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 +1342 -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: DRA-GRPO
|
| 5 |
+
tags:
|
| 6 |
+
- generated_from_trainer
|
| 7 |
+
- trl
|
| 8 |
+
- grpo
|
| 9 |
+
licence: license
|
| 10 |
+
---
|
| 11 |
+
|
| 12 |
+
# Model Card for DRA-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/DRA-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/lkg866y3)
|
| 31 |
+
|
| 32 |
+
|
| 33 |
+
This model was trained with GRPO, a method introduced in [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://huggingface.co/papers/2402.03300).
|
| 34 |
+
|
| 35 |
+
### Framework versions
|
| 36 |
+
|
| 37 |
+
- TRL: 0.16.0.dev0
|
| 38 |
+
- Transformers: 4.49.0
|
| 39 |
+
- Pytorch: 2.5.1
|
| 40 |
+
- Datasets: 3.6.0
|
| 41 |
+
- Tokenizers: 0.21.1
|
| 42 |
+
|
| 43 |
+
## Citations
|
| 44 |
+
|
| 45 |
+
Cite GRPO as:
|
| 46 |
+
|
| 47 |
+
```bibtex
|
| 48 |
+
@article{zhihong2024deepseekmath,
|
| 49 |
+
title = {{DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models}},
|
| 50 |
+
author = {Zhihong Shao and Peiyi Wang and Qihao Zhu and Runxin Xu and Junxiao Song and Mingchuan Zhang and Y. K. Li and Y. Wu and Daya Guo},
|
| 51 |
+
year = 2024,
|
| 52 |
+
eprint = {arXiv:2402.03300},
|
| 53 |
+
}
|
| 54 |
+
|
| 55 |
+
```
|
| 56 |
+
|
| 57 |
+
Cite TRL as:
|
| 58 |
+
|
| 59 |
+
```bibtex
|
| 60 |
+
@misc{vonwerra2022trl,
|
| 61 |
+
title = {{TRL: Transformer Reinforcement Learning}},
|
| 62 |
+
author = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang and Kashif Rasul and Quentin Gallouédec},
|
| 63 |
+
year = 2020,
|
| 64 |
+
journal = {GitHub repository},
|
| 65 |
+
publisher = {GitHub},
|
| 66 |
+
howpublished = {\url{https://github.com/huggingface/trl}}
|
| 67 |
+
}
|
| 68 |
+
```
|
all_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"total_flos": 0.0,
|
| 3 |
+
"train_loss": 3.568841239768084e-05,
|
| 4 |
+
"train_runtime": 22061.4799,
|
| 5 |
+
"train_samples": 7000,
|
| 6 |
+
"train_samples_per_second": 0.435,
|
| 7 |
+
"train_steps_per_second": 0.005
|
| 8 |
+
}
|
generation_config.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"bos_token_id": 151646,
|
| 4 |
+
"do_sample": true,
|
| 5 |
+
"eos_token_id": 151643,
|
| 6 |
+
"temperature": 0.6,
|
| 7 |
+
"top_p": 0.95,
|
| 8 |
+
"transformers_version": "4.49.0"
|
| 9 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"total_flos": 0.0,
|
| 3 |
+
"train_loss": 3.568841239768084e-05,
|
| 4 |
+
"train_runtime": 22061.4799,
|
| 5 |
+
"train_samples": 7000,
|
| 6 |
+
"train_samples_per_second": 0.435,
|
| 7 |
+
"train_steps_per_second": 0.005
|
| 8 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,1342 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 0.1713796058269066,
|
| 5 |
+
"eval_steps": 500,
|
| 6 |
+
"global_step": 100,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"completion_length": 2842.9896545410156,
|
| 13 |
+
"epoch": 0.001713796058269066,
|
| 14 |
+
"grad_norm": 0.16336098313331604,
|
| 15 |
+
"kl": 0.0,
|
| 16 |
+
"learning_rate": 1e-07,
|
| 17 |
+
"loss": 0.0,
|
| 18 |
+
"reward": 0.0586470989510417,
|
| 19 |
+
"reward_std": 0.1930437758564949,
|
| 20 |
+
"rewards/cosine_scaled_reward": -0.0629267618060112,
|
| 21 |
+
"rewards/format_reward": 0.3541666716337204,
|
| 22 |
+
"step": 1
|
| 23 |
+
},
|
| 24 |
+
{
|
| 25 |
+
"completion_length": 3090.7500610351562,
|
| 26 |
+
"epoch": 0.003427592116538132,
|
| 27 |
+
"grad_norm": 0.15128685534000397,
|
| 28 |
+
"kl": 0.0,
|
| 29 |
+
"learning_rate": 2e-07,
|
| 30 |
+
"loss": -0.0,
|
| 31 |
+
"reward": 0.013502426445484161,
|
| 32 |
+
"reward_std": 0.14039192721247673,
|
| 33 |
+
"rewards/cosine_scaled_reward": -0.10344066098332405,
|
| 34 |
+
"rewards/format_reward": 0.260416679084301,
|
| 35 |
+
"step": 2
|
| 36 |
+
},
|
| 37 |
+
{
|
| 38 |
+
"completion_length": 2600.1875610351562,
|
| 39 |
+
"epoch": 0.005141388174807198,
|
| 40 |
+
"grad_norm": 0.1910804808139801,
|
| 41 |
+
"kl": 3.987550735473633e-05,
|
| 42 |
+
"learning_rate": 3e-07,
|
| 43 |
+
"loss": 0.0,
|
| 44 |
+
"reward": 0.037791766226291656,
|
| 45 |
+
"reward_std": 0.1596581730991602,
|
| 46 |
+
"rewards/cosine_scaled_reward": -0.1509031739551574,
|
| 47 |
+
"rewards/format_reward": 0.4479166865348816,
|
| 48 |
+
"step": 3
|
| 49 |
+
},
|
| 50 |
+
{
|
| 51 |
+
"completion_length": 3076.8438110351562,
|
| 52 |
+
"epoch": 0.006855184233076264,
|
| 53 |
+
"grad_norm": 0.15531130135059357,
|
| 54 |
+
"kl": 4.374980926513672e-05,
|
| 55 |
+
"learning_rate": 4e-07,
|
| 56 |
+
"loss": 0.0,
|
| 57 |
+
"reward": -0.010211404645815492,
|
| 58 |
+
"reward_std": 0.18583546206355095,
|
| 59 |
+
"rewards/cosine_scaled_reward": -0.19252582639455795,
|
| 60 |
+
"rewards/format_reward": 0.3437500037252903,
|
| 61 |
+
"step": 4
|
| 62 |
+
},
|
| 63 |
+
{
|
| 64 |
+
"completion_length": 3045.604248046875,
|
| 65 |
+
"epoch": 0.00856898029134533,
|
| 66 |
+
"grad_norm": 0.11871814727783203,
|
| 67 |
+
"kl": 2.682209014892578e-05,
|
| 68 |
+
"learning_rate": 5e-07,
|
| 69 |
+
"loss": 0.0,
|
| 70 |
+
"reward": 0.0951396506279707,
|
| 71 |
+
"reward_std": 0.17187905684113503,
|
| 72 |
+
"rewards/cosine_scaled_reward": -0.026736530475318432,
|
| 73 |
+
"rewards/format_reward": 0.4270833358168602,
|
| 74 |
+
"step": 5
|
| 75 |
+
},
|
| 76 |
+
{
|
| 77 |
+
"completion_length": 2959.3438720703125,
|
| 78 |
+
"epoch": 0.010282776349614395,
|
| 79 |
+
"grad_norm": 0.13790491223335266,
|
| 80 |
+
"kl": 3.451108932495117e-05,
|
| 81 |
+
"learning_rate": 6e-07,
|
| 82 |
+
"loss": 0.0,
|
| 83 |
+
"reward": 0.11859842715784907,
|
| 84 |
+
"reward_std": 0.17868740856647491,
|
| 85 |
+
"rewards/cosine_scaled_reward": 0.010096830315887928,
|
| 86 |
+
"rewards/format_reward": 0.447916679084301,
|
| 87 |
+
"step": 6
|
| 88 |
+
},
|
| 89 |
+
{
|
| 90 |
+
"completion_length": 2884.6771850585938,
|
| 91 |
+
"epoch": 0.011996572407883462,
|
| 92 |
+
"grad_norm": 0.1611727625131607,
|
| 93 |
+
"kl": 2.6971101760864258e-05,
|
| 94 |
+
"learning_rate": 7e-07,
|
| 95 |
+
"loss": 0.0,
|
| 96 |
+
"reward": 0.03227711468935013,
|
| 97 |
+
"reward_std": 0.19281568750739098,
|
| 98 |
+
"rewards/cosine_scaled_reward": -0.12487578392028809,
|
| 99 |
+
"rewards/format_reward": 0.3750000149011612,
|
| 100 |
+
"step": 7
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"completion_length": 2835.6666870117188,
|
| 104 |
+
"epoch": 0.013710368466152529,
|
| 105 |
+
"grad_norm": 0.1475144475698471,
|
| 106 |
+
"kl": 3.361701965332031e-05,
|
| 107 |
+
"learning_rate": 8e-07,
|
| 108 |
+
"loss": 0.0,
|
| 109 |
+
"reward": 0.045130204409360886,
|
| 110 |
+
"reward_std": 0.18995673581957817,
|
| 111 |
+
"rewards/cosine_scaled_reward": -0.13529739249497652,
|
| 112 |
+
"rewards/format_reward": 0.44791667722165585,
|
| 113 |
+
"step": 8
|
| 114 |
+
},
|
| 115 |
+
{
|
| 116 |
+
"completion_length": 3167.3541870117188,
|
| 117 |
+
"epoch": 0.015424164524421594,
|
| 118 |
+
"grad_norm": 0.13677193224430084,
|
| 119 |
+
"kl": 2.440810203552246e-05,
|
| 120 |
+
"learning_rate": 9e-07,
|
| 121 |
+
"loss": 0.0,
|
| 122 |
+
"reward": 0.020484148059040308,
|
| 123 |
+
"reward_std": 0.16632195003330708,
|
| 124 |
+
"rewards/cosine_scaled_reward": -0.11485888389870524,
|
| 125 |
+
"rewards/format_reward": 0.3125000111758709,
|
| 126 |
+
"step": 9
|
| 127 |
+
},
|
| 128 |
+
{
|
| 129 |
+
"completion_length": 2629.4584350585938,
|
| 130 |
+
"epoch": 0.01713796058269066,
|
| 131 |
+
"grad_norm": 0.14359042048454285,
|
| 132 |
+
"kl": 2.0831823348999023e-05,
|
| 133 |
+
"learning_rate": 1e-06,
|
| 134 |
+
"loss": 0.0,
|
| 135 |
+
"reward": 0.1312947366386652,
|
| 136 |
+
"reward_std": 0.1554134227335453,
|
| 137 |
+
"rewards/cosine_scaled_reward": 0.009967727586627007,
|
| 138 |
+
"rewards/format_reward": 0.5000000074505806,
|
| 139 |
+
"step": 10
|
| 140 |
+
},
|
| 141 |
+
{
|
| 142 |
+
"completion_length": 2906.1875610351562,
|
| 143 |
+
"epoch": 0.018851756640959727,
|
| 144 |
+
"grad_norm": 0.20071381330490112,
|
| 145 |
+
"kl": 3.3527612686157227e-05,
|
| 146 |
+
"learning_rate": 9.997258721585931e-07,
|
| 147 |
+
"loss": 0.0,
|
| 148 |
+
"reward": 0.033789366483688354,
|
| 149 |
+
"reward_std": 0.19262398406863213,
|
| 150 |
+
"rewards/cosine_scaled_reward": -0.08468496927525848,
|
| 151 |
+
"rewards/format_reward": 0.3020833460614085,
|
| 152 |
+
"step": 11
|
| 153 |
+
},
|
| 154 |
+
{
|
| 155 |
+
"completion_length": 2942.4168090820312,
|
| 156 |
+
"epoch": 0.02056555269922879,
|
| 157 |
+
"grad_norm": 0.15123063325881958,
|
| 158 |
+
"kl": 2.7477741241455078e-05,
|
| 159 |
+
"learning_rate": 9.989038226169207e-07,
|
| 160 |
+
"loss": 0.0,
|
| 161 |
+
"reward": 0.052781688049435616,
|
| 162 |
+
"reward_std": 0.19968328438699245,
|
| 163 |
+
"rewards/cosine_scaled_reward": -0.08891183231025934,
|
| 164 |
+
"rewards/format_reward": 0.3854166865348816,
|
| 165 |
+
"step": 12
|
| 166 |
+
},
|
| 167 |
+
{
|
| 168 |
+
"completion_length": 2924.7501220703125,
|
| 169 |
+
"epoch": 0.022279348757497857,
|
| 170 |
+
"grad_norm": 0.1488049179315567,
|
| 171 |
+
"kl": 1.4431774616241455e-05,
|
| 172 |
+
"learning_rate": 9.975348529157229e-07,
|
| 173 |
+
"loss": 0.0,
|
| 174 |
+
"reward": 0.1416405034251511,
|
| 175 |
+
"reward_std": 0.20871220529079437,
|
| 176 |
+
"rewards/cosine_scaled_reward": 0.05599783733487129,
|
| 177 |
+
"rewards/format_reward": 0.4479166865348816,
|
| 178 |
+
"step": 13
|
| 179 |
+
},
|
| 180 |
+
{
|
| 181 |
+
"completion_length": 2709.6250915527344,
|
| 182 |
+
"epoch": 0.023993144815766924,
|
| 183 |
+
"grad_norm": 0.18846973776817322,
|
| 184 |
+
"kl": 3.9637088775634766e-05,
|
| 185 |
+
"learning_rate": 9.956206309337066e-07,
|
| 186 |
+
"loss": 0.0,
|
| 187 |
+
"reward": 0.06054589990526438,
|
| 188 |
+
"reward_std": 0.1687684953212738,
|
| 189 |
+
"rewards/cosine_scaled_reward": -0.10551007091999054,
|
| 190 |
+
"rewards/format_reward": 0.4479166865348816,
|
| 191 |
+
"step": 14
|
| 192 |
+
},
|
| 193 |
+
{
|
| 194 |
+
"completion_length": 1950.8646545410156,
|
| 195 |
+
"epoch": 0.02570694087403599,
|
| 196 |
+
"grad_norm": 0.18432216346263885,
|
| 197 |
+
"kl": 4.6759843826293945e-05,
|
| 198 |
+
"learning_rate": 9.931634888554935e-07,
|
| 199 |
+
"loss": 0.0,
|
| 200 |
+
"reward": 0.23200466111302376,
|
| 201 |
+
"reward_std": 0.2355622537434101,
|
| 202 |
+
"rewards/cosine_scaled_reward": 0.06849630922079086,
|
| 203 |
+
"rewards/format_reward": 0.7708333432674408,
|
| 204 |
+
"step": 15
|
| 205 |
+
},
|
| 206 |
+
{
|
| 207 |
+
"completion_length": 3007.6875,
|
| 208 |
+
"epoch": 0.027420736932305057,
|
| 209 |
+
"grad_norm": 0.15038132667541504,
|
| 210 |
+
"kl": 3.84598970413208e-05,
|
| 211 |
+
"learning_rate": 9.901664203302124e-07,
|
| 212 |
+
"loss": 0.0,
|
| 213 |
+
"reward": 0.06949906051158905,
|
| 214 |
+
"reward_std": 0.21871956065297127,
|
| 215 |
+
"rewards/cosine_scaled_reward": -0.07155923545360565,
|
| 216 |
+
"rewards/format_reward": 0.416666679084301,
|
| 217 |
+
"step": 16
|
| 218 |
+
},
|
| 219 |
+
{
|
| 220 |
+
"completion_length": 2888.8438110351562,
|
| 221 |
+
"epoch": 0.02913453299057412,
|
| 222 |
+
"grad_norm": 0.17018933594226837,
|
| 223 |
+
"kl": 5.860626697540283e-05,
|
| 224 |
+
"learning_rate": 9.866330768241983e-07,
|
| 225 |
+
"loss": 0.0,
|
| 226 |
+
"reward": 0.08268930693157017,
|
| 227 |
+
"reward_std": 0.2000725194811821,
|
| 228 |
+
"rewards/cosine_scaled_reward": -0.03420639084652066,
|
| 229 |
+
"rewards/format_reward": 0.3958333469927311,
|
| 230 |
+
"step": 17
|
| 231 |
+
},
|
| 232 |
+
{
|
| 233 |
+
"completion_length": 2994.1666870117188,
|
| 234 |
+
"epoch": 0.030848329048843187,
|
| 235 |
+
"grad_norm": 0.14627192914485931,
|
| 236 |
+
"kl": 3.9910897612571716e-05,
|
| 237 |
+
"learning_rate": 9.825677631722435e-07,
|
| 238 |
+
"loss": 0.0,
|
| 239 |
+
"reward": 0.1066256552003324,
|
| 240 |
+
"reward_std": 0.1765430960804224,
|
| 241 |
+
"rewards/cosine_scaled_reward": -0.012849166989326477,
|
| 242 |
+
"rewards/format_reward": 0.447916679084301,
|
| 243 |
+
"step": 18
|
| 244 |
+
},
|
| 245 |
+
{
|
| 246 |
+
"completion_length": 2986.885498046875,
|
| 247 |
+
"epoch": 0.032562125107112254,
|
| 248 |
+
"grad_norm": 0.11996503919363022,
|
| 249 |
+
"kl": 7.943809032440186e-05,
|
| 250 |
+
"learning_rate": 9.779754323328192e-07,
|
| 251 |
+
"loss": 0.0,
|
| 252 |
+
"reward": 0.037288530729711056,
|
| 253 |
+
"reward_std": 0.1901964209973812,
|
| 254 |
+
"rewards/cosine_scaled_reward": -0.08846686349716038,
|
| 255 |
+
"rewards/format_reward": 0.3229166753590107,
|
| 256 |
+
"step": 19
|
| 257 |
+
},
|
| 258 |
+
{
|
| 259 |
+
"completion_length": 3068.0000610351562,
|
| 260 |
+
"epoch": 0.03427592116538132,
|
| 261 |
+
"grad_norm": 0.15175943076610565,
|
| 262 |
+
"kl": 0.00019240379333496094,
|
| 263 |
+
"learning_rate": 9.728616793536587e-07,
|
| 264 |
+
"loss": 0.0,
|
| 265 |
+
"reward": 0.03987372852861881,
|
| 266 |
+
"reward_std": 0.18592991679906845,
|
| 267 |
+
"rewards/cosine_scaled_reward": -0.11358156427741051,
|
| 268 |
+
"rewards/format_reward": 0.3854166828095913,
|
| 269 |
+
"step": 20
|
| 270 |
+
},
|
| 271 |
+
{
|
| 272 |
+
"completion_length": 3086.6771240234375,
|
| 273 |
+
"epoch": 0.03598971722365039,
|
| 274 |
+
"grad_norm": 0.1517907977104187,
|
| 275 |
+
"kl": 0.00036978721618652344,
|
| 276 |
+
"learning_rate": 9.672327345550543e-07,
|
| 277 |
+
"loss": 0.0,
|
| 278 |
+
"reward": 0.03412920422852039,
|
| 279 |
+
"reward_std": 0.1651427522301674,
|
| 280 |
+
"rewards/cosine_scaled_reward": -0.09963592886924744,
|
| 281 |
+
"rewards/format_reward": 0.33333334419876337,
|
| 282 |
+
"step": 21
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"completion_length": 3277.1458740234375,
|
| 286 |
+
"epoch": 0.037703513281919454,
|
| 287 |
+
"grad_norm": 0.11310048401355743,
|
| 288 |
+
"kl": 0.00019124150276184082,
|
| 289 |
+
"learning_rate": 9.610954559391704e-07,
|
| 290 |
+
"loss": 0.0,
|
| 291 |
+
"reward": -0.0021694526076316833,
|
| 292 |
+
"reward_std": 0.1807849071919918,
|
| 293 |
+
"rewards/cosine_scaled_reward": -0.1448658797889948,
|
| 294 |
+
"rewards/format_reward": 0.2812500074505806,
|
| 295 |
+
"step": 22
|
| 296 |
+
},
|
| 297 |
+
{
|
| 298 |
+
"completion_length": 2485.1875915527344,
|
| 299 |
+
"epoch": 0.03941730934018852,
|
| 300 |
+
"grad_norm": 0.21775458753108978,
|
| 301 |
+
"kl": 0.0002810359001159668,
|
| 302 |
+
"learning_rate": 9.54457320834625e-07,
|
| 303 |
+
"loss": 0.0,
|
| 304 |
+
"reward": 0.1259636478498578,
|
| 305 |
+
"reward_std": 0.21954844892024994,
|
| 306 |
+
"rewards/cosine_scaled_reward": 0.0026862788945436478,
|
| 307 |
+
"rewards/format_reward": 0.4895833395421505,
|
| 308 |
+
"step": 23
|
| 309 |
+
},
|
| 310 |
+
{
|
| 311 |
+
"completion_length": 3387.8646240234375,
|
| 312 |
+
"epoch": 0.04113110539845758,
|
| 313 |
+
"grad_norm": 0.11246265470981598,
|
| 314 |
+
"kl": 0.00029206275939941406,
|
| 315 |
+
"learning_rate": 9.473264167865171e-07,
|
| 316 |
+
"loss": 0.0,
|
| 317 |
+
"reward": -0.0444450975919608,
|
| 318 |
+
"reward_std": 0.19019928015768528,
|
| 319 |
+
"rewards/cosine_scaled_reward": -0.17542381957173347,
|
| 320 |
+
"rewards/format_reward": 0.1770833358168602,
|
| 321 |
+
"step": 24
|
| 322 |
+
},
|
| 323 |
+
{
|
| 324 |
+
"completion_length": 3383.3229370117188,
|
| 325 |
+
"epoch": 0.04284490145672665,
|
| 326 |
+
"grad_norm": 0.11982468515634537,
|
| 327 |
+
"kl": 0.00018139183521270752,
|
| 328 |
+
"learning_rate": 9.397114317029974e-07,
|
| 329 |
+
"loss": 0.0,
|
| 330 |
+
"reward": -0.08237307704985142,
|
| 331 |
+
"reward_std": 0.10484589636325836,
|
| 332 |
+
"rewards/cosine_scaled_reward": -0.2248186320066452,
|
| 333 |
+
"rewards/format_reward": 0.1250000074505806,
|
| 334 |
+
"step": 25
|
| 335 |
+
},
|
| 336 |
+
{
|
| 337 |
+
"completion_length": 2839.0625610351562,
|
| 338 |
+
"epoch": 0.044558697514995714,
|
| 339 |
+
"grad_norm": 0.1347901076078415,
|
| 340 |
+
"kl": 0.0003955364227294922,
|
| 341 |
+
"learning_rate": 9.316216432703916e-07,
|
| 342 |
+
"loss": 0.0,
|
| 343 |
+
"reward": 0.11097948951646686,
|
| 344 |
+
"reward_std": 0.18024785444140434,
|
| 345 |
+
"rewards/cosine_scaled_reward": 0.0011768988333642483,
|
| 346 |
+
"rewards/format_reward": 0.4375000111758709,
|
| 347 |
+
"step": 26
|
| 348 |
+
},
|
| 349 |
+
{
|
| 350 |
+
"completion_length": 2636.5729370117188,
|
| 351 |
+
"epoch": 0.04627249357326478,
|
| 352 |
+
"grad_norm": 0.13496308028697968,
|
| 353 |
+
"kl": 0.0006954669952392578,
|
| 354 |
+
"learning_rate": 9.230669076497687e-07,
|
| 355 |
+
"loss": 0.0,
|
| 356 |
+
"reward": 0.05549928080290556,
|
| 357 |
+
"reward_std": 0.14237392507493496,
|
| 358 |
+
"rewards/cosine_scaled_reward": -0.12638001888990402,
|
| 359 |
+
"rewards/format_reward": 0.4687500111758709,
|
| 360 |
+
"step": 27
|
| 361 |
+
},
|
| 362 |
+
{
|
| 363 |
+
"completion_length": 2851.2396240234375,
|
| 364 |
+
"epoch": 0.04798628963153385,
|
| 365 |
+
"grad_norm": 0.16802968084812164,
|
| 366 |
+
"kl": 0.00039768218994140625,
|
| 367 |
+
"learning_rate": 9.140576474687263e-07,
|
| 368 |
+
"loss": 0.0,
|
| 369 |
+
"reward": 0.005757941631600261,
|
| 370 |
+
"reward_std": 0.16327618807554245,
|
| 371 |
+
"rewards/cosine_scaled_reward": -0.18160032108426094,
|
| 372 |
+
"rewards/format_reward": 0.385416679084301,
|
| 373 |
+
"step": 28
|
| 374 |
+
},
|
| 375 |
+
{
|
| 376 |
+
"completion_length": 2759.5208740234375,
|
| 377 |
+
"epoch": 0.049700085689802914,
|
| 378 |
+
"grad_norm": 0.1442675143480301,
|
| 379 |
+
"kl": 0.0005460977554321289,
|
| 380 |
+
"learning_rate": 9.046048391230247e-07,
|
| 381 |
+
"loss": 0.0,
|
| 382 |
+
"reward": 0.09482090920209885,
|
| 383 |
+
"reward_std": 0.16877998411655426,
|
| 384 |
+
"rewards/cosine_scaled_reward": -0.010849054902791977,
|
| 385 |
+
"rewards/format_reward": 0.3958333432674408,
|
| 386 |
+
"step": 29
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"completion_length": 3377.59375,
|
| 390 |
+
"epoch": 0.05141388174807198,
|
| 391 |
+
"grad_norm": 0.12018371373414993,
|
| 392 |
+
"kl": 0.0006362795829772949,
|
| 393 |
+
"learning_rate": 8.9471999940354e-07,
|
| 394 |
+
"loss": 0.0,
|
| 395 |
+
"reward": 0.028713669627904892,
|
| 396 |
+
"reward_std": 0.18587654829025269,
|
| 397 |
+
"rewards/cosine_scaled_reward": -0.07821773737668991,
|
| 398 |
+
"rewards/format_reward": 0.2708333358168602,
|
| 399 |
+
"step": 30
|
| 400 |
+
},
|
| 401 |
+
{
|
| 402 |
+
"completion_length": 3138.104248046875,
|
| 403 |
+
"epoch": 0.05312767780634105,
|
| 404 |
+
"grad_norm": 0.1430775374174118,
|
| 405 |
+
"kl": 0.0004627704620361328,
|
| 406 |
+
"learning_rate": 8.844151714648274e-07,
|
| 407 |
+
"loss": 0.0,
|
| 408 |
+
"reward": -0.008235585875809193,
|
| 409 |
+
"reward_std": 0.14394641481339931,
|
| 410 |
+
"rewards/cosine_scaled_reward": -0.13156688585877419,
|
| 411 |
+
"rewards/format_reward": 0.22916666977107525,
|
| 412 |
+
"step": 31
|
| 413 |
+
},
|
| 414 |
+
{
|
| 415 |
+
"completion_length": 2630.5938720703125,
|
| 416 |
+
"epoch": 0.054841473864610114,
|
| 417 |
+
"grad_norm": 0.17797145247459412,
|
| 418 |
+
"kl": 0.0012159347534179688,
|
| 419 |
+
"learning_rate": 8.737029101523929e-07,
|
| 420 |
+
"loss": 0.0,
|
| 421 |
+
"reward": 0.13316596718505025,
|
| 422 |
+
"reward_std": 0.22369278594851494,
|
| 423 |
+
"rewards/cosine_scaled_reward": -0.002324562519788742,
|
| 424 |
+
"rewards/format_reward": 0.5312500149011612,
|
| 425 |
+
"step": 32
|
| 426 |
+
},
|
| 427 |
+
{
|
| 428 |
+
"completion_length": 2592.5000610351562,
|
| 429 |
+
"epoch": 0.056555269922879174,
|
| 430 |
+
"grad_norm": 0.15806646645069122,
|
| 431 |
+
"kl": 0.0012989044189453125,
|
| 432 |
+
"learning_rate": 8.625962667065487e-07,
|
| 433 |
+
"loss": 0.0001,
|
| 434 |
+
"reward": 0.11066696606576443,
|
| 435 |
+
"reward_std": 0.14327863790094852,
|
| 436 |
+
"rewards/cosine_scaled_reward": -0.010827179066836834,
|
| 437 |
+
"rewards/format_reward": 0.4583333432674408,
|
| 438 |
+
"step": 33
|
| 439 |
+
},
|
| 440 |
+
{
|
| 441 |
+
"completion_length": 2626.5313110351562,
|
| 442 |
+
"epoch": 0.05826906598114824,
|
| 443 |
+
"grad_norm": 0.1691393107175827,
|
| 444 |
+
"kl": 0.0020818710327148438,
|
| 445 |
+
"learning_rate": 8.511087728614862e-07,
|
| 446 |
+
"loss": 0.0001,
|
| 447 |
+
"reward": 0.09107339568436146,
|
| 448 |
+
"reward_std": 0.18131284788250923,
|
| 449 |
+
"rewards/cosine_scaled_reward": -0.04873488505836576,
|
| 450 |
+
"rewards/format_reward": 0.4583333507180214,
|
| 451 |
+
"step": 34
|
| 452 |
+
},
|
| 453 |
+
{
|
| 454 |
+
"completion_length": 3051.9896850585938,
|
| 455 |
+
"epoch": 0.05998286203941731,
|
| 456 |
+
"grad_norm": 0.17819511890411377,
|
| 457 |
+
"kl": 0.0019023418426513672,
|
| 458 |
+
"learning_rate": 8.392544243589427e-07,
|
| 459 |
+
"loss": 0.0001,
|
| 460 |
+
"reward": 0.10884217114653438,
|
| 461 |
+
"reward_std": 0.23975158482789993,
|
| 462 |
+
"rewards/cosine_scaled_reward": 0.007055521011352539,
|
| 463 |
+
"rewards/format_reward": 0.416666679084301,
|
| 464 |
+
"step": 35
|
| 465 |
+
},
|
| 466 |
+
{
|
| 467 |
+
"completion_length": 2809.8646240234375,
|
| 468 |
+
"epoch": 0.061696658097686374,
|
| 469 |
+
"grad_norm": 0.14143270254135132,
|
| 470 |
+
"kl": 0.0004401206970214844,
|
| 471 |
+
"learning_rate": 8.270476638965461e-07,
|
| 472 |
+
"loss": 0.0,
|
| 473 |
+
"reward": 0.15133970603346825,
|
| 474 |
+
"reward_std": 0.25536274164915085,
|
| 475 |
+
"rewards/cosine_scaled_reward": 0.04777143523097038,
|
| 476 |
+
"rewards/format_reward": 0.5000000149011612,
|
| 477 |
+
"step": 36
|
| 478 |
+
},
|
| 479 |
+
{
|
| 480 |
+
"completion_length": 3153.0938110351562,
|
| 481 |
+
"epoch": 0.06341045415595545,
|
| 482 |
+
"grad_norm": 0.12571778893470764,
|
| 483 |
+
"kl": 0.0004382133483886719,
|
| 484 |
+
"learning_rate": 8.145033635316128e-07,
|
| 485 |
+
"loss": 0.0,
|
| 486 |
+
"reward": 0.020705484319478273,
|
| 487 |
+
"reward_std": 0.17584117874503136,
|
| 488 |
+
"rewards/cosine_scaled_reward": -0.13113276287913322,
|
| 489 |
+
"rewards/format_reward": 0.3437500074505806,
|
| 490 |
+
"step": 37
|
| 491 |
+
},
|
| 492 |
+
{
|
| 493 |
+
"completion_length": 3118.322998046875,
|
| 494 |
+
"epoch": 0.06512425021422451,
|
| 495 |
+
"grad_norm": 0.11900849640369415,
|
| 496 |
+
"kl": 0.00022101402282714844,
|
| 497 |
+
"learning_rate": 8.01636806561836e-07,
|
| 498 |
+
"loss": 0.0,
|
| 499 |
+
"reward": 0.07037997618317604,
|
| 500 |
+
"reward_std": 0.2210748754441738,
|
| 501 |
+
"rewards/cosine_scaled_reward": -0.028155604377388954,
|
| 502 |
+
"rewards/format_reward": 0.33333333395421505,
|
| 503 |
+
"step": 38
|
| 504 |
+
},
|
| 505 |
+
{
|
| 506 |
+
"completion_length": 2472.3646240234375,
|
| 507 |
+
"epoch": 0.06683804627249357,
|
| 508 |
+
"grad_norm": 0.1329745203256607,
|
| 509 |
+
"kl": 0.0012660026550292969,
|
| 510 |
+
"learning_rate": 7.884636689049422e-07,
|
| 511 |
+
"loss": 0.0001,
|
| 512 |
+
"reward": 0.11453905934467912,
|
| 513 |
+
"reward_std": 0.20037821307778358,
|
| 514 |
+
"rewards/cosine_scaled_reward": -0.056616831570863724,
|
| 515 |
+
"rewards/format_reward": 0.5625000037252903,
|
| 516 |
+
"step": 39
|
| 517 |
+
},
|
| 518 |
+
{
|
| 519 |
+
"completion_length": 3018.072998046875,
|
| 520 |
+
"epoch": 0.06855184233076264,
|
| 521 |
+
"grad_norm": 0.11555609107017517,
|
| 522 |
+
"kl": 0.00041091442108154297,
|
| 523 |
+
"learning_rate": 7.75e-07,
|
| 524 |
+
"loss": 0.0,
|
| 525 |
+
"reward": 0.01033751294016838,
|
| 526 |
+
"reward_std": 0.1491969907656312,
|
| 527 |
+
"rewards/cosine_scaled_reward": -0.13083185872528702,
|
| 528 |
+
"rewards/format_reward": 0.3020833386108279,
|
| 529 |
+
"step": 40
|
| 530 |
+
},
|
| 531 |
+
{
|
| 532 |
+
"completion_length": 2939.229248046875,
|
| 533 |
+
"epoch": 0.0702656383890317,
|
| 534 |
+
"grad_norm": 0.12907516956329346,
|
| 535 |
+
"kl": 0.0007364749908447266,
|
| 536 |
+
"learning_rate": 7.612622032536507e-07,
|
| 537 |
+
"loss": 0.0,
|
| 538 |
+
"reward": 0.12070174579275772,
|
| 539 |
+
"reward_std": 0.2728600651025772,
|
| 540 |
+
"rewards/cosine_scaled_reward": -0.001188491762150079,
|
| 541 |
+
"rewards/format_reward": 0.479166679084301,
|
| 542 |
+
"step": 41
|
| 543 |
+
},
|
| 544 |
+
{
|
| 545 |
+
"completion_length": 2293.854248046875,
|
| 546 |
+
"epoch": 0.07197943444730077,
|
| 547 |
+
"grad_norm": 0.15430894494056702,
|
| 548 |
+
"kl": 0.001895904541015625,
|
| 549 |
+
"learning_rate": 7.472670160550848e-07,
|
| 550 |
+
"loss": 0.0001,
|
| 551 |
+
"reward": 0.1563777169212699,
|
| 552 |
+
"reward_std": 0.2286355346441269,
|
| 553 |
+
"rewards/cosine_scaled_reward": -0.015645149163901806,
|
| 554 |
+
"rewards/format_reward": 0.6458333432674408,
|
| 555 |
+
"step": 42
|
| 556 |
+
},
|
| 557 |
+
{
|
| 558 |
+
"completion_length": 2741.0833740234375,
|
| 559 |
+
"epoch": 0.07369323050556983,
|
| 560 |
+
"grad_norm": 0.11995150148868561,
|
| 561 |
+
"kl": 0.0004520416259765625,
|
| 562 |
+
"learning_rate": 7.330314893841101e-07,
|
| 563 |
+
"loss": 0.0,
|
| 564 |
+
"reward": 0.13986177369952202,
|
| 565 |
+
"reward_std": 0.22300692275166512,
|
| 566 |
+
"rewards/cosine_scaled_reward": 0.05043928697705269,
|
| 567 |
+
"rewards/format_reward": 0.4479166753590107,
|
| 568 |
+
"step": 43
|
| 569 |
+
},
|
| 570 |
+
{
|
| 571 |
+
"completion_length": 2305.854217529297,
|
| 572 |
+
"epoch": 0.07540702656383891,
|
| 573 |
+
"grad_norm": 0.21461769938468933,
|
| 574 |
+
"kl": 0.0007772445678710938,
|
| 575 |
+
"learning_rate": 7.185729670371604e-07,
|
| 576 |
+
"loss": 0.0,
|
| 577 |
+
"reward": 0.07035261578857899,
|
| 578 |
+
"reward_std": 0.13939796388149261,
|
| 579 |
+
"rewards/cosine_scaled_reward": -0.09040920436382294,
|
| 580 |
+
"rewards/format_reward": 0.4583333386108279,
|
| 581 |
+
"step": 44
|
| 582 |
+
},
|
| 583 |
+
{
|
| 584 |
+
"completion_length": 2846.8958435058594,
|
| 585 |
+
"epoch": 0.07712082262210797,
|
| 586 |
+
"grad_norm": 0.13480690121650696,
|
| 587 |
+
"kl": 0.0009918212890625,
|
| 588 |
+
"learning_rate": 7.039090644965509e-07,
|
| 589 |
+
"loss": 0.0,
|
| 590 |
+
"reward": -0.018673544749617577,
|
| 591 |
+
"reward_std": 0.16758902743458748,
|
| 592 |
+
"rewards/cosine_scaled_reward": -0.20985213853418827,
|
| 593 |
+
"rewards/format_reward": 0.3437500149011612,
|
| 594 |
+
"step": 45
|
| 595 |
+
},
|
| 596 |
+
{
|
| 597 |
+
"completion_length": 2584.1563110351562,
|
| 598 |
+
"epoch": 0.07883461868037704,
|
| 599 |
+
"grad_norm": 0.17533442378044128,
|
| 600 |
+
"kl": 0.0022344589233398438,
|
| 601 |
+
"learning_rate": 6.890576474687263e-07,
|
| 602 |
+
"loss": 0.0001,
|
| 603 |
+
"reward": 0.022231539245694876,
|
| 604 |
+
"reward_std": 0.16662119142711163,
|
| 605 |
+
"rewards/cosine_scaled_reward": -0.17729704454541206,
|
| 606 |
+
"rewards/format_reward": 0.4375000223517418,
|
| 607 |
+
"step": 46
|
| 608 |
+
},
|
| 609 |
+
{
|
| 610 |
+
"completion_length": 3102.354248046875,
|
| 611 |
+
"epoch": 0.0805484147386461,
|
| 612 |
+
"grad_norm": 0.1247975304722786,
|
| 613 |
+
"kl": 0.0013914108276367188,
|
| 614 |
+
"learning_rate": 6.740368101176495e-07,
|
| 615 |
+
"loss": 0.0001,
|
| 616 |
+
"reward": -0.00813557580113411,
|
| 617 |
+
"reward_std": 0.1496633905917406,
|
| 618 |
+
"rewards/cosine_scaled_reward": -0.17180366069078445,
|
| 619 |
+
"rewards/format_reward": 0.31250000838190317,
|
| 620 |
+
"step": 47
|
| 621 |
+
},
|
| 622 |
+
{
|
| 623 |
+
"completion_length": 2603.8438415527344,
|
| 624 |
+
"epoch": 0.08226221079691516,
|
| 625 |
+
"grad_norm": 0.16067557036876678,
|
| 626 |
+
"kl": 0.0016384124755859375,
|
| 627 |
+
"learning_rate": 6.588648530198504e-07,
|
| 628 |
+
"loss": 0.0001,
|
| 629 |
+
"reward": 0.11067175259813666,
|
| 630 |
+
"reward_std": 0.179900873452425,
|
| 631 |
+
"rewards/cosine_scaled_reward": -0.013014900498092175,
|
| 632 |
+
"rewards/format_reward": 0.4583333507180214,
|
| 633 |
+
"step": 48
|
| 634 |
+
},
|
| 635 |
+
{
|
| 636 |
+
"completion_length": 3373.0313110351562,
|
| 637 |
+
"epoch": 0.08397600685518423,
|
| 638 |
+
"grad_norm": 0.10881118476390839,
|
| 639 |
+
"kl": 0.0002281665802001953,
|
| 640 |
+
"learning_rate": 6.435602608679916e-07,
|
| 641 |
+
"loss": 0.0,
|
| 642 |
+
"reward": 0.00016227364540100098,
|
| 643 |
+
"reward_std": 0.14608008973300457,
|
| 644 |
+
"rewards/cosine_scaled_reward": -0.10383528657257557,
|
| 645 |
+
"rewards/format_reward": 0.2083333358168602,
|
| 646 |
+
"step": 49
|
| 647 |
+
},
|
| 648 |
+
{
|
| 649 |
+
"completion_length": 3033.0208740234375,
|
| 650 |
+
"epoch": 0.0856898029134533,
|
| 651 |
+
"grad_norm": 0.13846911489963531,
|
| 652 |
+
"kl": 0.0012478828430175781,
|
| 653 |
+
"learning_rate": 6.281416799501187e-07,
|
| 654 |
+
"loss": 0.0001,
|
| 655 |
+
"reward": 0.06065037357620895,
|
| 656 |
+
"reward_std": 0.17429610714316368,
|
| 657 |
+
"rewards/cosine_scaled_reward": -0.05210728943347931,
|
| 658 |
+
"rewards/format_reward": 0.3437500074505806,
|
| 659 |
+
"step": 50
|
| 660 |
+
},
|
| 661 |
+
{
|
| 662 |
+
"completion_length": 2933.4793090820312,
|
| 663 |
+
"epoch": 0.08740359897172237,
|
| 664 |
+
"grad_norm": 0.13350166380405426,
|
| 665 |
+
"kl": 0.00033092498779296875,
|
| 666 |
+
"learning_rate": 6.126278954320294e-07,
|
| 667 |
+
"loss": 0.0,
|
| 668 |
+
"reward": -0.01886086305603385,
|
| 669 |
+
"reward_std": 0.13372720032930374,
|
| 670 |
+
"rewards/cosine_scaled_reward": -0.23485597968101501,
|
| 671 |
+
"rewards/format_reward": 0.3958333507180214,
|
| 672 |
+
"step": 51
|
| 673 |
+
},
|
| 674 |
+
{
|
| 675 |
+
"completion_length": 3236.791748046875,
|
| 676 |
+
"epoch": 0.08911739502999143,
|
| 677 |
+
"grad_norm": 0.10703922808170319,
|
| 678 |
+
"kl": 0.0003795623779296875,
|
| 679 |
+
"learning_rate": 5.97037808470444e-07,
|
| 680 |
+
"loss": 0.0,
|
| 681 |
+
"reward": 0.0947269550524652,
|
| 682 |
+
"reward_std": 0.2209848240017891,
|
| 683 |
+
"rewards/cosine_scaled_reward": 0.004282929934561253,
|
| 684 |
+
"rewards/format_reward": 0.3645833432674408,
|
| 685 |
+
"step": 52
|
| 686 |
+
},
|
| 687 |
+
{
|
| 688 |
+
"completion_length": 2535.1875610351562,
|
| 689 |
+
"epoch": 0.0908311910882605,
|
| 690 |
+
"grad_norm": 0.19642022252082825,
|
| 691 |
+
"kl": 0.0011463165283203125,
|
| 692 |
+
"learning_rate": 5.813904131848564e-07,
|
| 693 |
+
"loss": 0.0,
|
| 694 |
+
"reward": 0.074245510622859,
|
| 695 |
+
"reward_std": 0.20659937523305416,
|
| 696 |
+
"rewards/cosine_scaled_reward": -0.10727175627835095,
|
| 697 |
+
"rewards/format_reward": 0.5000000074505806,
|
| 698 |
+
"step": 53
|
| 699 |
+
},
|
| 700 |
+
{
|
| 701 |
+
"completion_length": 3126.1146240234375,
|
| 702 |
+
"epoch": 0.09254498714652956,
|
| 703 |
+
"grad_norm": 0.17827318608760834,
|
| 704 |
+
"kl": 0.001310110092163086,
|
| 705 |
+
"learning_rate": 5.657047735161255e-07,
|
| 706 |
+
"loss": 0.0001,
|
| 707 |
+
"reward": 0.0032570459879934788,
|
| 708 |
+
"reward_std": 0.15636450424790382,
|
| 709 |
+
"rewards/cosine_scaled_reward": -0.14380789827555418,
|
| 710 |
+
"rewards/format_reward": 0.30208333395421505,
|
| 711 |
+
"step": 54
|
| 712 |
+
},
|
| 713 |
+
{
|
| 714 |
+
"completion_length": 2929.9688720703125,
|
| 715 |
+
"epoch": 0.09425878320479864,
|
| 716 |
+
"grad_norm": 0.1380075216293335,
|
| 717 |
+
"kl": 0.001453399658203125,
|
| 718 |
+
"learning_rate": 5.5e-07,
|
| 719 |
+
"loss": 0.0001,
|
| 720 |
+
"reward": 0.045829253271222115,
|
| 721 |
+
"reward_std": 0.1648220308125019,
|
| 722 |
+
"rewards/cosine_scaled_reward": -0.06621671980246902,
|
| 723 |
+
"rewards/format_reward": 0.31250000558793545,
|
| 724 |
+
"step": 55
|
| 725 |
+
},
|
| 726 |
+
{
|
| 727 |
+
"completion_length": 2777.7083435058594,
|
| 728 |
+
"epoch": 0.0959725792630677,
|
| 729 |
+
"grad_norm": 0.15538543462753296,
|
| 730 |
+
"kl": 0.0006000995635986328,
|
| 731 |
+
"learning_rate": 5.342952264838747e-07,
|
| 732 |
+
"loss": 0.0,
|
| 733 |
+
"reward": 0.05388350097928196,
|
| 734 |
+
"reward_std": 0.1790449097752571,
|
| 735 |
+
"rewards/cosine_scaled_reward": -0.09807816706597805,
|
| 736 |
+
"rewards/format_reward": 0.4062500186264515,
|
| 737 |
+
"step": 56
|
| 738 |
+
},
|
| 739 |
+
{
|
| 740 |
+
"completion_length": 3187.697998046875,
|
| 741 |
+
"epoch": 0.09768637532133675,
|
| 742 |
+
"grad_norm": 0.10485029965639114,
|
| 743 |
+
"kl": 0.0006384849548339844,
|
| 744 |
+
"learning_rate": 5.186095868151436e-07,
|
| 745 |
+
"loss": 0.0,
|
| 746 |
+
"reward": 0.02806462941225618,
|
| 747 |
+
"reward_std": 0.1900525726377964,
|
| 748 |
+
"rewards/cosine_scaled_reward": -0.10978525504469872,
|
| 749 |
+
"rewards/format_reward": 0.3333333432674408,
|
| 750 |
+
"step": 57
|
| 751 |
+
},
|
| 752 |
+
{
|
| 753 |
+
"completion_length": 2514.208465576172,
|
| 754 |
+
"epoch": 0.09940017137960583,
|
| 755 |
+
"grad_norm": 0.15051576495170593,
|
| 756 |
+
"kl": 0.0010862350463867188,
|
| 757 |
+
"learning_rate": 5.02962191529556e-07,
|
| 758 |
+
"loss": 0.0,
|
| 759 |
+
"reward": 0.04878209065645933,
|
| 760 |
+
"reward_std": 0.17859612964093685,
|
| 761 |
+
"rewards/cosine_scaled_reward": -0.17480848729610443,
|
| 762 |
+
"rewards/format_reward": 0.5416666865348816,
|
| 763 |
+
"step": 58
|
| 764 |
+
},
|
| 765 |
+
{
|
| 766 |
+
"completion_length": 2698.4896240234375,
|
| 767 |
+
"epoch": 0.10111396743787489,
|
| 768 |
+
"grad_norm": 0.17716623842716217,
|
| 769 |
+
"kl": 0.00115966796875,
|
| 770 |
+
"learning_rate": 4.873721045679706e-07,
|
| 771 |
+
"loss": 0.0,
|
| 772 |
+
"reward": 0.18329142779111862,
|
| 773 |
+
"reward_std": 0.27025362849235535,
|
| 774 |
+
"rewards/cosine_scaled_reward": 0.0847717362921685,
|
| 775 |
+
"rewards/format_reward": 0.5520833507180214,
|
| 776 |
+
"step": 59
|
| 777 |
+
},
|
| 778 |
+
{
|
| 779 |
+
"completion_length": 2798.9375610351562,
|
| 780 |
+
"epoch": 0.10282776349614396,
|
| 781 |
+
"grad_norm": 0.17029182612895966,
|
| 782 |
+
"kl": 0.00177001953125,
|
| 783 |
+
"learning_rate": 4.7185832004988133e-07,
|
| 784 |
+
"loss": 0.0001,
|
| 785 |
+
"reward": 0.02042369358241558,
|
| 786 |
+
"reward_std": 0.13106655701994896,
|
| 787 |
+
"rewards/cosine_scaled_reward": -0.17489198595285416,
|
| 788 |
+
"rewards/format_reward": 0.4270833395421505,
|
| 789 |
+
"step": 60
|
| 790 |
+
},
|
| 791 |
+
{
|
| 792 |
+
"completion_length": 2818.84375,
|
| 793 |
+
"epoch": 0.10454155955441302,
|
| 794 |
+
"grad_norm": 0.14482490718364716,
|
| 795 |
+
"kl": 0.0011817216873168945,
|
| 796 |
+
"learning_rate": 4.5643973913200837e-07,
|
| 797 |
+
"loss": 0.0,
|
| 798 |
+
"reward": 0.06702410569414496,
|
| 799 |
+
"reward_std": 0.16219814494252205,
|
| 800 |
+
"rewards/cosine_scaled_reward": -0.07603749074041843,
|
| 801 |
+
"rewards/format_reward": 0.4166666865348816,
|
| 802 |
+
"step": 61
|
| 803 |
+
},
|
| 804 |
+
{
|
| 805 |
+
"completion_length": 3321.2604370117188,
|
| 806 |
+
"epoch": 0.1062553556126821,
|
| 807 |
+
"grad_norm": 0.12993496656417847,
|
| 808 |
+
"kl": 0.0014019012451171875,
|
| 809 |
+
"learning_rate": 4.4113514698014953e-07,
|
| 810 |
+
"loss": 0.0001,
|
| 811 |
+
"reward": -0.06851966120302677,
|
| 812 |
+
"reward_std": 0.1139833852648735,
|
| 813 |
+
"rewards/cosine_scaled_reward": -0.20644466951489449,
|
| 814 |
+
"rewards/format_reward": 0.14583333861082792,
|
| 815 |
+
"step": 62
|
| 816 |
+
},
|
| 817 |
+
{
|
| 818 |
+
"completion_length": 3189.135498046875,
|
| 819 |
+
"epoch": 0.10796915167095116,
|
| 820 |
+
"grad_norm": 0.1303822249174118,
|
| 821 |
+
"kl": 0.0011129379272460938,
|
| 822 |
+
"learning_rate": 4.2596318988235037e-07,
|
| 823 |
+
"loss": 0.0,
|
| 824 |
+
"reward": -0.018826739862561226,
|
| 825 |
+
"reward_std": 0.13178249262273312,
|
| 826 |
+
"rewards/cosine_scaled_reward": -0.13564759586006403,
|
| 827 |
+
"rewards/format_reward": 0.19791666977107525,
|
| 828 |
+
"step": 63
|
| 829 |
+
},
|
| 830 |
+
{
|
| 831 |
+
"completion_length": 2808.697998046875,
|
| 832 |
+
"epoch": 0.10968294772922023,
|
| 833 |
+
"grad_norm": 0.1368769407272339,
|
| 834 |
+
"kl": 0.0009212493896484375,
|
| 835 |
+
"learning_rate": 4.1094235253127374e-07,
|
| 836 |
+
"loss": 0.0,
|
| 837 |
+
"reward": 0.05554246995598078,
|
| 838 |
+
"reward_std": 0.17365428805351257,
|
| 839 |
+
"rewards/cosine_scaled_reward": -0.0875049065798521,
|
| 840 |
+
"rewards/format_reward": 0.3958333460614085,
|
| 841 |
+
"step": 64
|
| 842 |
+
},
|
| 843 |
+
{
|
| 844 |
+
"completion_length": 3047.6771240234375,
|
| 845 |
+
"epoch": 0.11139674378748929,
|
| 846 |
+
"grad_norm": 0.11966266483068466,
|
| 847 |
+
"kl": 0.0005755424499511719,
|
| 848 |
+
"learning_rate": 3.9609093550344907e-07,
|
| 849 |
+
"loss": 0.0,
|
| 850 |
+
"reward": 0.08035540394484997,
|
| 851 |
+
"reward_std": 0.1835373416543007,
|
| 852 |
+
"rewards/cosine_scaled_reward": -0.03351259673945606,
|
| 853 |
+
"rewards/format_reward": 0.385416679084301,
|
| 854 |
+
"step": 65
|
| 855 |
+
},
|
| 856 |
+
{
|
| 857 |
+
"completion_length": 2938.4166870117188,
|
| 858 |
+
"epoch": 0.11311053984575835,
|
| 859 |
+
"grad_norm": 0.15843810141086578,
|
| 860 |
+
"kl": 0.0007996559143066406,
|
| 861 |
+
"learning_rate": 3.8142703296283953e-07,
|
| 862 |
+
"loss": 0.0,
|
| 863 |
+
"reward": 0.02124944794923067,
|
| 864 |
+
"reward_std": 0.14316299185156822,
|
| 865 |
+
"rewards/cosine_scaled_reward": -0.09920864924788475,
|
| 866 |
+
"rewards/format_reward": 0.28125,
|
| 867 |
+
"step": 66
|
| 868 |
+
},
|
| 869 |
+
{
|
| 870 |
+
"completion_length": 2877.0834350585938,
|
| 871 |
+
"epoch": 0.11482433590402742,
|
| 872 |
+
"grad_norm": 0.16664999723434448,
|
| 873 |
+
"kl": 0.0018649101257324219,
|
| 874 |
+
"learning_rate": 3.6696851061588994e-07,
|
| 875 |
+
"loss": 0.0001,
|
| 876 |
+
"reward": 0.023960473015904427,
|
| 877 |
+
"reward_std": 0.17024591751396656,
|
| 878 |
+
"rewards/cosine_scaled_reward": -0.12505127862095833,
|
| 879 |
+
"rewards/format_reward": 0.3437500102445483,
|
| 880 |
+
"step": 67
|
| 881 |
+
},
|
| 882 |
+
{
|
| 883 |
+
"completion_length": 2517.4479370117188,
|
| 884 |
+
"epoch": 0.11653813196229648,
|
| 885 |
+
"grad_norm": 0.18517287075519562,
|
| 886 |
+
"kl": 0.0015734434127807617,
|
| 887 |
+
"learning_rate": 3.5273298394491515e-07,
|
| 888 |
+
"loss": 0.0001,
|
| 889 |
+
"reward": 0.12474549748003483,
|
| 890 |
+
"reward_std": 0.17121394164860249,
|
| 891 |
+
"rewards/cosine_scaled_reward": -0.03634676709771156,
|
| 892 |
+
"rewards/format_reward": 0.5625000074505806,
|
| 893 |
+
"step": 68
|
| 894 |
+
},
|
| 895 |
+
{
|
| 896 |
+
"completion_length": 2816.3646850585938,
|
| 897 |
+
"epoch": 0.11825192802056556,
|
| 898 |
+
"grad_norm": 0.16400621831417084,
|
| 899 |
+
"kl": 0.0013251304626464844,
|
| 900 |
+
"learning_rate": 3.387377967463493e-07,
|
| 901 |
+
"loss": 0.0001,
|
| 902 |
+
"reward": 0.009894740185700357,
|
| 903 |
+
"reward_std": 0.17024565115571022,
|
| 904 |
+
"rewards/cosine_scaled_reward": -0.16932371503207833,
|
| 905 |
+
"rewards/format_reward": 0.375,
|
| 906 |
+
"step": 69
|
| 907 |
+
},
|
| 908 |
+
{
|
| 909 |
+
"completion_length": 2789.4168090820312,
|
| 910 |
+
"epoch": 0.11996572407883462,
|
| 911 |
+
"grad_norm": 0.1394311487674713,
|
| 912 |
+
"kl": 0.0010738372802734375,
|
| 913 |
+
"learning_rate": 3.250000000000001e-07,
|
| 914 |
+
"loss": 0.0,
|
| 915 |
+
"reward": 0.1316700098104775,
|
| 916 |
+
"reward_std": 0.2457997277379036,
|
| 917 |
+
"rewards/cosine_scaled_reward": 0.04584646038711071,
|
| 918 |
+
"rewards/format_reward": 0.4270833395421505,
|
| 919 |
+
"step": 70
|
| 920 |
+
},
|
| 921 |
+
{
|
| 922 |
+
"completion_length": 2267.166717529297,
|
| 923 |
+
"epoch": 0.12167952013710369,
|
| 924 |
+
"grad_norm": 0.15189269185066223,
|
| 925 |
+
"kl": 0.001094818115234375,
|
| 926 |
+
"learning_rate": 3.115363310950578e-07,
|
| 927 |
+
"loss": 0.0,
|
| 928 |
+
"reward": 0.24627735652029514,
|
| 929 |
+
"reward_std": 0.1724311113357544,
|
| 930 |
+
"rewards/cosine_scaled_reward": 0.16811909526586533,
|
| 931 |
+
"rewards/format_reward": 0.6354166865348816,
|
| 932 |
+
"step": 71
|
| 933 |
+
},
|
| 934 |
+
{
|
| 935 |
+
"completion_length": 3065.7501220703125,
|
| 936 |
+
"epoch": 0.12339331619537275,
|
| 937 |
+
"grad_norm": 0.1408892422914505,
|
| 938 |
+
"kl": 0.0006256103515625,
|
| 939 |
+
"learning_rate": 2.9836319343816397e-07,
|
| 940 |
+
"loss": 0.0,
|
| 941 |
+
"reward": 0.0009416388347744942,
|
| 942 |
+
"reward_std": 0.1657393015921116,
|
| 943 |
+
"rewards/cosine_scaled_reward": -0.15875469404272735,
|
| 944 |
+
"rewards/format_reward": 0.3229166744276881,
|
| 945 |
+
"step": 72
|
| 946 |
+
},
|
| 947 |
+
{
|
| 948 |
+
"completion_length": 2957.0625610351562,
|
| 949 |
+
"epoch": 0.12510711225364182,
|
| 950 |
+
"grad_norm": 0.13583190739154816,
|
| 951 |
+
"kl": 0.00033473968505859375,
|
| 952 |
+
"learning_rate": 2.854966364683872e-07,
|
| 953 |
+
"loss": 0.0,
|
| 954 |
+
"reward": 0.06538257002830505,
|
| 955 |
+
"reward_std": 0.1970806121826172,
|
| 956 |
+
"rewards/cosine_scaled_reward": -0.09460783936083317,
|
| 957 |
+
"rewards/format_reward": 0.447916679084301,
|
| 958 |
+
"step": 73
|
| 959 |
+
},
|
| 960 |
+
{
|
| 961 |
+
"completion_length": 3106.010498046875,
|
| 962 |
+
"epoch": 0.1268209083119109,
|
| 963 |
+
"grad_norm": 0.15821737051010132,
|
| 964 |
+
"kl": 0.0009412765502929688,
|
| 965 |
+
"learning_rate": 2.729523361034538e-07,
|
| 966 |
+
"loss": 0.0,
|
| 967 |
+
"reward": 0.0028651878237724304,
|
| 968 |
+
"reward_std": 0.19838279485702515,
|
| 969 |
+
"rewards/cosine_scaled_reward": -0.13493257109075785,
|
| 970 |
+
"rewards/format_reward": 0.2812500111758709,
|
| 971 |
+
"step": 74
|
| 972 |
+
},
|
| 973 |
+
{
|
| 974 |
+
"completion_length": 2847.541748046875,
|
| 975 |
+
"epoch": 0.12853470437017994,
|
| 976 |
+
"grad_norm": 0.16942313313484192,
|
| 977 |
+
"kl": 0.0019121170043945312,
|
| 978 |
+
"learning_rate": 2.6074557564105724e-07,
|
| 979 |
+
"loss": 0.0001,
|
| 980 |
+
"reward": 0.0911966897547245,
|
| 981 |
+
"reward_std": 0.21281348168849945,
|
| 982 |
+
"rewards/cosine_scaled_reward": -0.028547534719109535,
|
| 983 |
+
"rewards/format_reward": 0.41666668467223644,
|
| 984 |
+
"step": 75
|
| 985 |
+
},
|
| 986 |
+
{
|
| 987 |
+
"completion_length": 2846.9896850585938,
|
| 988 |
+
"epoch": 0.13024850042844902,
|
| 989 |
+
"grad_norm": 0.1670897901058197,
|
| 990 |
+
"kl": 0.0010271072387695312,
|
| 991 |
+
"learning_rate": 2.488912271385139e-07,
|
| 992 |
+
"loss": 0.0,
|
| 993 |
+
"reward": 0.02618580963462591,
|
| 994 |
+
"reward_std": 0.15880979225039482,
|
| 995 |
+
"rewards/cosine_scaled_reward": -0.18800780980382115,
|
| 996 |
+
"rewards/format_reward": 0.4791666828095913,
|
| 997 |
+
"step": 76
|
| 998 |
+
},
|
| 999 |
+
{
|
| 1000 |
+
"completion_length": 3011.2188110351562,
|
| 1001 |
+
"epoch": 0.1319622964867181,
|
| 1002 |
+
"grad_norm": 0.1846207082271576,
|
| 1003 |
+
"kl": 0.0016541481018066406,
|
| 1004 |
+
"learning_rate": 2.374037332934512e-07,
|
| 1005 |
+
"loss": 0.0001,
|
| 1006 |
+
"reward": 0.01657121942844242,
|
| 1007 |
+
"reward_std": 0.19087375327944756,
|
| 1008 |
+
"rewards/cosine_scaled_reward": -0.10442618932574987,
|
| 1009 |
+
"rewards/format_reward": 0.27083334513008595,
|
| 1010 |
+
"step": 77
|
| 1011 |
+
},
|
| 1012 |
+
{
|
| 1013 |
+
"completion_length": 3298.625,
|
| 1014 |
+
"epoch": 0.13367609254498714,
|
| 1015 |
+
"grad_norm": 0.12679187953472137,
|
| 1016 |
+
"kl": 0.0008182525634765625,
|
| 1017 |
+
"learning_rate": 2.2629708984760706e-07,
|
| 1018 |
+
"loss": 0.0,
|
| 1019 |
+
"reward": -0.07863575895316899,
|
| 1020 |
+
"reward_std": 0.11674198694527149,
|
| 1021 |
+
"rewards/cosine_scaled_reward": -0.24837390705943108,
|
| 1022 |
+
"rewards/format_reward": 0.1875000074505806,
|
| 1023 |
+
"step": 78
|
| 1024 |
+
},
|
| 1025 |
+
{
|
| 1026 |
+
"completion_length": 3014.0521850585938,
|
| 1027 |
+
"epoch": 0.1353898886032562,
|
| 1028 |
+
"grad_norm": 0.19050873816013336,
|
| 1029 |
+
"kl": 0.0012989044189453125,
|
| 1030 |
+
"learning_rate": 2.1558482853517253e-07,
|
| 1031 |
+
"loss": 0.0001,
|
| 1032 |
+
"reward": 0.14426555391401052,
|
| 1033 |
+
"reward_std": 0.21018010936677456,
|
| 1034 |
+
"rewards/cosine_scaled_reward": 0.08679820597171783,
|
| 1035 |
+
"rewards/format_reward": 0.3958333507180214,
|
| 1036 |
+
"step": 79
|
| 1037 |
+
},
|
| 1038 |
+
{
|
| 1039 |
+
"completion_length": 2568.697998046875,
|
| 1040 |
+
"epoch": 0.13710368466152528,
|
| 1041 |
+
"grad_norm": 0.14511092007160187,
|
| 1042 |
+
"kl": 0.0016317367553710938,
|
| 1043 |
+
"learning_rate": 2.0528000059645995e-07,
|
| 1044 |
+
"loss": 0.0001,
|
| 1045 |
+
"reward": 0.15173329645767808,
|
| 1046 |
+
"reward_std": 0.1817643828690052,
|
| 1047 |
+
"rewards/cosine_scaled_reward": 0.00575945433229208,
|
| 1048 |
+
"rewards/format_reward": 0.5833333432674408,
|
| 1049 |
+
"step": 80
|
| 1050 |
+
},
|
| 1051 |
+
{
|
| 1052 |
+
"completion_length": 2273.5209045410156,
|
| 1053 |
+
"epoch": 0.13881748071979436,
|
| 1054 |
+
"grad_norm": 0.1482679694890976,
|
| 1055 |
+
"kl": 0.0021762847900390625,
|
| 1056 |
+
"learning_rate": 1.9539516087697517e-07,
|
| 1057 |
+
"loss": 0.0001,
|
| 1058 |
+
"reward": 0.17546646296977997,
|
| 1059 |
+
"reward_std": 0.18799066171050072,
|
| 1060 |
+
"rewards/cosine_scaled_reward": 0.05814665276557207,
|
| 1061 |
+
"rewards/format_reward": 0.5729166865348816,
|
| 1062 |
+
"step": 81
|
| 1063 |
+
},
|
| 1064 |
+
{
|
| 1065 |
+
"completion_length": 2761.7709350585938,
|
| 1066 |
+
"epoch": 0.1405312767780634,
|
| 1067 |
+
"grad_norm": 0.14379660785198212,
|
| 1068 |
+
"kl": 0.000904083251953125,
|
| 1069 |
+
"learning_rate": 1.8594235253127372e-07,
|
| 1070 |
+
"loss": 0.0,
|
| 1071 |
+
"reward": 0.09056159481406212,
|
| 1072 |
+
"reward_std": 0.19218357279896736,
|
| 1073 |
+
"rewards/cosine_scaled_reward": -0.07696967083029449,
|
| 1074 |
+
"rewards/format_reward": 0.5104166865348816,
|
| 1075 |
+
"step": 82
|
| 1076 |
+
},
|
| 1077 |
+
{
|
| 1078 |
+
"completion_length": 2424.1666870117188,
|
| 1079 |
+
"epoch": 0.14224507283633248,
|
| 1080 |
+
"grad_norm": 0.18656377494335175,
|
| 1081 |
+
"kl": 0.0023345947265625,
|
| 1082 |
+
"learning_rate": 1.7693309235023127e-07,
|
| 1083 |
+
"loss": 0.0001,
|
| 1084 |
+
"reward": 0.17556490686547477,
|
| 1085 |
+
"reward_std": 0.24106226861476898,
|
| 1086 |
+
"rewards/cosine_scaled_reward": 0.06774241663515568,
|
| 1087 |
+
"rewards/format_reward": 0.5520833432674408,
|
| 1088 |
+
"step": 83
|
| 1089 |
+
},
|
| 1090 |
+
{
|
| 1091 |
+
"completion_length": 2954.041748046875,
|
| 1092 |
+
"epoch": 0.14395886889460155,
|
| 1093 |
+
"grad_norm": 0.12463019788265228,
|
| 1094 |
+
"kl": 0.0012934207916259766,
|
| 1095 |
+
"learning_rate": 1.6837835672960831e-07,
|
| 1096 |
+
"loss": 0.0001,
|
| 1097 |
+
"reward": 0.07739502191543579,
|
| 1098 |
+
"reward_std": 0.14365380816161633,
|
| 1099 |
+
"rewards/cosine_scaled_reward": -0.024237127974629402,
|
| 1100 |
+
"rewards/format_reward": 0.3541666716337204,
|
| 1101 |
+
"step": 84
|
| 1102 |
+
},
|
| 1103 |
+
{
|
| 1104 |
+
"completion_length": 3007.1771850585938,
|
| 1105 |
+
"epoch": 0.1456726649528706,
|
| 1106 |
+
"grad_norm": 0.1296854168176651,
|
| 1107 |
+
"kl": 0.0006504058837890625,
|
| 1108 |
+
"learning_rate": 1.6028856829700258e-07,
|
| 1109 |
+
"loss": 0.0,
|
| 1110 |
+
"reward": 0.06423507491126657,
|
| 1111 |
+
"reward_std": 0.18408943712711334,
|
| 1112 |
+
"rewards/cosine_scaled_reward": -0.08204184658825397,
|
| 1113 |
+
"rewards/format_reward": 0.416666679084301,
|
| 1114 |
+
"step": 85
|
| 1115 |
+
},
|
| 1116 |
+
{
|
| 1117 |
+
"completion_length": 3371.8646240234375,
|
| 1118 |
+
"epoch": 0.14738646101113967,
|
| 1119 |
+
"grad_norm": 0.1433609277009964,
|
| 1120 |
+
"kl": 0.0012350082397460938,
|
| 1121 |
+
"learning_rate": 1.5267358321348285e-07,
|
| 1122 |
+
"loss": 0.0,
|
| 1123 |
+
"reward": -0.03188338316977024,
|
| 1124 |
+
"reward_std": 0.16921599954366684,
|
| 1125 |
+
"rewards/cosine_scaled_reward": -0.13588787615299225,
|
| 1126 |
+
"rewards/format_reward": 0.14583333861082792,
|
| 1127 |
+
"step": 86
|
| 1128 |
+
},
|
| 1129 |
+
{
|
| 1130 |
+
"completion_length": 3052.2501220703125,
|
| 1131 |
+
"epoch": 0.14910025706940874,
|
| 1132 |
+
"grad_norm": 0.13880982995033264,
|
| 1133 |
+
"kl": 0.001392364501953125,
|
| 1134 |
+
"learning_rate": 1.4554267916537495e-07,
|
| 1135 |
+
"loss": 0.0001,
|
| 1136 |
+
"reward": -0.018908033147454262,
|
| 1137 |
+
"reward_std": 0.1546711027622223,
|
| 1138 |
+
"rewards/cosine_scaled_reward": -0.18234001845121384,
|
| 1139 |
+
"rewards/format_reward": 0.2916666753590107,
|
| 1140 |
+
"step": 87
|
| 1141 |
+
},
|
| 1142 |
+
{
|
| 1143 |
+
"completion_length": 2693.0626220703125,
|
| 1144 |
+
"epoch": 0.15081405312767782,
|
| 1145 |
+
"grad_norm": 0.15144215524196625,
|
| 1146 |
+
"kl": 0.0007927417755126953,
|
| 1147 |
+
"learning_rate": 1.3890454406082956e-07,
|
| 1148 |
+
"loss": 0.0,
|
| 1149 |
+
"reward": 0.12248719483613968,
|
| 1150 |
+
"reward_std": 0.18367315083742142,
|
| 1151 |
+
"rewards/cosine_scaled_reward": 0.043038370087742805,
|
| 1152 |
+
"rewards/format_reward": 0.3958333507180214,
|
| 1153 |
+
"step": 88
|
| 1154 |
+
},
|
| 1155 |
+
{
|
| 1156 |
+
"completion_length": 3063.604248046875,
|
| 1157 |
+
"epoch": 0.15252784918594686,
|
| 1158 |
+
"grad_norm": 0.14191189408302307,
|
| 1159 |
+
"kl": 0.0008802413940429688,
|
| 1160 |
+
"learning_rate": 1.3276726544494571e-07,
|
| 1161 |
+
"loss": 0.0,
|
| 1162 |
+
"reward": 0.0365169154247269,
|
| 1163 |
+
"reward_std": 0.15275101736187935,
|
| 1164 |
+
"rewards/cosine_scaled_reward": -0.0832907222211361,
|
| 1165 |
+
"rewards/format_reward": 0.31250000558793545,
|
| 1166 |
+
"step": 89
|
| 1167 |
+
},
|
| 1168 |
+
{
|
| 1169 |
+
"completion_length": 2311.010467529297,
|
| 1170 |
+
"epoch": 0.15424164524421594,
|
| 1171 |
+
"grad_norm": 0.17067813873291016,
|
| 1172 |
+
"kl": 0.0018815994262695312,
|
| 1173 |
+
"learning_rate": 1.2713832064634125e-07,
|
| 1174 |
+
"loss": 0.0001,
|
| 1175 |
+
"reward": 0.1953146681189537,
|
| 1176 |
+
"reward_std": 0.2011973224580288,
|
| 1177 |
+
"rewards/cosine_scaled_reward": 0.08297107368707657,
|
| 1178 |
+
"rewards/format_reward": 0.6041666716337204,
|
| 1179 |
+
"step": 90
|
| 1180 |
+
},
|
| 1181 |
+
{
|
| 1182 |
+
"completion_length": 2986.729248046875,
|
| 1183 |
+
"epoch": 0.155955441302485,
|
| 1184 |
+
"grad_norm": 0.1307849884033203,
|
| 1185 |
+
"kl": 0.0007395744323730469,
|
| 1186 |
+
"learning_rate": 1.220245676671809e-07,
|
| 1187 |
+
"loss": 0.0,
|
| 1188 |
+
"reward": 0.05273827910423279,
|
| 1189 |
+
"reward_std": 0.22075917199254036,
|
| 1190 |
+
"rewards/cosine_scaled_reward": -0.054260533303022385,
|
| 1191 |
+
"rewards/format_reward": 0.3125000102445483,
|
| 1192 |
+
"step": 91
|
| 1193 |
+
},
|
| 1194 |
+
{
|
| 1195 |
+
"completion_length": 2813.3543090820312,
|
| 1196 |
+
"epoch": 0.15766923736075408,
|
| 1197 |
+
"grad_norm": 0.13788200914859772,
|
| 1198 |
+
"kl": 0.0018596649169921875,
|
| 1199 |
+
"learning_rate": 1.1743223682775649e-07,
|
| 1200 |
+
"loss": 0.0001,
|
| 1201 |
+
"reward": 0.02730799838900566,
|
| 1202 |
+
"reward_std": 0.15018530935049057,
|
| 1203 |
+
"rewards/cosine_scaled_reward": -0.13331225764704868,
|
| 1204 |
+
"rewards/format_reward": 0.375,
|
| 1205 |
+
"step": 92
|
| 1206 |
+
},
|
| 1207 |
+
{
|
| 1208 |
+
"completion_length": 3271.2500610351562,
|
| 1209 |
+
"epoch": 0.15938303341902313,
|
| 1210 |
+
"grad_norm": 0.1422564685344696,
|
| 1211 |
+
"kl": 0.001224517822265625,
|
| 1212 |
+
"learning_rate": 1.1336692317580158e-07,
|
| 1213 |
+
"loss": 0.0,
|
| 1214 |
+
"reward": 0.08444089069962502,
|
| 1215 |
+
"reward_std": 0.18314170092344284,
|
| 1216 |
+
"rewards/cosine_scaled_reward": 0.026102278381586075,
|
| 1217 |
+
"rewards/format_reward": 0.2812500037252903,
|
| 1218 |
+
"step": 93
|
| 1219 |
+
},
|
| 1220 |
+
{
|
| 1221 |
+
"completion_length": 3040.9688720703125,
|
| 1222 |
+
"epoch": 0.1610968294772922,
|
| 1223 |
+
"grad_norm": 0.1662042886018753,
|
| 1224 |
+
"kl": 0.0022430419921875,
|
| 1225 |
+
"learning_rate": 1.0983357966978745e-07,
|
| 1226 |
+
"loss": 0.0001,
|
| 1227 |
+
"reward": 0.028276806697249413,
|
| 1228 |
+
"reward_std": 0.17833666875958443,
|
| 1229 |
+
"rewards/cosine_scaled_reward": -0.14247922599315643,
|
| 1230 |
+
"rewards/format_reward": 0.3958333544433117,
|
| 1231 |
+
"step": 94
|
| 1232 |
+
},
|
| 1233 |
+
{
|
| 1234 |
+
"completion_length": 3048.197998046875,
|
| 1235 |
+
"epoch": 0.16281062553556128,
|
| 1236 |
+
"grad_norm": 0.12611675262451172,
|
| 1237 |
+
"kl": 0.0008878707885742188,
|
| 1238 |
+
"learning_rate": 1.068365111445064e-07,
|
| 1239 |
+
"loss": 0.0,
|
| 1240 |
+
"reward": 0.07369702425785363,
|
| 1241 |
+
"reward_std": 0.18385101296007633,
|
| 1242 |
+
"rewards/cosine_scaled_reward": -0.07403519004583359,
|
| 1243 |
+
"rewards/format_reward": 0.4375000074505806,
|
| 1244 |
+
"step": 95
|
| 1245 |
+
},
|
| 1246 |
+
{
|
| 1247 |
+
"completion_length": 3014.7918090820312,
|
| 1248 |
+
"epoch": 0.16452442159383032,
|
| 1249 |
+
"grad_norm": 0.16144989430904388,
|
| 1250 |
+
"kl": 0.0016870498657226562,
|
| 1251 |
+
"learning_rate": 1.0437936906629334e-07,
|
| 1252 |
+
"loss": 0.0001,
|
| 1253 |
+
"reward": 0.07583070383407176,
|
| 1254 |
+
"reward_std": 0.21927380189299583,
|
| 1255 |
+
"rewards/cosine_scaled_reward": -0.03254310414195061,
|
| 1256 |
+
"rewards/format_reward": 0.3645833395421505,
|
| 1257 |
+
"step": 96
|
| 1258 |
+
},
|
| 1259 |
+
{
|
| 1260 |
+
"completion_length": 2385.0521240234375,
|
| 1261 |
+
"epoch": 0.1662382176520994,
|
| 1262 |
+
"grad_norm": 0.15538226068019867,
|
| 1263 |
+
"kl": 0.0020532608032226562,
|
| 1264 |
+
"learning_rate": 1.0246514708427701e-07,
|
| 1265 |
+
"loss": 0.0001,
|
| 1266 |
+
"reward": 0.09558990353252739,
|
| 1267 |
+
"reward_std": 0.14071721211075783,
|
| 1268 |
+
"rewards/cosine_scaled_reward": -0.1022979412227869,
|
| 1269 |
+
"rewards/format_reward": 0.5833333507180214,
|
| 1270 |
+
"step": 97
|
| 1271 |
+
},
|
| 1272 |
+
{
|
| 1273 |
+
"completion_length": 2722.2500610351562,
|
| 1274 |
+
"epoch": 0.16795201371036847,
|
| 1275 |
+
"grad_norm": 0.16014371812343597,
|
| 1276 |
+
"kl": 0.0013885498046875,
|
| 1277 |
+
"learning_rate": 1.0109617738307911e-07,
|
| 1278 |
+
"loss": 0.0001,
|
| 1279 |
+
"reward": 0.024186712922528386,
|
| 1280 |
+
"reward_std": 0.16764022037386894,
|
| 1281 |
+
"rewards/cosine_scaled_reward": -0.1459345854818821,
|
| 1282 |
+
"rewards/format_reward": 0.385416679084301,
|
| 1283 |
+
"step": 98
|
| 1284 |
+
},
|
| 1285 |
+
{
|
| 1286 |
+
"completion_length": 2599.0625610351562,
|
| 1287 |
+
"epoch": 0.16966580976863754,
|
| 1288 |
+
"grad_norm": 0.132161945104599,
|
| 1289 |
+
"kl": 0.0023794174194335938,
|
| 1290 |
+
"learning_rate": 1.002741278414069e-07,
|
| 1291 |
+
"loss": 0.0001,
|
| 1292 |
+
"reward": 0.11149843037128448,
|
| 1293 |
+
"reward_std": 0.1786147654056549,
|
| 1294 |
+
"rewards/cosine_scaled_reward": -0.06053047860041261,
|
| 1295 |
+
"rewards/format_reward": 0.5625000149011612,
|
| 1296 |
+
"step": 99
|
| 1297 |
+
},
|
| 1298 |
+
{
|
| 1299 |
+
"completion_length": 2782.8126220703125,
|
| 1300 |
+
"epoch": 0.1713796058269066,
|
| 1301 |
+
"grad_norm": 0.15893962979316711,
|
| 1302 |
+
"kl": 0.0020580291748046875,
|
| 1303 |
+
"learning_rate": 1e-07,
|
| 1304 |
+
"loss": 0.0001,
|
| 1305 |
+
"reward": 0.0819746982306242,
|
| 1306 |
+
"reward_std": 0.2312169447541237,
|
| 1307 |
+
"rewards/cosine_scaled_reward": -0.04185225034598261,
|
| 1308 |
+
"rewards/format_reward": 0.4062500074505806,
|
| 1309 |
+
"step": 100
|
| 1310 |
+
},
|
| 1311 |
+
{
|
| 1312 |
+
"epoch": 0.1713796058269066,
|
| 1313 |
+
"step": 100,
|
| 1314 |
+
"total_flos": 0.0,
|
| 1315 |
+
"train_loss": 3.568841239768084e-05,
|
| 1316 |
+
"train_runtime": 22061.4799,
|
| 1317 |
+
"train_samples_per_second": 0.435,
|
| 1318 |
+
"train_steps_per_second": 0.005
|
| 1319 |
+
}
|
| 1320 |
+
],
|
| 1321 |
+
"logging_steps": 1,
|
| 1322 |
+
"max_steps": 100,
|
| 1323 |
+
"num_input_tokens_seen": 0,
|
| 1324 |
+
"num_train_epochs": 1,
|
| 1325 |
+
"save_steps": 50,
|
| 1326 |
+
"stateful_callbacks": {
|
| 1327 |
+
"TrainerControl": {
|
| 1328 |
+
"args": {
|
| 1329 |
+
"should_epoch_stop": false,
|
| 1330 |
+
"should_evaluate": false,
|
| 1331 |
+
"should_log": false,
|
| 1332 |
+
"should_save": true,
|
| 1333 |
+
"should_training_stop": true
|
| 1334 |
+
},
|
| 1335 |
+
"attributes": {}
|
| 1336 |
+
}
|
| 1337 |
+
},
|
| 1338 |
+
"total_flos": 0.0,
|
| 1339 |
+
"train_batch_size": 4,
|
| 1340 |
+
"trial_name": null,
|
| 1341 |
+
"trial_params": null
|
| 1342 |
+
}
|