912 |
+
"grad_norm": 0.16383042931556702,
|
| 913 |
+
"kl": 0.001850128173828125,
|
| 914 |
+
"learning_rate": 7.282358947176205e-07,
|
| 915 |
+
"loss": 0.0001,
|
| 916 |
+
"reward": 0.6666666865348816,
|
| 917 |
+
"reward_std": 0.38071464747190475,
|
| 918 |
+
"rewards/accuracy_reward": 0.19791667722165585,
|
| 919 |
+
"rewards/format_reward": 0.4687500149011612,
|
| 920 |
+
"step": 65
|
| 921 |
+
},
|
| 922 |
+
{
|
| 923 |
+
"completion_length": 2014.6354370117188,
|
| 924 |
+
"entropy": 0.33642578125,
|
| 925 |
+
"epoch": 0.07542857142857143,
|
| 926 |
+
"grad_norm": 0.26954010128974915,
|
| 927 |
+
"kl": 0.008558273315429688,
|
| 928 |
+
"learning_rate": 7.185729670371604e-07,
|
| 929 |
+
"loss": 0.0003,
|
| 930 |
+
"reward": 0.9375000447034836,
|
| 931 |
+
"reward_std": 0.35192636400461197,
|
| 932 |
+
"rewards/accuracy_reward": 0.34375000838190317,
|
| 933 |
+
"rewards/format_reward": 0.59375,
|
| 934 |
+
"step": 66
|
| 935 |
+
},
|
| 936 |
+
{
|
| 937 |
+
"completion_length": 3587.479248046875,
|
| 938 |
+
"entropy": 0.36572265625,
|
| 939 |
+
"epoch": 0.07657142857142857,
|
| 940 |
+
"grad_norm": 0.08370436728000641,
|
| 941 |
+
"kl": 0.001728057861328125,
|
| 942 |
+
"learning_rate": 7.08818754121241e-07,
|
| 943 |
+
"loss": 0.0001,
|
| 944 |
+
"reward": 0.22916667442768812,
|
| 945 |
+
"reward_std": 0.2259194478392601,
|
| 946 |
+
"rewards/accuracy_reward": 0.010416666977107525,
|
| 947 |
+
"rewards/format_reward": 0.21875000279396772,
|
| 948 |
+
"step": 67
|
| 949 |
+
},
|
| 950 |
+
{
|
| 951 |
+
"completion_length": 2018.3750457763672,
|
| 952 |
+
"entropy": 0.36474609375,
|
| 953 |
+
"epoch": 0.07771428571428571,
|
| 954 |
+
"grad_norm": 0.18221524357795715,
|
| 955 |
+
"kl": 0.0046844482421875,
|
| 956 |
+
"learning_rate": 6.989785380482312e-07,
|
| 957 |
+
"loss": 0.0002,
|
| 958 |
+
"reward": 0.916666716337204,
|
| 959 |
+
"reward_std": 0.35510556399822235,
|
| 960 |
+
"rewards/accuracy_reward": 0.2604166716337204,
|
| 961 |
+
"rewards/format_reward": 0.6562500149011612,
|
| 962 |
+
"step": 68
|
| 963 |
+
},
|
| 964 |
+
{
|
| 965 |
+
"completion_length": 2231.1979370117188,
|
| 966 |
+
"entropy": 0.41796875,
|
| 967 |
+
"epoch": 0.07885714285714286,
|
| 968 |
+
"grad_norm": 0.19485469162464142,
|
| 969 |
+
"kl": 0.0045928955078125,
|
| 970 |
+
"learning_rate": 6.890576474687263e-07,
|
| 971 |
+
"loss": 0.0002,
|
| 972 |
+
"reward": 0.6666666939854622,
|
| 973 |
+
"reward_std": 0.3663952201604843,
|
| 974 |
+
"rewards/accuracy_reward": 0.06250000186264515,
|
| 975 |
+
"rewards/format_reward": 0.6041666865348816,
|
| 976 |
+
"step": 69
|
| 977 |
+
},
|
| 978 |
+
{
|
| 979 |
+
"completion_length": 3180.8854370117188,
|
| 980 |
+
"entropy": 0.38232421875,
|
| 981 |
+
"epoch": 0.08,
|
| 982 |
+
"grad_norm": 0.13257598876953125,
|
| 983 |
+
"kl": 0.00274658203125,
|
| 984 |
+
"learning_rate": 6.790614547199906e-07,
|
| 985 |
+
"loss": 0.0001,
|
| 986 |
+
"reward": 0.45833334885537624,
|
| 987 |
+
"reward_std": 0.4246904104948044,
|
| 988 |
+
"rewards/accuracy_reward": 0.07291666977107525,
|
| 989 |
+
"rewards/format_reward": 0.38541666977107525,
|
| 990 |
+
"step": 70
|
| 991 |
+
},
|
| 992 |
+
{
|
| 993 |
+
"completion_length": 2643.1771240234375,
|
| 994 |
+
"entropy": 0.43798828125,
|
| 995 |
+
"epoch": 0.08114285714285714,
|
| 996 |
+
"grad_norm": 0.13770414888858795,
|
| 997 |
+
"kl": 0.0029726028442382812,
|
| 998 |
+
"learning_rate": 6.68995372916741e-07,
|
| 999 |
+
"loss": 0.0001,
|
| 1000 |
+
"reward": 0.7500000149011612,
|
| 1001 |
+
"reward_std": 0.30269280821084976,
|
| 1002 |
+
"rewards/accuracy_reward": 0.2500000074505806,
|
| 1003 |
+
"rewards/format_reward": 0.5000000149011612,
|
| 1004 |
+
"step": 71
|
| 1005 |
+
},
|
| 1006 |
+
{
|
| 1007 |
+
"completion_length": 2940.6146545410156,
|
| 1008 |
+
"entropy": 0.4892578125,
|
| 1009 |
+
"epoch": 0.08228571428571428,
|
| 1010 |
+
"grad_norm": 0.20104120671749115,
|
| 1011 |
+
"kl": 0.0030574798583984375,
|
| 1012 |
+
"learning_rate": 6.588648530198504e-07,
|
| 1013 |
+
"loss": 0.0001,
|
| 1014 |
+
"reward": 0.5000000149011612,
|
| 1015 |
+
"reward_std": 0.47806398570537567,
|
| 1016 |
+
"rewards/accuracy_reward": 0.06250000186264515,
|
| 1017 |
+
"rewards/format_reward": 0.4375000111758709,
|
| 1018 |
+
"step": 72
|
| 1019 |
+
},
|
| 1020 |
+
{
|
| 1021 |
+
"completion_length": 3779.4583740234375,
|
| 1022 |
+
"entropy": 0.513671875,
|
| 1023 |
+
"epoch": 0.08342857142857144,
|
| 1024 |
+
"grad_norm": 0.09181614220142365,
|
| 1025 |
+
"kl": 0.0015926361083984375,
|
| 1026 |
+
"learning_rate": 6.486753808845564e-07,
|
| 1027 |
+
"loss": 0.0001,
|
| 1028 |
+
"reward": 0.3229166744276881,
|
| 1029 |
+
"reward_std": 0.4466712549328804,
|
| 1030 |
+
"rewards/accuracy_reward": 0.125,
|
| 1031 |
+
"rewards/format_reward": 0.19791667070239782,
|
| 1032 |
+
"step": 73
|
| 1033 |
+
},
|
| 1034 |
+
{
|
| 1035 |
+
"completion_length": 3236.4688720703125,
|
| 1036 |
+
"entropy": 0.4130859375,
|
| 1037 |
+
"epoch": 0.08457142857142858,
|
| 1038 |
+
"grad_norm": 0.1495177149772644,
|
| 1039 |
+
"kl": 0.0028803348541259766,
|
| 1040 |
+
"learning_rate": 6.384324742897735e-07,
|
| 1041 |
+
"loss": 0.0001,
|
| 1042 |
+
"reward": 0.645833358168602,
|
| 1043 |
+
"reward_std": 0.4744175747036934,
|
| 1044 |
+
"rewards/accuracy_reward": 0.26041666977107525,
|
| 1045 |
+
"rewards/format_reward": 0.385416679084301,
|
| 1046 |
+
"step": 74
|
| 1047 |
+
},
|
| 1048 |
+
{
|
| 1049 |
+
"completion_length": 3159.635498046875,
|
| 1050 |
+
"entropy": 0.404296875,
|
| 1051 |
+
"epoch": 0.08571428571428572,
|
| 1052 |
+
"grad_norm": 0.11836569011211395,
|
| 1053 |
+
"kl": 0.0026121139526367188,
|
| 1054 |
+
"learning_rate": 6.281416799501187e-07,
|
| 1055 |
+
"loss": 0.0001,
|
| 1056 |
+
"reward": 0.697916679084301,
|
| 1057 |
+
"reward_std": 0.46560388058423996,
|
| 1058 |
+
"rewards/accuracy_reward": 0.22916666697710752,
|
| 1059 |
+
"rewards/format_reward": 0.4687500111758709,
|
| 1060 |
+
"step": 75
|
| 1061 |
+
},
|
| 1062 |
+
{
|
| 1063 |
+
"completion_length": 2450.041748046875,
|
| 1064 |
+
"entropy": 0.412353515625,
|
| 1065 |
+
"epoch": 0.08685714285714285,
|
| 1066 |
+
"grad_norm": 0.1628679782152176,
|
| 1067 |
+
"kl": 0.0018434524536132812,
|
| 1068 |
+
"learning_rate": 6.178085705122674e-07,
|
| 1069 |
+
"loss": 0.0001,
|
| 1070 |
+
"reward": 0.6875,
|
| 1071 |
+
"reward_std": 0.31900282949209213,
|
| 1072 |
+
"rewards/accuracy_reward": 0.08333333395421505,
|
| 1073 |
+
"rewards/format_reward": 0.6041666716337204,
|
| 1074 |
+
"step": 76
|
| 1075 |
+
},
|
| 1076 |
+
{
|
| 1077 |
+
"completion_length": 3208.3751220703125,
|
| 1078 |
+
"entropy": 0.453125,
|
| 1079 |
+
"epoch": 0.088,
|
| 1080 |
+
"grad_norm": 0.1415146142244339,
|
| 1081 |
+
"kl": 0.002140045166015625,
|
| 1082 |
+
"learning_rate": 6.074387415372676e-07,
|
| 1083 |
+
"loss": 0.0001,
|
| 1084 |
+
"reward": 0.5104166818782687,
|
| 1085 |
+
"reward_std": 0.44949568808078766,
|
| 1086 |
+
"rewards/accuracy_reward": 0.09375000558793545,
|
| 1087 |
+
"rewards/format_reward": 0.4166666744276881,
|
| 1088 |
+
"step": 77
|
| 1089 |
+
},
|
| 1090 |
+
{
|
| 1091 |
+
"completion_length": 2896.6771240234375,
|
| 1092 |
+
"entropy": 0.3759765625,
|
| 1093 |
+
"epoch": 0.08914285714285715,
|
| 1094 |
+
"grad_norm": 0.13142429292201996,
|
| 1095 |
+
"kl": 0.0013580322265625,
|
| 1096 |
+
"learning_rate": 5.97037808470444e-07,
|
| 1097 |
+
"loss": 0.0001,
|
| 1098 |
+
"reward": 0.7500000298023224,
|
| 1099 |
+
"reward_std": 0.5779594928026199,
|
| 1100 |
+
"rewards/accuracy_reward": 0.250000006519258,
|
| 1101 |
+
"rewards/format_reward": 0.5000000223517418,
|
| 1102 |
+
"step": 78
|
| 1103 |
+
},
|
| 1104 |
+
{
|
| 1105 |
+
"completion_length": 2312.0938110351562,
|
| 1106 |
+
"entropy": 0.3544921875,
|
| 1107 |
+
"epoch": 0.09028571428571429,
|
| 1108 |
+
"grad_norm": 0.1474093347787857,
|
| 1109 |
+
"kl": 0.0019969940185546875,
|
| 1110 |
+
"learning_rate": 5.866114036005362e-07,
|
| 1111 |
+
"loss": 0.0001,
|
| 1112 |
+
"reward": 0.8125000149011612,
|
| 1113 |
+
"reward_std": 0.36936958134174347,
|
| 1114 |
+
"rewards/accuracy_reward": 0.20833333674818277,
|
| 1115 |
+
"rewards/format_reward": 0.604166679084301,
|
| 1116 |
+
"step": 79
|
| 1117 |
+
},
|
| 1118 |
+
{
|
| 1119 |
+
"completion_length": 3418.0833740234375,
|
| 1120 |
+
"entropy": 0.49755859375,
|
| 1121 |
+
"epoch": 0.09142857142857143,
|
| 1122 |
+
"grad_norm": 0.14068344235420227,
|
| 1123 |
+
"kl": 0.002742767333984375,
|
| 1124 |
+
"learning_rate": 5.761651730097142e-07,
|
| 1125 |
+
"loss": 0.0001,
|
| 1126 |
+
"reward": 0.5208333544433117,
|
| 1127 |
+
"reward_std": 0.4246201291680336,
|
| 1128 |
+
"rewards/accuracy_reward": 0.16666666697710752,
|
| 1129 |
+
"rewards/format_reward": 0.3541666716337204,
|
| 1130 |
+
"step": 80
|
| 1131 |
+
},
|
| 1132 |
+
{
|
| 1133 |
+
"completion_length": 2992.0729370117188,
|
| 1134 |
+
"entropy": 0.56787109375,
|
| 1135 |
+
"epoch": 0.09257142857142857,
|
| 1136 |
+
"grad_norm": 0.1596606820821762,
|
| 1137 |
+
"kl": 0.005756378173828125,
|
| 1138 |
+
"learning_rate": 5.657047735161255e-07,
|
| 1139 |
+
"loss": 0.0002,
|
| 1140 |
+
"reward": 0.4895833432674408,
|
| 1141 |
+
"reward_std": 0.3185732662677765,
|
| 1142 |
+
"rewards/accuracy_reward": 0.11458333395421505,
|
| 1143 |
+
"rewards/format_reward": 0.3750000074505806,
|
| 1144 |
+
"step": 81
|
| 1145 |
+
},
|
| 1146 |
+
{
|
| 1147 |
+
"completion_length": 2483.9584350585938,
|
| 1148 |
+
"entropy": 0.39306640625,
|
| 1149 |
+
"epoch": 0.09371428571428571,
|
| 1150 |
+
"grad_norm": 0.13652034103870392,
|
| 1151 |
+
"kl": 0.002727508544921875,
|
| 1152 |
+
"learning_rate": 5.552358696106288e-07,
|
| 1153 |
+
"loss": 0.0001,
|
| 1154 |
+
"reward": 0.8020833432674408,
|
| 1155 |
+
"reward_std": 0.24248424544930458,
|
| 1156 |
+
"rewards/accuracy_reward": 0.3020833395421505,
|
| 1157 |
+
"rewards/format_reward": 0.5000000074505806,
|
| 1158 |
+
"step": 82
|
| 1159 |
+
},
|
| 1160 |
+
{
|
| 1161 |
+
"completion_length": 2964.8646850585938,
|
| 1162 |
+
"entropy": 0.48486328125,
|
| 1163 |
+
"epoch": 0.09485714285714286,
|
| 1164 |
+
"grad_norm": 0.10738710314035416,
|
| 1165 |
+
"kl": 0.00264739990234375,
|
| 1166 |
+
"learning_rate": 5.447641303893714e-07,
|
| 1167 |
+
"loss": 0.0001,
|
| 1168 |
+
"reward": 0.5312500074505806,
|
| 1169 |
+
"reward_std": 0.3985592797398567,
|
| 1170 |
+
"rewards/accuracy_reward": 0.1770833395421505,
|
| 1171 |
+
"rewards/format_reward": 0.3541666716337204,
|
| 1172 |
+
"step": 83
|
| 1173 |
+
},
|
| 1174 |
+
{
|
| 1175 |
+
"completion_length": 3069.7396240234375,
|
| 1176 |
+
"entropy": 0.45263671875,
|
| 1177 |
+
"epoch": 0.096,
|
| 1178 |
+
"grad_norm": 0.14286305010318756,
|
| 1179 |
+
"kl": 0.0017604827880859375,
|
| 1180 |
+
"learning_rate": 5.342952264838747e-07,
|
| 1181 |
+
"loss": 0.0001,
|
| 1182 |
+
"reward": 0.739583358168602,
|
| 1183 |
+
"reward_std": 0.44136959314346313,
|
| 1184 |
+
"rewards/accuracy_reward": 0.25,
|
| 1185 |
+
"rewards/format_reward": 0.4895833432674408,
|
| 1186 |
+
"step": 84
|
| 1187 |
+
},
|
| 1188 |
+
{
|
| 1189 |
+
"completion_length": 2664.2188110351562,
|
| 1190 |
+
"entropy": 0.324951171875,
|
| 1191 |
+
"epoch": 0.09714285714285714,
|
| 1192 |
+
"grad_norm": 0.13478516042232513,
|
| 1193 |
+
"kl": 0.002002716064453125,
|
| 1194 |
+
"learning_rate": 5.238348269902859e-07,
|
| 1195 |
+
"loss": 0.0001,
|
| 1196 |
+
"reward": 0.7604166716337204,
|
| 1197 |
+
"reward_std": 0.5178688690066338,
|
| 1198 |
+
"rewards/accuracy_reward": 0.15625000186264515,
|
| 1199 |
+
"rewards/format_reward": 0.6041666679084301,
|
| 1200 |
+
"step": 85
|
| 1201 |
+
},
|
| 1202 |
+
{
|
| 1203 |
+
"completion_length": 2774.291748046875,
|
| 1204 |
+
"entropy": 0.465576171875,
|
| 1205 |
+
"epoch": 0.09828571428571428,
|
| 1206 |
+
"grad_norm": 0.16704270243644714,
|
| 1207 |
+
"kl": 0.00365447998046875,
|
| 1208 |
+
"learning_rate": 5.133885963994639e-07,
|
| 1209 |
+
"loss": 0.0001,
|
| 1210 |
+
"reward": 0.6250000102445483,
|
| 1211 |
+
"reward_std": 0.2606133744120598,
|
| 1212 |
+
"rewards/accuracy_reward": 0.16666666977107525,
|
| 1213 |
+
"rewards/format_reward": 0.4583333386108279,
|
| 1214 |
+
"step": 86
|
| 1215 |
+
},
|
| 1216 |
+
{
|
| 1217 |
+
"completion_length": 2400.302215576172,
|
| 1218 |
+
"entropy": 0.4453125,
|
| 1219 |
+
"epoch": 0.09942857142857142,
|
| 1220 |
+
"grad_norm": 0.23015545308589935,
|
| 1221 |
+
"kl": 0.0040435791015625,
|
| 1222 |
+
"learning_rate": 5.02962191529556e-07,
|
| 1223 |
+
"loss": 0.0002,
|
| 1224 |
+
"reward": 0.8125000447034836,
|
| 1225 |
+
"reward_std": 0.5173326060175896,
|
| 1226 |
+
"rewards/accuracy_reward": 0.18750001024454832,
|
| 1227 |
+
"rewards/format_reward": 0.625,
|
| 1228 |
+
"step": 87
|
| 1229 |
+
},
|
| 1230 |
+
{
|
| 1231 |
+
"completion_length": 2469.7396697998047,
|
| 1232 |
+
"entropy": 0.41943359375,
|
| 1233 |
+
"epoch": 0.10057142857142858,
|
| 1234 |
+
"grad_norm": 0.16470564901828766,
|
| 1235 |
+
"kl": 0.0041961669921875,
|
| 1236 |
+
"learning_rate": 4.925612584627324e-07,
|
| 1237 |
+
"loss": 0.0002,
|
| 1238 |
+
"reward": 1.0208333730697632,
|
| 1239 |
+
"reward_std": 0.693773627281189,
|
| 1240 |
+
"rewards/accuracy_reward": 0.3750000186264515,
|
| 1241 |
+
"rewards/format_reward": 0.6458333507180214,
|
| 1242 |
+
"step": 88
|
| 1243 |
+
},
|
| 1244 |
+
{
|
| 1245 |
+
"completion_length": 2834.635498046875,
|
| 1246 |
+
"entropy": 0.37939453125,
|
| 1247 |
+
"epoch": 0.10171428571428572,
|
| 1248 |
+
"grad_norm": 0.1899234652519226,
|
| 1249 |
+
"kl": 0.003200531005859375,
|
| 1250 |
+
"learning_rate": 4.821914294877326e-07,
|
| 1251 |
+
"loss": 0.0001,
|
| 1252 |
+
"reward": 0.6562500149011612,
|
| 1253 |
+
"reward_std": 0.5321320816874504,
|
| 1254 |
+
"rewards/accuracy_reward": 0.17708333395421505,
|
| 1255 |
+
"rewards/format_reward": 0.479166679084301,
|
| 1256 |
+
"step": 89
|
| 1257 |
+
},
|
| 1258 |
+
{
|
| 1259 |
+
"completion_length": 2226.3959045410156,
|
| 1260 |
+
"entropy": 0.59619140625,
|
| 1261 |
+
"epoch": 0.10285714285714286,
|
| 1262 |
+
"grad_norm": 0.15134188532829285,
|
| 1263 |
+
"kl": 0.0078887939453125,
|
| 1264 |
+
"learning_rate": 4.7185832004988133e-07,
|
| 1265 |
+
"loss": 0.0003,
|
| 1266 |
+
"reward": 0.6458333563059568,
|
| 1267 |
+
"reward_std": 0.21650634706020355,
|
| 1268 |
+
"rewards/accuracy_reward": 0.031250000931322575,
|
| 1269 |
+
"rewards/format_reward": 0.6145833460614085,
|
| 1270 |
+
"step": 90
|
| 1271 |
+
},
|
| 1272 |
+
{
|
| 1273 |
+
"completion_length": 2510.8334045410156,
|
| 1274 |
+
"entropy": 0.423828125,
|
| 1275 |
+
"epoch": 0.104,
|
| 1276 |
+
"grad_norm": 0.15344281494617462,
|
| 1277 |
+
"kl": 0.0041046142578125,
|
| 1278 |
+
"learning_rate": 4.6156752571022637e-07,
|
| 1279 |
+
"loss": 0.0002,
|
| 1280 |
+
"reward": 0.8958333637565374,
|
| 1281 |
+
"reward_std": 0.40820014476776123,
|
| 1282 |
+
"rewards/accuracy_reward": 0.2604166669771075,
|
| 1283 |
+
"rewards/format_reward": 0.6354166818782687,
|
| 1284 |
+
"step": 91
|
| 1285 |
+
},
|
| 1286 |
+
{
|
| 1287 |
+
"completion_length": 2551.062530517578,
|
| 1288 |
+
"entropy": 0.390869140625,
|
| 1289 |
+
"epoch": 0.10514285714285715,
|
| 1290 |
+
"grad_norm": 0.12282641232013702,
|
| 1291 |
+
"kl": 0.00506591796875,
|
| 1292 |
+
"learning_rate": 4.513246191154434e-07,
|
| 1293 |
+
"loss": 0.0002,
|
| 1294 |
+
"reward": 0.6770833432674408,
|
| 1295 |
+
"reward_std": 0.3805246874690056,
|
| 1296 |
+
"rewards/accuracy_reward": 0.09375000093132257,
|
| 1297 |
+
"rewards/format_reward": 0.5833333432674408,
|
| 1298 |
+
"step": 92
|
| 1299 |
+
},
|
| 1300 |
+
{
|
| 1301 |
+
"completion_length": 3784.7188110351562,
|
| 1302 |
+
"entropy": 0.638671875,
|
| 1303 |
+
"epoch": 0.10628571428571429,
|
| 1304 |
+
"grad_norm": 0.2033790946006775,
|
| 1305 |
+
"kl": 0.0058460235595703125,
|
| 1306 |
+
"learning_rate": 4.4113514698014953e-07,
|
| 1307 |
+
"loss": 0.0002,
|
| 1308 |
+
"reward": 0.0729166679084301,
|
| 1309 |
+
"reward_std": 0.18205293267965317,
|
| 1310 |
+
"rewards/accuracy_reward": 0.0,
|
| 1311 |
+
"rewards/format_reward": 0.0729166679084301,
|
| 1312 |
+
"step": 93
|
| 1313 |
+
},
|
| 1314 |
+
{
|
| 1315 |
+
"completion_length": 2915.9375,
|
| 1316 |
+
"entropy": 0.5439453125,
|
| 1317 |
+
"epoch": 0.10742857142857143,
|
| 1318 |
+
"grad_norm": 0.19546350836753845,
|
| 1319 |
+
"kl": 0.004344940185546875,
|
| 1320 |
+
"learning_rate": 4.3100462708325914e-07,
|
| 1321 |
+
"loss": 0.0002,
|
| 1322 |
+
"reward": 0.5833333395421505,
|
| 1323 |
+
"reward_std": 0.427902989089489,
|
| 1324 |
+
"rewards/accuracy_reward": 0.18750000279396772,
|
| 1325 |
+
"rewards/format_reward": 0.3958333358168602,
|
| 1326 |
+
"step": 94
|
| 1327 |
+
},
|
| 1328 |
+
{
|
| 1329 |
+
"completion_length": 3644.4688110351562,
|
| 1330 |
+
"entropy": 0.47607421875,
|
| 1331 |
+
"epoch": 0.10857142857142857,
|
| 1332 |
+
"grad_norm": 0.09118141978979111,
|
| 1333 |
+
"kl": 0.0022993087768554688,
|
| 1334 |
+
"learning_rate": 4.209385452800095e-07,
|
| 1335 |
+
"loss": 0.0001,
|
| 1336 |
+
"reward": 0.3958333358168602,
|
| 1337 |
+
"reward_std": 0.4476298391819,
|
| 1338 |
+
"rewards/accuracy_reward": 0.1041666679084301,
|
| 1339 |
+
"rewards/format_reward": 0.2916666679084301,
|
| 1340 |
+
"step": 95
|
| 1341 |
+
},
|
| 1342 |
+
{
|
| 1343 |
+
"completion_length": 2458.0833740234375,
|
| 1344 |
+
"entropy": 0.37353515625,
|
| 1345 |
+
"epoch": 0.10971428571428571,
|
| 1346 |
+
"grad_norm": 0.16451668739318848,
|
| 1347 |
+
"kl": 0.0038547515869140625,
|
| 1348 |
+
"learning_rate": 4.1094235253127374e-07,
|
| 1349 |
+
"loss": 0.0002,
|
| 1350 |
+
"reward": 0.8750000447034836,
|
| 1351 |
+
"reward_std": 0.4621882885694504,
|
| 1352 |
+
"rewards/accuracy_reward": 0.2708333348855376,
|
| 1353 |
+
"rewards/format_reward": 0.6041666865348816,
|
| 1354 |
+
"step": 96
|
| 1355 |
+
},
|
| 1356 |
+
{
|
| 1357 |
+
"completion_length": 2523.0833740234375,
|
| 1358 |
+
"entropy": 0.421875,
|
| 1359 |
+
"epoch": 0.11085714285714286,
|
| 1360 |
+
"grad_norm": 0.1885625571012497,
|
| 1361 |
+
"kl": 0.003082275390625,
|
| 1362 |
+
"learning_rate": 4.0102146195176887e-07,
|
| 1363 |
+
"loss": 0.0001,
|
| 1364 |
+
"reward": 0.927083358168602,
|
| 1365 |
+
"reward_std": 0.5721743106842041,
|
| 1366 |
+
"rewards/accuracy_reward": 0.2708333348855376,
|
| 1367 |
+
"rewards/format_reward": 0.6562500223517418,
|
| 1368 |
+
"step": 97
|
| 1369 |
+
},
|
| 1370 |
+
{
|
| 1371 |
+
"completion_length": 2336.5625915527344,
|
| 1372 |
+
"entropy": 0.384033203125,
|
| 1373 |
+
"epoch": 0.112,
|
| 1374 |
+
"grad_norm": 0.15700186789035797,
|
| 1375 |
+
"kl": 0.0025310516357421875,
|
| 1376 |
+
"learning_rate": 3.911812458787591e-07,
|
| 1377 |
+
"loss": 0.0001,
|
| 1378 |
+
"reward": 0.7916666865348816,
|
| 1379 |
+
"reward_std": 0.2110944464802742,
|
| 1380 |
+
"rewards/accuracy_reward": 0.13541666697710752,
|
| 1381 |
+
"rewards/format_reward": 0.6562500223517418,
|
| 1382 |
+
"step": 98
|
| 1383 |
+
},
|
| 1384 |
+
{
|
| 1385 |
+
"completion_length": 2524.8958740234375,
|
| 1386 |
+
"entropy": 0.379150390625,
|
| 1387 |
+
"epoch": 0.11314285714285714,
|
| 1388 |
+
"grad_norm": 0.1743498593568802,
|
| 1389 |
+
"kl": 0.003711700439453125,
|
| 1390 |
+
"learning_rate": 3.8142703296283953e-07,
|
| 1391 |
+
"loss": 0.0001,
|
| 1392 |
+
"reward": 0.8125000204890966,
|
| 1393 |
+
"reward_std": 0.4803639128804207,
|
| 1394 |
+
"rewards/accuracy_reward": 0.2708333432674408,
|
| 1395 |
+
"rewards/format_reward": 0.5416666697710752,
|
| 1396 |
+
"step": 99
|
| 1397 |
+
},
|
| 1398 |
+
{
|
| 1399 |
+
"completion_length": 2306.0521545410156,
|
| 1400 |
+
"entropy": 0.35009765625,
|
| 1401 |
+
"epoch": 0.11428571428571428,
|
| 1402 |
+
"grad_norm": 0.12883873283863068,
|
| 1403 |
+
"kl": 0.003726959228515625,
|
| 1404 |
+
"learning_rate": 3.7176410528237945e-07,
|
| 1405 |
+
"loss": 0.0001,
|
| 1406 |
+
"reward": 1.0416666865348816,
|
| 1407 |
+
"reward_std": 0.44550345838069916,
|
| 1408 |
+
"rewards/accuracy_reward": 0.3437500074505806,
|
| 1409 |
+
"rewards/format_reward": 0.6979166716337204,
|
| 1410 |
+
"step": 100
|
| 1411 |
+
},
|
| 1412 |
+
{
|
| 1413 |
+
"completion_length": 2115.5313110351562,
|
| 1414 |
+
"entropy": 0.437255859375,
|
| 1415 |
+
"epoch": 0.11542857142857142,
|
| 1416 |
+
"grad_norm": 0.1613183170557022,
|
| 1417 |
+
"kl": 0.00330352783203125,
|
| 1418 |
+
"learning_rate": 3.62197695483182e-07,
|
| 1419 |
+
"loss": 0.0001,
|
| 1420 |
+
"reward": 0.833333358168602,
|
| 1421 |
+
"reward_std": 0.2652370296418667,
|
| 1422 |
+
"rewards/accuracy_reward": 0.15625000558793545,
|
| 1423 |
+
"rewards/format_reward": 0.6770833358168602,
|
| 1424 |
+
"step": 101
|
| 1425 |
+
},
|
| 1426 |
+
{
|
| 1427 |
+
"completion_length": 1823.4167175292969,
|
| 1428 |
+
"entropy": 0.369384765625,
|
| 1429 |
+
"epoch": 0.11657142857142858,
|
| 1430 |
+
"grad_norm": 0.11039572954177856,
|
| 1431 |
+
"kl": 0.0045318603515625,
|
| 1432 |
+
"learning_rate": 3.5273298394491515e-07,
|
| 1433 |
+
"loss": 0.0002,
|
| 1434 |
+
"reward": 0.9375000298023224,
|
| 1435 |
+
"reward_std": 0.2463684342801571,
|
| 1436 |
+
"rewards/accuracy_reward": 0.12500000279396772,
|
| 1437 |
+
"rewards/format_reward": 0.8125000298023224,
|
| 1438 |
+
"step": 102
|
| 1439 |
+
},
|
| 1440 |
+
{
|
| 1441 |
+
"completion_length": 2301.593780517578,
|
| 1442 |
+
"entropy": 0.376953125,
|
| 1443 |
+
"epoch": 0.11771428571428572,
|
| 1444 |
+
"grad_norm": 0.27249184250831604,
|
| 1445 |
+
"kl": 0.00434112548828125,
|
| 1446 |
+
"learning_rate": 3.433750959758446e-07,
|
| 1447 |
+
"loss": 0.0002,
|
| 1448 |
+
"reward": 0.895833358168602,
|
| 1449 |
+
"reward_std": 0.5925451144576073,
|
| 1450 |
+
"rewards/accuracy_reward": 0.19791667815297842,
|
| 1451 |
+
"rewards/format_reward": 0.6979166865348816,
|
| 1452 |
+
"step": 103
|
| 1453 |
+
},
|
| 1454 |
+
{
|
| 1455 |
+
"completion_length": 2650.4063720703125,
|
| 1456 |
+
"entropy": 0.45458984375,
|
| 1457 |
+
"epoch": 0.11885714285714286,
|
| 1458 |
+
"grad_norm": 0.14005857706069946,
|
| 1459 |
+
"kl": 0.00519561767578125,
|
| 1460 |
+
"learning_rate": 3.3412909903738936e-07,
|
| 1461 |
+
"loss": 0.0002,
|
| 1462 |
+
"reward": 0.572916679084301,
|
| 1463 |
+
"reward_std": 0.42935075983405113,
|
| 1464 |
+
"rewards/accuracy_reward": 0.09375000186264515,
|
| 1465 |
+
"rewards/format_reward": 0.479166679084301,
|
| 1466 |
+
"step": 104
|
| 1467 |
+
},
|
| 1468 |
+
{
|
| 1469 |
+
"completion_length": 2248.6666870117188,
|
| 1470 |
+
"entropy": 0.35205078125,
|
| 1471 |
+
"epoch": 0.12,
|
| 1472 |
+
"grad_norm": 0.18584585189819336,
|
| 1473 |
+
"kl": 0.0032196044921875,
|
| 1474 |
+
"learning_rate": 3.250000000000001e-07,
|
| 1475 |
+
"loss": 0.0001,
|
| 1476 |
+
"reward": 0.885416716337204,
|
| 1477 |
+
"reward_std": 0.5438356846570969,
|
| 1478 |
+
"rewards/accuracy_reward": 0.25000000838190317,
|
| 1479 |
+
"rewards/format_reward": 0.6354166865348816,
|
| 1480 |
+
"step": 105
|
| 1481 |
+
},
|
| 1482 |
+
{
|
| 1483 |
+
"completion_length": 2213.1250915527344,
|
| 1484 |
+
"entropy": 0.30859375,
|
| 1485 |
+
"epoch": 0.12114285714285715,
|
| 1486 |
+
"grad_norm": 0.13285432755947113,
|
| 1487 |
+
"kl": 0.003154754638671875,
|
| 1488 |
+
"learning_rate": 3.159927424318531e-07,
|
| 1489 |
+
"loss": 0.0001,
|
| 1490 |
+
"reward": 1.0625000204890966,
|
| 1491 |
+
"reward_std": 0.4201487675309181,
|
| 1492 |
+
"rewards/accuracy_reward": 0.40625,
|
| 1493 |
+
"rewards/format_reward": 0.6562500055879354,
|
| 1494 |
+
"step": 106
|
| 1495 |
+
},
|
| 1496 |
+
{
|
| 1497 |
+
"completion_length": 2495.4166870117188,
|
| 1498 |
+
"entropy": 0.54296875,
|
| 1499 |
+
"epoch": 0.12228571428571429,
|
| 1500 |
+
"grad_norm": 0.2279452681541443,
|
| 1501 |
+
"kl": 0.005664825439453125,
|
| 1502 |
+
"learning_rate": 3.0711220392181934e-07,
|
| 1503 |
+
"loss": 0.0002,
|
| 1504 |
+
"reward": 0.8437500298023224,
|
| 1505 |
+
"reward_std": 0.39580530673265457,
|
| 1506 |
+
"rewards/accuracy_reward": 0.19791666697710752,
|
| 1507 |
+
"rewards/format_reward": 0.645833358168602,
|
| 1508 |
+
"step": 107
|
| 1509 |
+
},
|
| 1510 |
+
{
|
| 1511 |
+
"completion_length": 2470.9896545410156,
|
| 1512 |
+
"entropy": 0.41259765625,
|
| 1513 |
+
"epoch": 0.12342857142857143,
|
| 1514 |
+
"grad_norm": 0.1852155178785324,
|
| 1515 |
+
"kl": 0.005123138427734375,
|
| 1516 |
+
"learning_rate": 2.9836319343816397e-07,
|
| 1517 |
+
"loss": 0.0002,
|
| 1518 |
+
"reward": 0.7604167014360428,
|
| 1519 |
+
"reward_std": 0.44793232530355453,
|
| 1520 |
+
"rewards/accuracy_reward": 0.1770833358168602,
|
| 1521 |
+
"rewards/format_reward": 0.583333358168602,
|
| 1522 |
+
"step": 108
|
| 1523 |
+
},
|
| 1524 |
+
{
|
| 1525 |
+
"completion_length": 2746.9166870117188,
|
| 1526 |
+
"entropy": 0.41796875,
|
| 1527 |
+
"epoch": 0.12457142857142857,
|
| 1528 |
+
"grad_norm": 0.15881556272506714,
|
| 1529 |
+
"kl": 0.00360107421875,
|
| 1530 |
+
"learning_rate": 2.897504487244061e-07,
|
| 1531 |
+
"loss": 0.0001,
|
| 1532 |
+
"reward": 0.6458333656191826,
|
| 1533 |
+
"reward_std": 0.3334706202149391,
|
| 1534 |
+
"rewards/accuracy_reward": 0.15625,
|
| 1535 |
+
"rewards/format_reward": 0.4895833507180214,
|
| 1536 |
+
"step": 109
|
| 1537 |
+
},
|
| 1538 |
+
{
|
| 1539 |
+
"completion_length": 2482.5000610351562,
|
| 1540 |
+
"entropy": 0.400390625,
|
| 1541 |
+
"epoch": 0.12571428571428572,
|
| 1542 |
+
"grad_norm": 0.1606108397245407,
|
| 1543 |
+
"kl": 0.0030384063720703125,
|
| 1544 |
+
"learning_rate": 2.812786337337463e-07,
|
| 1545 |
+
"loss": 0.0001,
|
| 1546 |
+
"reward": 0.8750000298023224,
|
| 1547 |
+
"reward_std": 0.560508705675602,
|
| 1548 |
+
"rewards/accuracy_reward": 0.22916667349636555,
|
| 1549 |
+
"rewards/format_reward": 0.645833358168602,
|
| 1550 |
+
"step": 110
|
| 1551 |
+
},
|
| 1552 |
+
{
|
| 1553 |
+
"completion_length": 2674.791748046875,
|
| 1554 |
+
"entropy": 0.486572265625,
|
| 1555 |
+
"epoch": 0.12685714285714286,
|
| 1556 |
+
"grad_norm": 0.14915555715560913,
|
| 1557 |
+
"kl": 0.00421905517578125,
|
| 1558 |
+
"learning_rate": 2.729523361034538e-07,
|
| 1559 |
+
"loss": 0.0002,
|
| 1560 |
+
"reward": 0.614583358168602,
|
| 1561 |
+
"reward_std": 0.38519187271595,
|
| 1562 |
+
"rewards/accuracy_reward": 0.12500000186264515,
|
| 1563 |
+
"rewards/format_reward": 0.4895833507180214,
|
| 1564 |
+
"step": 111
|
| 1565 |
+
},
|
| 1566 |
+
{
|
| 1567 |
+
"completion_length": 2968.5834350585938,
|
| 1568 |
+
"entropy": 0.470703125,
|
| 1569 |
+
"epoch": 0.128,
|
| 1570 |
+
"grad_norm": 0.2177121490240097,
|
| 1571 |
+
"kl": 0.0030193328857421875,
|
| 1572 |
+
"learning_rate": 2.6477606467058035e-07,
|
| 1573 |
+
"loss": 0.0001,
|
| 1574 |
+
"reward": 0.8229166865348816,
|
| 1575 |
+
"reward_std": 0.5380749329924583,
|
| 1576 |
+
"rewards/accuracy_reward": 0.2708333386108279,
|
| 1577 |
+
"rewards/format_reward": 0.5520833507180214,
|
| 1578 |
+
"step": 112
|
| 1579 |
+
},
|
| 1580 |
+
{
|
| 1581 |
+
"completion_length": 1850.3125457763672,
|
| 1582 |
+
"entropy": 0.37451171875,
|
| 1583 |
+
"epoch": 0.12914285714285714,
|
| 1584 |
+
"grad_norm": 0.18240460753440857,
|
| 1585 |
+
"kl": 0.004482269287109375,
|
| 1586 |
+
"learning_rate": 2.567542470303452e-07,
|
| 1587 |
+
"loss": 0.0002,
|
| 1588 |
+
"reward": 0.9375000149011612,
|
| 1589 |
+
"reward_std": 0.39232632517814636,
|
| 1590 |
+
"rewards/accuracy_reward": 0.1979166716337204,
|
| 1591 |
+
"rewards/format_reward": 0.7395833432674408,
|
| 1592 |
+
"step": 113
|
| 1593 |
+
},
|
| 1594 |
+
{
|
| 1595 |
+
"completion_length": 1991.9479675292969,
|
| 1596 |
+
"entropy": 0.34228515625,
|
| 1597 |
+
"epoch": 0.13028571428571428,
|
| 1598 |
+
"grad_norm": 0.12856441736221313,
|
| 1599 |
+
"kl": 0.00384521484375,
|
| 1600 |
+
"learning_rate": 2.488912271385139e-07,
|
| 1601 |
+
"loss": 0.0002,
|
| 1602 |
+
"reward": 0.9687500298023224,
|
| 1603 |
+
"reward_std": 0.41063307225704193,
|
| 1604 |
+
"rewards/accuracy_reward": 0.16666666977107525,
|
| 1605 |
+
"rewards/format_reward": 0.802083358168602,
|
| 1606 |
+
"step": 114
|
| 1607 |
+
},
|
| 1608 |
+
{
|
| 1609 |
+
"completion_length": 2642.2813110351562,
|
| 1610 |
+
"entropy": 0.470703125,
|
| 1611 |
+
"epoch": 0.13142857142857142,
|
| 1612 |
+
"grad_norm": 0.1557956039905548,
|
| 1613 |
+
"kl": 0.005847930908203125,
|
| 1614 |
+
"learning_rate": 2.411912629590699e-07,
|
| 1615 |
+
"loss": 0.0002,
|
| 1616 |
+
"reward": 0.7500000149011612,
|
| 1617 |
+
"reward_std": 0.3538191542029381,
|
| 1618 |
+
"rewards/accuracy_reward": 0.2500000074505806,
|
| 1619 |
+
"rewards/format_reward": 0.5000000149011612,
|
| 1620 |
+
"step": 115
|
| 1621 |
+
},
|
| 1622 |
+
{
|
| 1623 |
+
"completion_length": 3452.9271240234375,
|
| 1624 |
+
"entropy": 0.5498046875,
|
| 1625 |
+
"epoch": 0.13257142857142856,
|
| 1626 |
+
"grad_norm": 0.12163397669792175,
|
| 1627 |
+
"kl": 0.00417327880859375,
|
| 1628 |
+
"learning_rate": 2.336585241584522e-07,
|
| 1629 |
+
"loss": 0.0002,
|
| 1630 |
+
"reward": 0.3125000009313226,
|
| 1631 |
+
"reward_std": 0.3494237996637821,
|
| 1632 |
+
"rewards/accuracy_reward": 0.09375000279396772,
|
| 1633 |
+
"rewards/format_reward": 0.21875000558793545,
|
| 1634 |
+
"step": 116
|
| 1635 |
+
},
|
| 1636 |
+
{
|
| 1637 |
+
"completion_length": 2859.000030517578,
|
| 1638 |
+
"entropy": 0.5322265625,
|
| 1639 |
+
"epoch": 0.1337142857142857,
|
| 1640 |
+
"grad_norm": 0.23236258327960968,
|
| 1641 |
+
"kl": 0.0061187744140625,
|
| 1642 |
+
"learning_rate": 2.2629708984760706e-07,
|
| 1643 |
+
"loss": 0.0002,
|
| 1644 |
+
"reward": 0.48958336375653744,
|
| 1645 |
+
"reward_std": 0.34661681205034256,
|
| 1646 |
+
"rewards/accuracy_reward": 0.052083334885537624,
|
| 1647 |
+
"rewards/format_reward": 0.4375000027939677,
|
| 1648 |
+
"step": 117
|
| 1649 |
+
},
|
| 1650 |
+
{
|
| 1651 |
+
"completion_length": 2824.354248046875,
|
| 1652 |
+
"entropy": 0.3935546875,
|
| 1653 |
+
"epoch": 0.13485714285714287,
|
| 1654 |
+
"grad_norm": 0.11868440359830856,
|
| 1655 |
+
"kl": 0.002780914306640625,
|
| 1656 |
+
"learning_rate": 2.1911094637307714e-07,
|
| 1657 |
+
"loss": 0.0001,
|
| 1658 |
+
"reward": 1.0416666865348816,
|
| 1659 |
+
"reward_std": 0.6117755249142647,
|
| 1660 |
+
"rewards/accuracy_reward": 0.4270833358168602,
|
| 1661 |
+
"rewards/format_reward": 0.6145833432674408,
|
| 1662 |
+
"step": 118
|
| 1663 |
+
},
|
| 1664 |
+
{
|
| 1665 |
+
"completion_length": 2026.5937805175781,
|
| 1666 |
+
"entropy": 0.446533203125,
|
| 1667 |
+
"epoch": 0.136,
|
| 1668 |
+
"grad_norm": 0.1918383240699768,
|
| 1669 |
+
"kl": 0.00518035888671875,
|
| 1670 |
+
"learning_rate": 2.1210398515832536e-07,
|
| 1671 |
+
"loss": 0.0002,
|
| 1672 |
+
"reward": 0.989583358168602,
|
| 1673 |
+
"reward_std": 0.3108450919389725,
|
| 1674 |
+
"rewards/accuracy_reward": 0.2604166716337204,
|
| 1675 |
+
"rewards/format_reward": 0.7291666865348816,
|
| 1676 |
+
"step": 119
|
| 1677 |
+
},
|
| 1678 |
+
{
|
| 1679 |
+
"completion_length": 1974.4270935058594,
|
| 1680 |
+
"entropy": 0.43359375,
|
| 1681 |
+
"epoch": 0.13714285714285715,
|
| 1682 |
+
"grad_norm": 0.19848716259002686,
|
| 1683 |
+
"kl": 0.006320953369140625,
|
| 1684 |
+
"learning_rate": 2.0528000059645995e-07,
|
| 1685 |
+
"loss": 0.0003,
|
| 1686 |
+
"reward": 0.864583358168602,
|
| 1687 |
+
"reward_std": 0.4301687255501747,
|
| 1688 |
+
"rewards/accuracy_reward": 0.16666666697710752,
|
| 1689 |
+
"rewards/format_reward": 0.6979166865348816,
|
| 1690 |
+
"step": 120
|
| 1691 |
+
},
|
| 1692 |
+
{
|
| 1693 |
+
"completion_length": 1066.2396240234375,
|
| 1694 |
+
"entropy": 0.30810546875,
|
| 1695 |
+
"epoch": 0.1382857142857143,
|
| 1696 |
+
"grad_norm": 0.15486152470111847,
|
| 1697 |
+
"kl": 0.00482940673828125,
|
| 1698 |
+
"learning_rate": 1.986426879955034e-07,
|
| 1699 |
+
"loss": 0.0002,
|
| 1700 |
+
"reward": 1.2500000298023224,
|
| 1701 |
+
"reward_std": 0.25760992616415024,
|
| 1702 |
+
"rewards/accuracy_reward": 0.3020833386108279,
|
| 1703 |
+
"rewards/format_reward": 0.9479166865348816,
|
| 1704 |
+
"step": 121
|
| 1705 |
+
},
|
| 1706 |
+
{
|
| 1707 |
+
"completion_length": 2534.8958740234375,
|
| 1708 |
+
"entropy": 0.4521484375,
|
| 1709 |
+
"epoch": 0.13942857142857143,
|
| 1710 |
+
"grad_norm": 0.12797077000141144,
|
| 1711 |
+
"kl": 0.00380706787109375,
|
| 1712 |
+
"learning_rate": 1.9219564157731844e-07,
|
| 1713 |
+
"loss": 0.0002,
|
| 1714 |
+
"reward": 0.8229167014360428,
|
| 1715 |
+
"reward_std": 0.3391122668981552,
|
| 1716 |
+
"rewards/accuracy_reward": 0.20833333861082792,
|
| 1717 |
+
"rewards/format_reward": 0.6145833432674408,
|
| 1718 |
+
"step": 122
|
| 1719 |
+
},
|
| 1720 |
+
{
|
| 1721 |
+
"completion_length": 2572.791748046875,
|
| 1722 |
+
"entropy": 0.451416015625,
|
| 1723 |
+
"epoch": 0.14057142857142857,
|
| 1724 |
+
"grad_norm": 0.1227995902299881,
|
| 1725 |
+
"kl": 0.0032405853271484375,
|
| 1726 |
+
"learning_rate": 1.8594235253127372e-07,
|
| 1727 |
+
"loss": 0.0001,
|
| 1728 |
+
"reward": 0.7083333544433117,
|
| 1729 |
+
"reward_std": 0.3739900141954422,
|
| 1730 |
+
"rewards/accuracy_reward": 0.1354166679084301,
|
| 1731 |
+
"rewards/format_reward": 0.5729166828095913,
|
| 1732 |
+
"step": 123
|
| 1733 |
+
},
|
| 1734 |
+
{
|
| 1735 |
+
"completion_length": 2198.6563110351562,
|
| 1736 |
+
"entropy": 0.33984375,
|
| 1737 |
+
"epoch": 0.1417142857142857,
|
| 1738 |
+
"grad_norm": 0.2465568333864212,
|
| 1739 |
+
"kl": 0.015537261962890625,
|
| 1740 |
+
"learning_rate": 1.7988620712370195e-07,
|
| 1741 |
+
"loss": 0.0006,
|
| 1742 |
+
"reward": 0.9687500298023224,
|
| 1743 |
+
"reward_std": 0.5658619552850723,
|
| 1744 |
+
"rewards/accuracy_reward": 0.25,
|
| 1745 |
+
"rewards/format_reward": 0.7187500298023224,
|
| 1746 |
+
"step": 124
|
| 1747 |
+
},
|
| 1748 |
+
{
|
| 1749 |
+
"completion_length": 2865.250030517578,
|
| 1750 |
+
"entropy": 0.431640625,
|
| 1751 |
+
"epoch": 0.14285714285714285,
|
| 1752 |
+
"grad_norm": 0.10473211109638214,
|
| 1753 |
+
"kl": 0.00345611572265625,
|
| 1754 |
+
"learning_rate": 1.7403048486417868e-07,
|
| 1755 |
+
"loss": 0.0001,
|
| 1756 |
+
"reward": 0.6979166697710752,
|
| 1757 |
+
"reward_std": 0.3267679661512375,
|
| 1758 |
+
"rewards/accuracy_reward": 0.30208333395421505,
|
| 1759 |
+
"rewards/format_reward": 0.3958333386108279,
|
| 1760 |
+
"step": 125
|
| 1761 |
+
},
|
| 1762 |
+
{
|
| 1763 |
+
"completion_length": 2886.2188720703125,
|
| 1764 |
+
"entropy": 0.4560546875,
|
| 1765 |
+
"epoch": 0.144,
|
| 1766 |
+
"grad_norm": 0.09086798876523972,
|
| 1767 |
+
"kl": 0.0029506683349609375,
|
| 1768 |
+
"learning_rate": 1.6837835672960831e-07,
|
| 1769 |
+
"loss": 0.0001,
|
| 1770 |
+
"reward": 0.7395833358168602,
|
| 1771 |
+
"reward_std": 0.3677559196949005,
|
| 1772 |
+
"rewards/accuracy_reward": 0.2083333432674408,
|
| 1773 |
+
"rewards/format_reward": 0.5312500149011612,
|
| 1774 |
+
"step": 126
|
| 1775 |
+
},
|
| 1776 |
+
{
|
| 1777 |
+
"completion_length": 2771.3438415527344,
|
| 1778 |
+
"entropy": 0.41796875,
|
| 1779 |
+
"epoch": 0.14514285714285713,
|
| 1780 |
+
"grad_norm": 0.15809084475040436,
|
| 1781 |
+
"kl": 0.004436492919921875,
|
| 1782 |
+
"learning_rate": 1.6293288344708566e-07,
|
| 1783 |
+
"loss": 0.0002,
|
| 1784 |
+
"reward": 0.635416679084301,
|
| 1785 |
+
"reward_std": 0.49130160734057426,
|
| 1786 |
+
"rewards/accuracy_reward": 0.0937500037252903,
|
| 1787 |
+
"rewards/format_reward": 0.541666679084301,
|
| 1788 |
+
"step": 127
|
| 1789 |
+
},
|
| 1790 |
+
{
|
| 1791 |
+
"completion_length": 2852.291748046875,
|
| 1792 |
+
"entropy": 0.5322265625,
|
| 1793 |
+
"epoch": 0.1462857142857143,
|
| 1794 |
+
"grad_norm": 0.17982318997383118,
|
| 1795 |
+
"kl": 0.004756927490234375,
|
| 1796 |
+
"learning_rate": 1.5769701383645698e-07,
|
| 1797 |
+
"loss": 0.0002,
|
| 1798 |
+
"reward": 0.9166666967794299,
|
| 1799 |
+
"reward_std": 0.5036755502223969,
|
| 1800 |
+
"rewards/accuracy_reward": 0.3645833358168602,
|
| 1801 |
+
"rewards/format_reward": 0.5520833535119891,
|
| 1802 |
+
"step": 128
|
| 1803 |
+
},
|
| 1804 |
+
{
|
| 1805 |
+
"completion_length": 3367.9583740234375,
|
| 1806 |
+
"entropy": 0.49658203125,
|
| 1807 |
+
"epoch": 0.14742857142857144,
|
| 1808 |
+
"grad_norm": 0.1593668907880783,
|
| 1809 |
+
"kl": 0.00467681884765625,
|
| 1810 |
+
"learning_rate": 1.5267358321348285e-07,
|
| 1811 |
+
"loss": 0.0002,
|
| 1812 |
+
"reward": 0.4583333507180214,
|
| 1813 |
+
"reward_std": 0.4737073630094528,
|
| 1814 |
+
"rewards/accuracy_reward": 0.1666666753590107,
|
| 1815 |
+
"rewards/format_reward": 0.2916666679084301,
|
| 1816 |
+
"step": 129
|
| 1817 |
+
},
|
| 1818 |
+
{
|
| 1819 |
+
"completion_length": 2817.7709350585938,
|
| 1820 |
+
"entropy": 0.5,
|
| 1821 |
+
"epoch": 0.14857142857142858,
|
| 1822 |
+
"grad_norm": 0.16384749114513397,
|
| 1823 |
+
"kl": 0.00354766845703125,
|
| 1824 |
+
"learning_rate": 1.4786531185446452e-07,
|
| 1825 |
+
"loss": 0.0001,
|
| 1826 |
+
"reward": 0.479166679084301,
|
| 1827 |
+
"reward_std": 0.40873220562934875,
|
| 1828 |
+
"rewards/accuracy_reward": 0.06250000186264515,
|
| 1829 |
+
"rewards/format_reward": 0.4166666716337204,
|
| 1830 |
+
"step": 130
|
| 1831 |
+
},
|
| 1832 |
+
{
|
| 1833 |
+
"completion_length": 2720.8021850585938,
|
| 1834 |
+
"entropy": 0.49658203125,
|
| 1835 |
+
"epoch": 0.14971428571428572,
|
| 1836 |
+
"grad_norm": 0.24187295138835907,
|
| 1837 |
+
"kl": 0.004924774169921875,
|
| 1838 |
+
"learning_rate": 1.432748035231658e-07,
|
| 1839 |
+
"loss": 0.0002,
|
| 1840 |
+
"reward": 0.9062500298023224,
|
| 1841 |
+
"reward_std": 0.504379153251648,
|
| 1842 |
+
"rewards/accuracy_reward": 0.375,
|
| 1843 |
+
"rewards/format_reward": 0.5312500223517418,
|
| 1844 |
+
"step": 131
|
| 1845 |
+
},
|
| 1846 |
+
{
|
| 1847 |
+
"completion_length": 2590.0521545410156,
|
| 1848 |
+
"entropy": 0.4248046875,
|
| 1849 |
+
"epoch": 0.15085714285714286,
|
| 1850 |
+
"grad_norm": 0.13521018624305725,
|
| 1851 |
+
"kl": 0.003265380859375,
|
| 1852 |
+
"learning_rate": 1.3890454406082956e-07,
|
| 1853 |
+
"loss": 0.0001,
|
| 1854 |
+
"reward": 0.833333358168602,
|
| 1855 |
+
"reward_std": 0.5318443104624748,
|
| 1856 |
+
"rewards/accuracy_reward": 0.2916666744276881,
|
| 1857 |
+
"rewards/format_reward": 0.541666679084301,
|
| 1858 |
+
"step": 132
|
| 1859 |
+
},
|
| 1860 |
+
{
|
| 1861 |
+
"completion_length": 3037.8854370117188,
|
| 1862 |
+
"entropy": 0.49072265625,
|
| 1863 |
+
"epoch": 0.152,
|
| 1864 |
+
"grad_norm": 0.17249611020088196,
|
| 1865 |
+
"kl": 0.004932403564453125,
|
| 1866 |
+
"learning_rate": 1.3475690004005097e-07,
|
| 1867 |
+
"loss": 0.0002,
|
| 1868 |
+
"reward": 0.5104166865348816,
|
| 1869 |
+
"reward_std": 0.2723224312067032,
|
| 1870 |
+
"rewards/accuracy_reward": 0.1041666716337204,
|
| 1871 |
+
"rewards/format_reward": 0.4062500149011612,
|
| 1872 |
+
"step": 133
|
| 1873 |
+
},
|
| 1874 |
+
{
|
| 1875 |
+
"completion_length": 2485.322998046875,
|
| 1876 |
+
"entropy": 0.5224609375,
|
| 1877 |
+
"epoch": 0.15314285714285714,
|
| 1878 |
+
"grad_norm": 0.16694706678390503,
|
| 1879 |
+
"kl": 0.00672149658203125,
|
| 1880 |
+
"learning_rate": 1.308341174832359e-07,
|
| 1881 |
+
"loss": 0.0003,
|
| 1882 |
+
"reward": 0.8750000149011612,
|
| 1883 |
+
"reward_std": 0.49518734961748123,
|
| 1884 |
+
"rewards/accuracy_reward": 0.2500000102445483,
|
| 1885 |
+
"rewards/format_reward": 0.6250000149011612,
|
| 1886 |
+
"step": 134
|
| 1887 |
+
},
|
| 1888 |
+
{
|
| 1889 |
+
"completion_length": 1746.3958740234375,
|
| 1890 |
+
"entropy": 0.373046875,
|
| 1891 |
+
"epoch": 0.15428571428571428,
|
| 1892 |
+
"grad_norm": 0.1858878880739212,
|
| 1893 |
+
"kl": 0.007049560546875,
|
| 1894 |
+
"learning_rate": 1.2713832064634125e-07,
|
| 1895 |
+
"loss": 0.0003,
|
| 1896 |
+
"reward": 1.1145833432674408,
|
| 1897 |
+
"reward_std": 0.3352552205324173,
|
| 1898 |
+
"rewards/accuracy_reward": 0.43750000558793545,
|
| 1899 |
+
"rewards/format_reward": 0.6770833432674408,
|
| 1900 |
+
"step": 135
|
| 1901 |
+
},
|
| 1902 |
+
{
|
| 1903 |
+
"completion_length": 2145.4584045410156,
|
| 1904 |
+
"entropy": 0.35986328125,
|
| 1905 |
+
"epoch": 0.15542857142857142,
|
| 1906 |
+
"grad_norm": 0.196150541305542,
|
| 1907 |
+
"kl": 0.00507354736328125,
|
| 1908 |
+
"learning_rate": 1.2367151086855187e-07,
|
| 1909 |
+
"loss": 0.0002,
|
| 1910 |
+
"reward": 0.9895833358168602,
|
| 1911 |
+
"reward_std": 0.6333772391080856,
|
| 1912 |
+
"rewards/accuracy_reward": 0.3125000074505806,
|
| 1913 |
+
"rewards/format_reward": 0.6770833507180214,
|
| 1914 |
+
"step": 136
|
| 1915 |
+
},
|
| 1916 |
+
{
|
| 1917 |
+
"completion_length": 2813.510467529297,
|
| 1918 |
+
"entropy": 0.387939453125,
|
| 1919 |
+
"epoch": 0.15657142857142858,
|
| 1920 |
+
"grad_norm": 0.13014180958271027,
|
| 1921 |
+
"kl": 0.00402069091796875,
|
| 1922 |
+
"learning_rate": 1.2043556548852063e-07,
|
| 1923 |
+
"loss": 0.0002,
|
| 1924 |
+
"reward": 0.677083358168602,
|
| 1925 |
+
"reward_std": 0.5024930611252785,
|
| 1926 |
+
"rewards/accuracy_reward": 0.1458333395421505,
|
| 1927 |
+
"rewards/format_reward": 0.5312500260770321,
|
| 1928 |
+
"step": 137
|
| 1929 |
+
},
|
| 1930 |
+
{
|
| 1931 |
+
"completion_length": 2061.0000610351562,
|
| 1932 |
+
"entropy": 0.35107421875,
|
| 1933 |
+
"epoch": 0.15771428571428572,
|
| 1934 |
+
"grad_norm": 0.10559725016355515,
|
| 1935 |
+
"kl": 0.0037689208984375,
|
| 1936 |
+
"learning_rate": 1.1743223682775649e-07,
|
| 1937 |
+
"loss": 0.0002,
|
| 1938 |
+
"reward": 0.9270833730697632,
|
| 1939 |
+
"reward_std": 0.30403000861406326,
|
| 1940 |
+
"rewards/accuracy_reward": 0.18750000558793545,
|
| 1941 |
+
"rewards/format_reward": 0.739583358168602,
|
| 1942 |
+
"step": 138
|
| 1943 |
+
},
|
| 1944 |
+
{
|
| 1945 |
+
"completion_length": 3106.291748046875,
|
| 1946 |
+
"entropy": 0.55615234375,
|
| 1947 |
+
"epoch": 0.15885714285714286,
|
| 1948 |
+
"grad_norm": 0.15882417559623718,
|
| 1949 |
+
"kl": 0.00519561767578125,
|
| 1950 |
+
"learning_rate": 1.1466315124171128e-07,
|
| 1951 |
+
"loss": 0.0002,
|
| 1952 |
+
"reward": 0.708333358168602,
|
| 1953 |
+
"reward_std": 0.5456142984330654,
|
| 1954 |
+
"rewards/accuracy_reward": 0.1666666753590107,
|
| 1955 |
+
"rewards/format_reward": 0.541666679084301,
|
| 1956 |
+
"step": 139
|
| 1957 |
+
},
|
| 1958 |
+
{
|
| 1959 |
+
"completion_length": 2395.885498046875,
|
| 1960 |
+
"entropy": 0.4853515625,
|
| 1961 |
+
"epoch": 0.16,
|
| 1962 |
+
"grad_norm": 0.2862064242362976,
|
| 1963 |
+
"kl": 0.006809234619140625,
|
| 1964 |
+
"learning_rate": 1.1212980823907929e-07,
|
| 1965 |
+
"loss": 0.0003,
|
| 1966 |
+
"reward": 0.7500000298023224,
|
| 1967 |
+
"reward_std": 0.38956041634082794,
|
| 1968 |
+
"rewards/accuracy_reward": 0.1666666679084301,
|
| 1969 |
+
"rewards/format_reward": 0.5833333507180214,
|
| 1970 |
+
"step": 140
|
| 1971 |
+
},
|
| 1972 |
+
{
|
| 1973 |
+
"completion_length": 1969.2292175292969,
|
| 1974 |
+
"entropy": 0.33935546875,
|
| 1975 |
+
"epoch": 0.16114285714285714,
|
| 1976 |
+
"grad_norm": 0.17285719513893127,
|
| 1977 |
+
"kl": 0.0047760009765625,
|
| 1978 |
+
"learning_rate": 1.0983357966978745e-07,
|
| 1979 |
+
"loss": 0.0002,
|
| 1980 |
+
"reward": 0.9895833730697632,
|
| 1981 |
+
"reward_std": 0.520443569868803,
|
| 1982 |
+
"rewards/accuracy_reward": 0.2187500074505806,
|
| 1983 |
+
"rewards/format_reward": 0.770833358168602,
|
| 1984 |
+
"step": 141
|
| 1985 |
+
},
|
| 1986 |
+
{
|
| 1987 |
+
"completion_length": 2584.9688110351562,
|
| 1988 |
+
"entropy": 0.447998046875,
|
| 1989 |
+
"epoch": 0.16228571428571428,
|
| 1990 |
+
"grad_norm": 0.13187262415885925,
|
| 1991 |
+
"kl": 0.0047740936279296875,
|
| 1992 |
+
"learning_rate": 1.0777570898211405e-07,
|
| 1993 |
+
"loss": 0.0002,
|
| 1994 |
+
"reward": 0.9166667014360428,
|
| 1995 |
+
"reward_std": 0.4435262605547905,
|
| 1996 |
+
"rewards/accuracy_reward": 0.2187500111758709,
|
| 1997 |
+
"rewards/format_reward": 0.6979166865348816,
|
| 1998 |
+
"step": 142
|
| 1999 |
+
},
|
| 2000 |
+
{
|
| 2001 |
+
"completion_length": 2300.7500610351562,
|
| 2002 |
+
"entropy": 0.4326171875,
|
| 2003 |
+
"epoch": 0.16342857142857142,
|
| 2004 |
+
"grad_norm": 0.25766721367836,
|
| 2005 |
+
"kl": 0.00977325439453125,
|
| 2006 |
+
"learning_rate": 1.0595731054933934e-07,
|
| 2007 |
+
"loss": 0.0004,
|
| 2008 |
+
"reward": 0.6875000074505806,
|
| 2009 |
+
"reward_std": 0.3443669453263283,
|
| 2010 |
+
"rewards/accuracy_reward": 0.0833333358168602,
|
| 2011 |
+
"rewards/format_reward": 0.6041666716337204,
|
| 2012 |
+
"step": 143
|
| 2013 |
+
},
|
| 2014 |
+
{
|
| 2015 |
+
"completion_length": 2849.4688110351562,
|
| 2016 |
+
"entropy": 0.45556640625,
|
| 2017 |
+
"epoch": 0.16457142857142856,
|
| 2018 |
+
"grad_norm": 0.14796483516693115,
|
| 2019 |
+
"kl": 0.00493621826171875,
|
| 2020 |
+
"learning_rate": 1.0437936906629334e-07,
|
| 2021 |
+
"loss": 0.0002,
|
| 2022 |
+
"reward": 0.677083358168602,
|
| 2023 |
+
"reward_std": 0.4717573896050453,
|
| 2024 |
+
"rewards/accuracy_reward": 0.2187500037252903,
|
| 2025 |
+
"rewards/format_reward": 0.4583333432674408,
|
| 2026 |
+
"step": 144
|
| 2027 |
+
},
|
| 2028 |
+
{
|
| 2029 |
+
"completion_length": 1885.4479522705078,
|
| 2030 |
+
"entropy": 0.353515625,
|
| 2031 |
+
"epoch": 0.1657142857142857,
|
| 2032 |
+
"grad_norm": 0.1617114096879959,
|
| 2033 |
+
"kl": 0.005008697509765625,
|
| 2034 |
+
"learning_rate": 1.0304273901612565e-07,
|
| 2035 |
+
"loss": 0.0002,
|
| 2036 |
+
"reward": 1.0208333730697632,
|
| 2037 |
+
"reward_std": 0.3080247640609741,
|
| 2038 |
+
"rewards/accuracy_reward": 0.3020833460614085,
|
| 2039 |
+
"rewards/format_reward": 0.7187500149011612,
|
| 2040 |
+
"step": 145
|
| 2041 |
+
},
|
| 2042 |
+
{
|
| 2043 |
+
"completion_length": 1947.0312805175781,
|
| 2044 |
+
"entropy": 0.376953125,
|
| 2045 |
+
"epoch": 0.16685714285714287,
|
| 2046 |
+
"grad_norm": 0.11680302768945694,
|
| 2047 |
+
"kl": 0.0033721923828125,
|
| 2048 |
+
"learning_rate": 1.0194814420758804e-07,
|
| 2049 |
+
"loss": 0.0001,
|
| 2050 |
+
"reward": 0.8750000149011612,
|
| 2051 |
+
"reward_std": 0.22604453563690186,
|
| 2052 |
+
"rewards/accuracy_reward": 0.07291666977107525,
|
| 2053 |
+
"rewards/format_reward": 0.8020833432674408,
|
| 2054 |
+
"step": 146
|
| 2055 |
+
},
|
| 2056 |
+
{
|
| 2057 |
+
"completion_length": 2215.7084045410156,
|
| 2058 |
+
"entropy": 0.392578125,
|
| 2059 |
+
"epoch": 0.168,
|
| 2060 |
+
"grad_norm": 0.21416479349136353,
|
| 2061 |
+
"kl": 0.00583648681640625,
|
| 2062 |
+
"learning_rate": 1.0109617738307911e-07,
|
| 2063 |
+
"loss": 0.0002,
|
| 2064 |
+
"reward": 0.8437500149011612,
|
| 2065 |
+
"reward_std": 0.5214347615838051,
|
| 2066 |
+
"rewards/accuracy_reward": 0.1770833395421505,
|
| 2067 |
+
"rewards/format_reward": 0.6666666865348816,
|
| 2068 |
+
"step": 147
|
| 2069 |
+
},
|
| 2070 |
+
{
|
| 2071 |
+
"completion_length": 1631.229248046875,
|
| 2072 |
+
"entropy": 0.302734375,
|
| 2073 |
+
"epoch": 0.16914285714285715,
|
| 2074 |
+
"grad_norm": 0.17106008529663086,
|
| 2075 |
+
"kl": 0.00476837158203125,
|
| 2076 |
+
"learning_rate": 1.0048729989766394e-07,
|
| 2077 |
+
"loss": 0.0002,
|
| 2078 |
+
"reward": 0.9895833730697632,
|
| 2079 |
+
"reward_std": 0.29220427572727203,
|
| 2080 |
+
"rewards/accuracy_reward": 0.13541666977107525,
|
| 2081 |
+
"rewards/format_reward": 0.8541666865348816,
|
| 2082 |
+
"step": 148
|
| 2083 |
+
},
|
| 2084 |
+
{
|
| 2085 |
+
"completion_length": 2407.8229370117188,
|
| 2086 |
+
"entropy": 0.3408203125,
|
| 2087 |
+
"epoch": 0.1702857142857143,
|
| 2088 |
+
"grad_norm": 0.15983973443508148,
|
| 2089 |
+
"kl": 0.008087158203125,
|
| 2090 |
+
"learning_rate": 1.0012184146924223e-07,
|
| 2091 |
+
"loss": 0.0003,
|
| 2092 |
+
"reward": 0.9270833432674408,
|
| 2093 |
+
"reward_std": 0.48707588016986847,
|
| 2094 |
+
"rewards/accuracy_reward": 0.2500000074505806,
|
| 2095 |
+
"rewards/format_reward": 0.6770833432674408,
|
| 2096 |
+
"step": 149
|
| 2097 |
+
},
|
| 2098 |
+
{
|
| 2099 |
+
"completion_length": 2178.729217529297,
|
| 2100 |
+
"entropy": 0.36962890625,
|
| 2101 |
+
"epoch": 0.17142857142857143,
|
| 2102 |
+
"grad_norm": 0.19081099331378937,
|
| 2103 |
+
"kl": 0.004444122314453125,
|
| 2104 |
+
"learning_rate": 1e-07,
|
| 2105 |
+
"loss": 0.0002,
|
| 2106 |
+
"reward": 1.0312500298023224,
|
| 2107 |
+
"reward_std": 0.5531396120786667,
|
| 2108 |
+
"rewards/accuracy_reward": 0.291666679084301,
|
| 2109 |
+
"rewards/format_reward": 0.7395833432674408,
|
| 2110 |
+
"step": 150
|
| 2111 |
+
},
|
| 2112 |
+
{
|
| 2113 |
+
"epoch": 0.17142857142857143,
|
| 2114 |
+
"step": 150,
|
| 2115 |
+
"total_flos": 0.0,
|
| 2116 |
+
"train_loss": 0.00011944215420650531,
|
| 2117 |
+
"train_runtime": 12092.6435,
|
| 2118 |
+
"train_samples_per_second": 1.191,
|
| 2119 |
+
"train_steps_per_second": 0.012
|
| 2120 |
+
}
|
| 2121 |
+
],
|
| 2122 |
+
"logging_steps": 1,
|
| 2123 |
+
"max_steps": 150,
|
| 2124 |
+
"num_input_tokens_seen": 0,
|
| 2125 |
+
"num_train_epochs": 1,
|
| 2126 |
+
"save_steps": 50,
|
| 2127 |
+
"stateful_callbacks": {
|
| 2128 |
+
"TrainerControl": {
|
| 2129 |
+
"args": {
|
| 2130 |
+
"should_epoch_stop": false,
|
| 2131 |
+
"should_evaluate": false,
|
| 2132 |
+
"should_log": false,
|
| 2133 |
+
"should_save": true,
|
| 2134 |
+
"should_training_stop": true
|
| 2135 |
+
},
|
| 2136 |
+
"attributes": {}
|
| 2137 |
+
}
|
| 2138 |
+
},
|
| 2139 |
+
"total_flos": 0.0,
|
| 2140 |
+
"train_batch_size": 6,
|
| 2141 |
+
"trial_name": null,
|
| 2142 |
+
"trial_params": null
|
| 2143 |
+
}
|