Model save
Browse files- README.md +68 -0
- all_results.json +8 -0
- generation_config.json +6 -0
- train_results.json +8 -0
- trainer_state.json +1567 -0
README.md
ADDED
|
@@ -0,0 +1,68 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
base_model: allenai/OLMoE-1B-7B-0125-Instruct
|
| 3 |
+
library_name: transformers
|
| 4 |
+
model_name: OLMoE-1B-7B-0125-Instruct-grpo-E8-D8000
|
| 5 |
+
tags:
|
| 6 |
+
- generated_from_trainer
|
| 7 |
+
- trl
|
| 8 |
+
- grpo
|
| 9 |
+
licence: license
|
| 10 |
+
---
|
| 11 |
+
|
| 12 |
+
# Model Card for OLMoE-1B-7B-0125-Instruct-grpo-E8-D8000
|
| 13 |
+
|
| 14 |
+
This model is a fine-tuned version of [allenai/OLMoE-1B-7B-0125-Instruct](https://huggingface.co/allenai/OLMoE-1B-7B-0125-Instruct).
|
| 15 |
+
It has been trained using [TRL](https://github.com/huggingface/trl).
|
| 16 |
+
|
| 17 |
+
## Quick start
|
| 18 |
+
|
| 19 |
+
```python
|
| 20 |
+
from transformers import pipeline
|
| 21 |
+
|
| 22 |
+
question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
|
| 23 |
+
generator = pipeline("text-generation", model="chenggong1995/OLMoE-1B-7B-0125-Instruct-grpo-E8-D8000", 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/gongc1995-city-university-of-hong-kong/huggingface/runs/0scgpt1n)
|
| 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: 3.3.2
|
| 41 |
+
- Tokenizers: 0.21.0
|
| 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": 0.04277147677156233,
|
| 4 |
+
"train_runtime": 16581.2814,
|
| 5 |
+
"train_samples": 8000,
|
| 6 |
+
"train_samples_per_second": 3.86,
|
| 7 |
+
"train_steps_per_second": 0.03
|
| 8 |
+
}
|
generation_config.json
ADDED
|
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"eos_token_id": 50279,
|
| 4 |
+
"pad_token_id": 1,
|
| 5 |
+
"transformers_version": "4.49.0"
|
| 6 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"total_flos": 0.0,
|
| 3 |
+
"train_loss": 0.04277147677156233,
|
| 4 |
+
"train_runtime": 16581.2814,
|
| 5 |
+
"train_samples": 8000,
|
| 6 |
+
"train_samples_per_second": 3.86,
|
| 7 |
+
"train_steps_per_second": 0.03
|
| 8 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,1567 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 7.992,
|
| 5 |
+
"eval_steps": 100,
|
| 6 |
+
"global_step": 496,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"clip_ratio": 0.0,
|
| 13 |
+
"completion_length": 575.0117378234863,
|
| 14 |
+
"epoch": 0.08,
|
| 15 |
+
"grad_norm": 0.5159733891487122,
|
| 16 |
+
"kl": 0.0005777359008789062,
|
| 17 |
+
"learning_rate": 2e-07,
|
| 18 |
+
"loss": 0.0211,
|
| 19 |
+
"reward": 0.41477865651249884,
|
| 20 |
+
"reward_std": 0.3093508366495371,
|
| 21 |
+
"rewards/accuracy_reward": 0.08098958514165133,
|
| 22 |
+
"rewards/format_reward": 0.007812500209547579,
|
| 23 |
+
"rewards/tag_count_reward": 0.32597657293081284,
|
| 24 |
+
"step": 5
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"clip_ratio": 0.0,
|
| 28 |
+
"completion_length": 569.7036651611328,
|
| 29 |
+
"epoch": 0.16,
|
| 30 |
+
"grad_norm": 0.5224132537841797,
|
| 31 |
+
"kl": 0.001135540008544922,
|
| 32 |
+
"learning_rate": 4e-07,
|
| 33 |
+
"loss": 0.0197,
|
| 34 |
+
"reward": 0.41028647050261496,
|
| 35 |
+
"reward_std": 0.30269680023193357,
|
| 36 |
+
"rewards/accuracy_reward": 0.07421875218860804,
|
| 37 |
+
"rewards/format_reward": 0.005729166814126075,
|
| 38 |
+
"rewards/tag_count_reward": 0.3303385525941849,
|
| 39 |
+
"step": 10
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"clip_ratio": 0.0,
|
| 43 |
+
"completion_length": 586.758869934082,
|
| 44 |
+
"epoch": 0.24,
|
| 45 |
+
"grad_norm": 0.3704665005207062,
|
| 46 |
+
"kl": 0.0023041725158691405,
|
| 47 |
+
"learning_rate": 6e-07,
|
| 48 |
+
"loss": 0.0267,
|
| 49 |
+
"reward": 0.473372408002615,
|
| 50 |
+
"reward_std": 0.3123403754085302,
|
| 51 |
+
"rewards/accuracy_reward": 0.08750000239815563,
|
| 52 |
+
"rewards/format_reward": 0.009635416907258332,
|
| 53 |
+
"rewards/tag_count_reward": 0.3762369878590107,
|
| 54 |
+
"step": 15
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"clip_ratio": 0.0,
|
| 58 |
+
"completion_length": 589.0830902099609,
|
| 59 |
+
"epoch": 0.32,
|
| 60 |
+
"grad_norm": 1.0813833475112915,
|
| 61 |
+
"kl": 0.0357421875,
|
| 62 |
+
"learning_rate": 8e-07,
|
| 63 |
+
"loss": 0.032,
|
| 64 |
+
"reward": 0.6087890841066838,
|
| 65 |
+
"reward_std": 0.32254689410328863,
|
| 66 |
+
"rewards/accuracy_reward": 0.07447916907258331,
|
| 67 |
+
"rewards/format_reward": 0.025781250488944352,
|
| 68 |
+
"rewards/tag_count_reward": 0.5085286632180214,
|
| 69 |
+
"step": 20
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"clip_ratio": 0.0,
|
| 73 |
+
"completion_length": 598.2661651611328,
|
| 74 |
+
"epoch": 0.4,
|
| 75 |
+
"grad_norm": 0.3315693140029907,
|
| 76 |
+
"kl": 0.0186279296875,
|
| 77 |
+
"learning_rate": 1e-06,
|
| 78 |
+
"loss": 0.0404,
|
| 79 |
+
"reward": 0.6679036632180214,
|
| 80 |
+
"reward_std": 0.3646054796874523,
|
| 81 |
+
"rewards/accuracy_reward": 0.08750000272411854,
|
| 82 |
+
"rewards/format_reward": 0.044791668001562354,
|
| 83 |
+
"rewards/tag_count_reward": 0.5356119990348815,
|
| 84 |
+
"step": 25
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"clip_ratio": 0.0,
|
| 88 |
+
"completion_length": 603.8648612976074,
|
| 89 |
+
"epoch": 0.48,
|
| 90 |
+
"grad_norm": 0.34996700286865234,
|
| 91 |
+
"kl": 0.0280853271484375,
|
| 92 |
+
"learning_rate": 1.2e-06,
|
| 93 |
+
"loss": 0.0484,
|
| 94 |
+
"reward": 0.7606770992279053,
|
| 95 |
+
"reward_std": 0.42633651196956635,
|
| 96 |
+
"rewards/accuracy_reward": 0.0877604190260172,
|
| 97 |
+
"rewards/format_reward": 0.08750000251457095,
|
| 98 |
+
"rewards/tag_count_reward": 0.5854166865348815,
|
| 99 |
+
"step": 30
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"clip_ratio": 0.0,
|
| 103 |
+
"completion_length": 598.5612190246582,
|
| 104 |
+
"epoch": 0.56,
|
| 105 |
+
"grad_norm": 0.3991927206516266,
|
| 106 |
+
"kl": 0.0320770263671875,
|
| 107 |
+
"learning_rate": 1.4e-06,
|
| 108 |
+
"loss": 0.0474,
|
| 109 |
+
"reward": 0.8298828288912773,
|
| 110 |
+
"reward_std": 0.46221103593707086,
|
| 111 |
+
"rewards/accuracy_reward": 0.07890625211875886,
|
| 112 |
+
"rewards/format_reward": 0.12395833758637309,
|
| 113 |
+
"rewards/tag_count_reward": 0.6270182520151139,
|
| 114 |
+
"step": 35
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"clip_ratio": 0.0,
|
| 118 |
+
"completion_length": 599.6370010375977,
|
| 119 |
+
"epoch": 0.64,
|
| 120 |
+
"grad_norm": 0.32159847021102905,
|
| 121 |
+
"kl": 0.0319732666015625,
|
| 122 |
+
"learning_rate": 1.6e-06,
|
| 123 |
+
"loss": 0.057,
|
| 124 |
+
"reward": 1.0265625357627868,
|
| 125 |
+
"reward_std": 0.5343712478876114,
|
| 126 |
+
"rewards/accuracy_reward": 0.06770833521150052,
|
| 127 |
+
"rewards/format_reward": 0.25104167461395266,
|
| 128 |
+
"rewards/tag_count_reward": 0.7078125223517417,
|
| 129 |
+
"step": 40
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"clip_ratio": 0.0,
|
| 133 |
+
"completion_length": 579.9875183105469,
|
| 134 |
+
"epoch": 0.72,
|
| 135 |
+
"grad_norm": 0.3186890482902527,
|
| 136 |
+
"kl": 0.0441070556640625,
|
| 137 |
+
"learning_rate": 1.8e-06,
|
| 138 |
+
"loss": 0.0737,
|
| 139 |
+
"reward": 1.2731771171092987,
|
| 140 |
+
"reward_std": 0.5938673868775368,
|
| 141 |
+
"rewards/accuracy_reward": 0.058072918513789776,
|
| 142 |
+
"rewards/format_reward": 0.446614596247673,
|
| 143 |
+
"rewards/tag_count_reward": 0.7684896036982536,
|
| 144 |
+
"step": 45
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"clip_ratio": 0.0,
|
| 148 |
+
"completion_length": 568.0612159729004,
|
| 149 |
+
"epoch": 0.8,
|
| 150 |
+
"grad_norm": 0.28663673996925354,
|
| 151 |
+
"kl": 0.0451324462890625,
|
| 152 |
+
"learning_rate": 2e-06,
|
| 153 |
+
"loss": 0.0684,
|
| 154 |
+
"reward": 1.470442745089531,
|
| 155 |
+
"reward_std": 0.5922676548361778,
|
| 156 |
+
"rewards/accuracy_reward": 0.06432291842065752,
|
| 157 |
+
"rewards/format_reward": 0.5750000163912773,
|
| 158 |
+
"rewards/tag_count_reward": 0.8311198085546494,
|
| 159 |
+
"step": 50
|
| 160 |
+
},
|
| 161 |
+
{
|
| 162 |
+
"clip_ratio": 0.0,
|
| 163 |
+
"completion_length": 539.609130859375,
|
| 164 |
+
"epoch": 0.88,
|
| 165 |
+
"grad_norm": 0.2857872247695923,
|
| 166 |
+
"kl": 0.0481170654296875,
|
| 167 |
+
"learning_rate": 1.9993798522846506e-06,
|
| 168 |
+
"loss": 0.0487,
|
| 169 |
+
"reward": 1.6721354514360427,
|
| 170 |
+
"reward_std": 0.5311627298593521,
|
| 171 |
+
"rewards/accuracy_reward": 0.06093750160653144,
|
| 172 |
+
"rewards/format_reward": 0.7151041895151138,
|
| 173 |
+
"rewards/tag_count_reward": 0.8960937678813934,
|
| 174 |
+
"step": 55
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"clip_ratio": 0.0,
|
| 178 |
+
"completion_length": 541.1231971740723,
|
| 179 |
+
"epoch": 0.96,
|
| 180 |
+
"grad_norm": 0.273100882768631,
|
| 181 |
+
"kl": 0.05677490234375,
|
| 182 |
+
"learning_rate": 1.9975201783049802e-06,
|
| 183 |
+
"loss": 0.0228,
|
| 184 |
+
"reward": 1.768489620089531,
|
| 185 |
+
"reward_std": 0.4786494873464108,
|
| 186 |
+
"rewards/accuracy_reward": 0.052343751583248374,
|
| 187 |
+
"rewards/format_reward": 0.7924479350447655,
|
| 188 |
+
"rewards/tag_count_reward": 0.9236979350447655,
|
| 189 |
+
"step": 60
|
| 190 |
+
},
|
| 191 |
+
{
|
| 192 |
+
"clip_ratio": 0.0,
|
| 193 |
+
"completion_length": 534.5526145935058,
|
| 194 |
+
"epoch": 1.048,
|
| 195 |
+
"grad_norm": 0.23924335837364197,
|
| 196 |
+
"kl": 0.053448486328125,
|
| 197 |
+
"learning_rate": 1.994423284606128e-06,
|
| 198 |
+
"loss": 0.0239,
|
| 199 |
+
"reward": 1.8424479603767394,
|
| 200 |
+
"reward_std": 0.4263153623789549,
|
| 201 |
+
"rewards/accuracy_reward": 0.06171875204890966,
|
| 202 |
+
"rewards/format_reward": 0.8372396007180214,
|
| 203 |
+
"rewards/tag_count_reward": 0.9434895992279053,
|
| 204 |
+
"step": 65
|
| 205 |
+
},
|
| 206 |
+
{
|
| 207 |
+
"clip_ratio": 0.0,
|
| 208 |
+
"completion_length": 553.1002761840821,
|
| 209 |
+
"epoch": 1.1280000000000001,
|
| 210 |
+
"grad_norm": 0.24568906426429749,
|
| 211 |
+
"kl": 0.0635986328125,
|
| 212 |
+
"learning_rate": 1.990093012251199e-06,
|
| 213 |
+
"loss": 0.0147,
|
| 214 |
+
"reward": 1.8621745228767395,
|
| 215 |
+
"reward_std": 0.37093897461891173,
|
| 216 |
+
"rewards/accuracy_reward": 0.0471354179084301,
|
| 217 |
+
"rewards/format_reward": 0.8682291880249977,
|
| 218 |
+
"rewards/tag_count_reward": 0.9468099132180214,
|
| 219 |
+
"step": 70
|
| 220 |
+
},
|
| 221 |
+
{
|
| 222 |
+
"clip_ratio": 0.0,
|
| 223 |
+
"completion_length": 502.05652542114257,
|
| 224 |
+
"epoch": 1.208,
|
| 225 |
+
"grad_norm": 0.2926892638206482,
|
| 226 |
+
"kl": 0.06356201171875,
|
| 227 |
+
"learning_rate": 1.9845347320572078e-06,
|
| 228 |
+
"loss": 0.0112,
|
| 229 |
+
"reward": 1.8821615010499955,
|
| 230 |
+
"reward_std": 0.3585973784327507,
|
| 231 |
+
"rewards/accuracy_reward": 0.05546875158324838,
|
| 232 |
+
"rewards/format_reward": 0.873697929084301,
|
| 233 |
+
"rewards/tag_count_reward": 0.9529948100447655,
|
| 234 |
+
"step": 75
|
| 235 |
+
},
|
| 236 |
+
{
|
| 237 |
+
"clip_ratio": 0.0,
|
| 238 |
+
"completion_length": 495.64324111938475,
|
| 239 |
+
"epoch": 1.288,
|
| 240 |
+
"grad_norm": 0.24777106940746307,
|
| 241 |
+
"kl": 0.0506591796875,
|
| 242 |
+
"learning_rate": 1.977755337933682e-06,
|
| 243 |
+
"loss": 0.0009,
|
| 244 |
+
"reward": 1.8709635853767395,
|
| 245 |
+
"reward_std": 0.3816621009260416,
|
| 246 |
+
"rewards/accuracy_reward": 0.05234375139698386,
|
| 247 |
+
"rewards/format_reward": 0.8700521066784859,
|
| 248 |
+
"rewards/tag_count_reward": 0.9485677301883697,
|
| 249 |
+
"step": 80
|
| 250 |
+
},
|
| 251 |
+
{
|
| 252 |
+
"clip_ratio": 0.0,
|
| 253 |
+
"completion_length": 485.4924621582031,
|
| 254 |
+
"epoch": 1.3679999999999999,
|
| 255 |
+
"grad_norm": 0.24417904019355774,
|
| 256 |
+
"kl": 0.057867431640625,
|
| 257 |
+
"learning_rate": 1.9697632383321754e-06,
|
| 258 |
+
"loss": 0.0134,
|
| 259 |
+
"reward": 1.9161458700895309,
|
| 260 |
+
"reward_std": 0.35734121650457384,
|
| 261 |
+
"rewards/accuracy_reward": 0.06484375179279596,
|
| 262 |
+
"rewards/format_reward": 0.8960937723517418,
|
| 263 |
+
"rewards/tag_count_reward": 0.9552083522081375,
|
| 264 |
+
"step": 85
|
| 265 |
+
},
|
| 266 |
+
{
|
| 267 |
+
"clip_ratio": 0.0,
|
| 268 |
+
"completion_length": 500.1916786193848,
|
| 269 |
+
"epoch": 1.448,
|
| 270 |
+
"grad_norm": 0.23515592515468597,
|
| 271 |
+
"kl": 0.0543365478515625,
|
| 272 |
+
"learning_rate": 1.960568345817306e-06,
|
| 273 |
+
"loss": 0.0188,
|
| 274 |
+
"reward": 1.9377604484558106,
|
| 275 |
+
"reward_std": 0.33337037824094296,
|
| 276 |
+
"rewards/accuracy_reward": 0.07031250232830644,
|
| 277 |
+
"rewards/format_reward": 0.9088541835546493,
|
| 278 |
+
"rewards/tag_count_reward": 0.9585937723517418,
|
| 279 |
+
"step": 90
|
| 280 |
+
},
|
| 281 |
+
{
|
| 282 |
+
"clip_ratio": 0.0,
|
| 283 |
+
"completion_length": 495.4093902587891,
|
| 284 |
+
"epoch": 1.528,
|
| 285 |
+
"grad_norm": 0.26944634318351746,
|
| 286 |
+
"kl": 0.048583984375,
|
| 287 |
+
"learning_rate": 1.9501820647722456e-06,
|
| 288 |
+
"loss": 0.0118,
|
| 289 |
+
"reward": 1.9946615099906921,
|
| 290 |
+
"reward_std": 0.29993436820805075,
|
| 291 |
+
"rewards/accuracy_reward": 0.09088541890960186,
|
| 292 |
+
"rewards/format_reward": 0.9322916865348816,
|
| 293 |
+
"rewards/tag_count_reward": 0.9714843943715096,
|
| 294 |
+
"step": 95
|
| 295 |
+
},
|
| 296 |
+
{
|
| 297 |
+
"epoch": 1.608,
|
| 298 |
+
"grad_norm": 0.2572639584541321,
|
| 299 |
+
"learning_rate": 1.938617277253916e-06,
|
| 300 |
+
"loss": 0.0144,
|
| 301 |
+
"step": 100
|
| 302 |
+
},
|
| 303 |
+
{
|
| 304 |
+
"epoch": 1.608,
|
| 305 |
+
"eval_clip_ratio": 0.0,
|
| 306 |
+
"eval_completion_length": 484.87662212665265,
|
| 307 |
+
"eval_kl": 0.057382436899038464,
|
| 308 |
+
"eval_loss": 0.015068341977894306,
|
| 309 |
+
"eval_reward": 1.952323785194984,
|
| 310 |
+
"eval_reward_std": 0.38110188108224136,
|
| 311 |
+
"eval_rewards/accuracy_reward": 0.10897436236532834,
|
| 312 |
+
"eval_rewards/format_reward": 0.89262822499642,
|
| 313 |
+
"eval_rewards/tag_count_reward": 0.9507211676010718,
|
| 314 |
+
"eval_runtime": 151.6564,
|
| 315 |
+
"eval_samples_per_second": 0.653,
|
| 316 |
+
"eval_steps_per_second": 0.02,
|
| 317 |
+
"step": 100
|
| 318 |
+
},
|
| 319 |
+
{
|
| 320 |
+
"clip_ratio": 0.0,
|
| 321 |
+
"completion_length": 506.50118865966795,
|
| 322 |
+
"epoch": 1.688,
|
| 323 |
+
"grad_norm": 0.2550075650215149,
|
| 324 |
+
"kl": 0.05778656005859375,
|
| 325 |
+
"learning_rate": 1.925888327015434e-06,
|
| 326 |
+
"loss": 0.0138,
|
| 327 |
+
"reward": 2.005696675181389,
|
| 328 |
+
"reward_std": 0.35690731797367337,
|
| 329 |
+
"rewards/accuracy_reward": 0.1282552123069763,
|
| 330 |
+
"rewards/format_reward": 0.9125000208616256,
|
| 331 |
+
"rewards/tag_count_reward": 0.9649414233863354,
|
| 332 |
+
"step": 105
|
| 333 |
+
},
|
| 334 |
+
{
|
| 335 |
+
"clip_ratio": 0.0,
|
| 336 |
+
"completion_length": 532.497412109375,
|
| 337 |
+
"epoch": 1.768,
|
| 338 |
+
"grad_norm": 0.22832101583480835,
|
| 339 |
+
"kl": 0.106121826171875,
|
| 340 |
+
"learning_rate": 1.9120110017156167e-06,
|
| 341 |
+
"loss": 0.0192,
|
| 342 |
+
"reward": 1.9944662123918533,
|
| 343 |
+
"reward_std": 0.3744897425174713,
|
| 344 |
+
"rewards/accuracy_reward": 0.13489583702757954,
|
| 345 |
+
"rewards/format_reward": 0.9020833536982537,
|
| 346 |
+
"rewards/tag_count_reward": 0.9574869960546494,
|
| 347 |
+
"step": 110
|
| 348 |
+
},
|
| 349 |
+
{
|
| 350 |
+
"clip_ratio": 0.0,
|
| 351 |
+
"completion_length": 505.0661560058594,
|
| 352 |
+
"epoch": 1.8479999999999999,
|
| 353 |
+
"grad_norm": 0.23704102635383606,
|
| 354 |
+
"kl": 0.055731201171875,
|
| 355 |
+
"learning_rate": 1.8970025133376251e-06,
|
| 356 |
+
"loss": 0.0128,
|
| 357 |
+
"reward": 2.009244841337204,
|
| 358 |
+
"reward_std": 0.33148185834288596,
|
| 359 |
+
"rewards/accuracy_reward": 0.12421875419095159,
|
| 360 |
+
"rewards/format_reward": 0.9169271022081376,
|
| 361 |
+
"rewards/tag_count_reward": 0.968098983168602,
|
| 362 |
+
"step": 115
|
| 363 |
+
},
|
| 364 |
+
{
|
| 365 |
+
"clip_ratio": 0.0,
|
| 366 |
+
"completion_length": 519.771109008789,
|
| 367 |
+
"epoch": 1.928,
|
| 368 |
+
"grad_norm": 0.2956863045692444,
|
| 369 |
+
"kl": 0.060504150390625,
|
| 370 |
+
"learning_rate": 1.8808814768410154e-06,
|
| 371 |
+
"loss": 0.0325,
|
| 372 |
+
"reward": 2.0304688274860383,
|
| 373 |
+
"reward_std": 0.3420341990888119,
|
| 374 |
+
"rewards/accuracy_reward": 0.13880208730697632,
|
| 375 |
+
"rewards/format_reward": 0.9231771007180214,
|
| 376 |
+
"rewards/tag_count_reward": 0.9684896007180214,
|
| 377 |
+
"step": 120
|
| 378 |
+
},
|
| 379 |
+
{
|
| 380 |
+
"clip_ratio": 0.0,
|
| 381 |
+
"completion_length": 512.1099105834961,
|
| 382 |
+
"epoch": 2.016,
|
| 383 |
+
"grad_norm": 0.22449135780334473,
|
| 384 |
+
"kl": 0.06649169921875,
|
| 385 |
+
"learning_rate": 1.8636678870736926e-06,
|
| 386 |
+
"loss": 0.0198,
|
| 387 |
+
"reward": 2.039778712391853,
|
| 388 |
+
"reward_std": 0.3516721487045288,
|
| 389 |
+
"rewards/accuracy_reward": 0.15156250335276128,
|
| 390 |
+
"rewards/format_reward": 0.9226562693715096,
|
| 391 |
+
"rewards/tag_count_reward": 0.9655599161982537,
|
| 392 |
+
"step": 125
|
| 393 |
+
},
|
| 394 |
+
{
|
| 395 |
+
"clip_ratio": 0.0,
|
| 396 |
+
"completion_length": 530.5299659729004,
|
| 397 |
+
"epoch": 2.096,
|
| 398 |
+
"grad_norm": 0.22710081934928894,
|
| 399 |
+
"kl": 0.0750244140625,
|
| 400 |
+
"learning_rate": 1.8453830939723912e-06,
|
| 401 |
+
"loss": 0.0315,
|
| 402 |
+
"reward": 2.042122468352318,
|
| 403 |
+
"reward_std": 0.37393513284623625,
|
| 404 |
+
"rewards/accuracy_reward": 0.16276042107492686,
|
| 405 |
+
"rewards/format_reward": 0.9140625193715095,
|
| 406 |
+
"rewards/tag_count_reward": 0.9652994975447655,
|
| 407 |
+
"step": 130
|
| 408 |
+
},
|
| 409 |
+
{
|
| 410 |
+
"clip_ratio": 0.0,
|
| 411 |
+
"completion_length": 549.8502754211426,
|
| 412 |
+
"epoch": 2.176,
|
| 413 |
+
"grad_norm": 0.28876593708992004,
|
| 414 |
+
"kl": 0.09521484375,
|
| 415 |
+
"learning_rate": 1.8260497760824456e-06,
|
| 416 |
+
"loss": 0.0474,
|
| 417 |
+
"reward": 2.0036458909511565,
|
| 418 |
+
"reward_std": 0.3934386149048805,
|
| 419 |
+
"rewards/accuracy_reward": 0.15026042023673652,
|
| 420 |
+
"rewards/format_reward": 0.8989583522081375,
|
| 421 |
+
"rewards/tag_count_reward": 0.9544271066784858,
|
| 422 |
+
"step": 135
|
| 423 |
+
},
|
| 424 |
+
{
|
| 425 |
+
"clip_ratio": 0.0,
|
| 426 |
+
"completion_length": 612.0536613464355,
|
| 427 |
+
"epoch": 2.2560000000000002,
|
| 428 |
+
"grad_norm": 0.8413172960281372,
|
| 429 |
+
"kl": 0.14200439453125,
|
| 430 |
+
"learning_rate": 1.8056919124296956e-06,
|
| 431 |
+
"loss": 0.1363,
|
| 432 |
+
"reward": 1.8713542193174362,
|
| 433 |
+
"reward_std": 0.536376877874136,
|
| 434 |
+
"rewards/accuracy_reward": 0.15989583814516664,
|
| 435 |
+
"rewards/format_reward": 0.7997395992279053,
|
| 436 |
+
"rewards/tag_count_reward": 0.911718773841858,
|
| 437 |
+
"step": 140
|
| 438 |
+
},
|
| 439 |
+
{
|
| 440 |
+
"clip_ratio": 0.0,
|
| 441 |
+
"completion_length": 1079.364094543457,
|
| 442 |
+
"epoch": 2.336,
|
| 443 |
+
"grad_norm": 0.6317133903503418,
|
| 444 |
+
"kl": 0.334130859375,
|
| 445 |
+
"learning_rate": 1.784334752779408e-06,
|
| 446 |
+
"loss": 0.28,
|
| 447 |
+
"reward": 1.164257851243019,
|
| 448 |
+
"reward_std": 0.6237203784286975,
|
| 449 |
+
"rewards/accuracy_reward": 0.08359375211875886,
|
| 450 |
+
"rewards/format_reward": 0.42317709475755694,
|
| 451 |
+
"rewards/tag_count_reward": 0.6574869945645332,
|
| 452 |
+
"step": 145
|
| 453 |
+
},
|
| 454 |
+
{
|
| 455 |
+
"clip_ratio": 0.0,
|
| 456 |
+
"completion_length": 1432.9927520751953,
|
| 457 |
+
"epoch": 2.416,
|
| 458 |
+
"grad_norm": 2.980980634689331,
|
| 459 |
+
"kl": 0.5193359375,
|
| 460 |
+
"learning_rate": 1.7620047863191097e-06,
|
| 461 |
+
"loss": 0.3054,
|
| 462 |
+
"reward": 0.794856795668602,
|
| 463 |
+
"reward_std": 0.5762915194034577,
|
| 464 |
+
"rewards/accuracy_reward": 0.042187501187436284,
|
| 465 |
+
"rewards/format_reward": 0.2578125067055225,
|
| 466 |
+
"rewards/tag_count_reward": 0.49485678449273107,
|
| 467 |
+
"step": 150
|
| 468 |
+
},
|
| 469 |
+
{
|
| 470 |
+
"clip_ratio": 0.0,
|
| 471 |
+
"completion_length": 579.490380859375,
|
| 472 |
+
"epoch": 2.496,
|
| 473 |
+
"grad_norm": 1.2213436365127563,
|
| 474 |
+
"kl": 0.2888671875,
|
| 475 |
+
"learning_rate": 1.7387297088041692e-06,
|
| 476 |
+
"loss": 0.1924,
|
| 477 |
+
"reward": 1.7270182639360427,
|
| 478 |
+
"reward_std": 0.5595091491937637,
|
| 479 |
+
"rewards/accuracy_reward": 0.1138020868645981,
|
| 480 |
+
"rewards/format_reward": 0.7473958514630794,
|
| 481 |
+
"rewards/tag_count_reward": 0.8658203348517418,
|
| 482 |
+
"step": 155
|
| 483 |
+
},
|
| 484 |
+
{
|
| 485 |
+
"clip_ratio": 0.0,
|
| 486 |
+
"completion_length": 401.75886688232424,
|
| 487 |
+
"epoch": 2.576,
|
| 488 |
+
"grad_norm": 0.8756619095802307,
|
| 489 |
+
"kl": 0.21771240234375,
|
| 490 |
+
"learning_rate": 1.7145383882068778e-06,
|
| 491 |
+
"loss": 0.02,
|
| 492 |
+
"reward": 1.8417318046092988,
|
| 493 |
+
"reward_std": 0.4633196383714676,
|
| 494 |
+
"rewards/accuracy_reward": 0.08697916942182929,
|
| 495 |
+
"rewards/format_reward": 0.8268229380249977,
|
| 496 |
+
"rewards/tag_count_reward": 0.9279297098517418,
|
| 497 |
+
"step": 160
|
| 498 |
+
},
|
| 499 |
+
{
|
| 500 |
+
"clip_ratio": 0.0,
|
| 501 |
+
"completion_length": 424.29662628173827,
|
| 502 |
+
"epoch": 2.656,
|
| 503 |
+
"grad_norm": 0.4502270221710205,
|
| 504 |
+
"kl": 0.13651123046875,
|
| 505 |
+
"learning_rate": 1.6894608289116341e-06,
|
| 506 |
+
"loss": 0.0061,
|
| 507 |
+
"reward": 1.8492838889360428,
|
| 508 |
+
"reward_std": 0.44692824259400366,
|
| 509 |
+
"rewards/accuracy_reward": 0.08515625311993062,
|
| 510 |
+
"rewards/format_reward": 0.8325521051883698,
|
| 511 |
+
"rewards/tag_count_reward": 0.9315755411982536,
|
| 512 |
+
"step": 165
|
| 513 |
+
},
|
| 514 |
+
{
|
| 515 |
+
"clip_ratio": 0.0,
|
| 516 |
+
"completion_length": 438.4726676940918,
|
| 517 |
+
"epoch": 2.7359999999999998,
|
| 518 |
+
"grad_norm": 0.37484192848205566,
|
| 519 |
+
"kl": 0.14183349609375,
|
| 520 |
+
"learning_rate": 1.663528134500646e-06,
|
| 521 |
+
"loss": 0.0114,
|
| 522 |
+
"reward": 1.8793620347976685,
|
| 523 |
+
"reward_std": 0.43491987958550454,
|
| 524 |
+
"rewards/accuracy_reward": 0.08671875244472176,
|
| 525 |
+
"rewards/format_reward": 0.8562500238418579,
|
| 526 |
+
"rewards/tag_count_reward": 0.9363932520151138,
|
| 527 |
+
"step": 170
|
| 528 |
+
},
|
| 529 |
+
{
|
| 530 |
+
"clip_ratio": 0.0,
|
| 531 |
+
"completion_length": 443.8067802429199,
|
| 532 |
+
"epoch": 2.816,
|
| 533 |
+
"grad_norm": 2.34854793548584,
|
| 534 |
+
"kl": 0.1514892578125,
|
| 535 |
+
"learning_rate": 1.6367724691762965e-06,
|
| 536 |
+
"loss": 0.0216,
|
| 537 |
+
"reward": 1.8856771260499954,
|
| 538 |
+
"reward_std": 0.44159807488322256,
|
| 539 |
+
"rewards/accuracy_reward": 0.09062500288709999,
|
| 540 |
+
"rewards/format_reward": 0.8583333492279053,
|
| 541 |
+
"rewards/tag_count_reward": 0.9367187693715096,
|
| 542 |
+
"step": 175
|
| 543 |
+
},
|
| 544 |
+
{
|
| 545 |
+
"clip_ratio": 0.0,
|
| 546 |
+
"completion_length": 428.94193878173826,
|
| 547 |
+
"epoch": 2.896,
|
| 548 |
+
"grad_norm": 0.9070159792900085,
|
| 549 |
+
"kl": 0.166015625,
|
| 550 |
+
"learning_rate": 1.6092270178680326e-06,
|
| 551 |
+
"loss": 0.0118,
|
| 552 |
+
"reward": 1.823046910762787,
|
| 553 |
+
"reward_std": 0.49542137011885645,
|
| 554 |
+
"rewards/accuracy_reward": 0.08932291937526315,
|
| 555 |
+
"rewards/format_reward": 0.8166666850447655,
|
| 556 |
+
"rewards/tag_count_reward": 0.9170573070645333,
|
| 557 |
+
"step": 180
|
| 558 |
+
},
|
| 559 |
+
{
|
| 560 |
+
"clip_ratio": 0.0,
|
| 561 |
+
"completion_length": 422.6744903564453,
|
| 562 |
+
"epoch": 2.976,
|
| 563 |
+
"grad_norm": 0.5818607211112976,
|
| 564 |
+
"kl": 0.22315673828125,
|
| 565 |
+
"learning_rate": 1.5809259450732493e-06,
|
| 566 |
+
"loss": 0.0246,
|
| 567 |
+
"reward": 1.8421875447034837,
|
| 568 |
+
"reward_std": 0.48024289682507515,
|
| 569 |
+
"rewards/accuracy_reward": 0.08046875253785402,
|
| 570 |
+
"rewards/format_reward": 0.8382812723517418,
|
| 571 |
+
"rewards/tag_count_reward": 0.9234375193715095,
|
| 572 |
+
"step": 185
|
| 573 |
+
},
|
| 574 |
+
{
|
| 575 |
+
"clip_ratio": 0.0,
|
| 576 |
+
"completion_length": 408.40235137939453,
|
| 577 |
+
"epoch": 3.064,
|
| 578 |
+
"grad_norm": 1.1170427799224854,
|
| 579 |
+
"kl": 0.3676025390625,
|
| 580 |
+
"learning_rate": 1.5519043524832167e-06,
|
| 581 |
+
"loss": 0.0297,
|
| 582 |
+
"reward": 1.8282552510499954,
|
| 583 |
+
"reward_std": 0.49146031588315964,
|
| 584 |
+
"rewards/accuracy_reward": 0.0791666692122817,
|
| 585 |
+
"rewards/format_reward": 0.8320312708616256,
|
| 586 |
+
"rewards/tag_count_reward": 0.9170573130249977,
|
| 587 |
+
"step": 190
|
| 588 |
+
},
|
| 589 |
+
{
|
| 590 |
+
"clip_ratio": 0.0,
|
| 591 |
+
"completion_length": 394.2843864440918,
|
| 592 |
+
"epoch": 3.144,
|
| 593 |
+
"grad_norm": 2.2475008964538574,
|
| 594 |
+
"kl": 0.3608642578125,
|
| 595 |
+
"learning_rate": 1.522198235446617e-06,
|
| 596 |
+
"loss": 0.0145,
|
| 597 |
+
"reward": 1.911328172683716,
|
| 598 |
+
"reward_std": 0.43749314919114113,
|
| 599 |
+
"rewards/accuracy_reward": 0.09192708560731262,
|
| 600 |
+
"rewards/format_reward": 0.8776041835546493,
|
| 601 |
+
"rewards/tag_count_reward": 0.9417968943715096,
|
| 602 |
+
"step": 195
|
| 603 |
+
},
|
| 604 |
+
{
|
| 605 |
+
"epoch": 3.224,
|
| 606 |
+
"grad_norm": 0.7201040387153625,
|
| 607 |
+
"learning_rate": 1.4918444383246736e-06,
|
| 608 |
+
"loss": 0.0284,
|
| 609 |
+
"step": 200
|
| 610 |
+
},
|
| 611 |
+
{
|
| 612 |
+
"epoch": 3.224,
|
| 613 |
+
"eval_clip_ratio": 0.0,
|
| 614 |
+
"eval_completion_length": 421.5977736253005,
|
| 615 |
+
"eval_kl": 0.96875,
|
| 616 |
+
"eval_loss": 0.050301969051361084,
|
| 617 |
+
"eval_reward": 1.9030449115313017,
|
| 618 |
+
"eval_reward_std": 0.45874117200191206,
|
| 619 |
+
"eval_rewards/accuracy_reward": 0.11217949052269642,
|
| 620 |
+
"eval_rewards/format_reward": 0.8589743742576013,
|
| 621 |
+
"eval_rewards/tag_count_reward": 0.9318910470375648,
|
| 622 |
+
"eval_runtime": 143.8236,
|
| 623 |
+
"eval_samples_per_second": 0.688,
|
| 624 |
+
"eval_steps_per_second": 0.021,
|
| 625 |
+
"step": 200
|
| 626 |
+
},
|
| 627 |
+
{
|
| 628 |
+
"clip_ratio": 0.0,
|
| 629 |
+
"completion_length": 436.63086853027346,
|
| 630 |
+
"epoch": 3.304,
|
| 631 |
+
"grad_norm": 3.313720941543579,
|
| 632 |
+
"kl": 0.51187744140625,
|
| 633 |
+
"learning_rate": 1.4608806087932616e-06,
|
| 634 |
+
"loss": 0.0457,
|
| 635 |
+
"reward": 1.9524414613842964,
|
| 636 |
+
"reward_std": 0.4343319511041045,
|
| 637 |
+
"rewards/accuracy_reward": 0.13098958672489971,
|
| 638 |
+
"rewards/format_reward": 0.8803385585546494,
|
| 639 |
+
"rewards/tag_count_reward": 0.9411133043467999,
|
| 640 |
+
"step": 205
|
| 641 |
+
},
|
| 642 |
+
{
|
| 643 |
+
"clip_ratio": 0.0,
|
| 644 |
+
"completion_length": 436.59454345703125,
|
| 645 |
+
"epoch": 3.384,
|
| 646 |
+
"grad_norm": 3.354438066482544,
|
| 647 |
+
"kl": 0.853369140625,
|
| 648 |
+
"learning_rate": 1.4293451511486655e-06,
|
| 649 |
+
"loss": 0.0496,
|
| 650 |
+
"reward": 1.9438151478767396,
|
| 651 |
+
"reward_std": 0.456201434135437,
|
| 652 |
+
"rewards/accuracy_reward": 0.13255208693444728,
|
| 653 |
+
"rewards/format_reward": 0.8734375208616256,
|
| 654 |
+
"rewards/tag_count_reward": 0.9378255411982537,
|
| 655 |
+
"step": 210
|
| 656 |
+
},
|
| 657 |
+
{
|
| 658 |
+
"clip_ratio": 0.0,
|
| 659 |
+
"completion_length": 438.16745910644534,
|
| 660 |
+
"epoch": 3.464,
|
| 661 |
+
"grad_norm": 1.7152291536331177,
|
| 662 |
+
"kl": 0.615625,
|
| 663 |
+
"learning_rate": 1.3972771786749071e-06,
|
| 664 |
+
"loss": 0.0382,
|
| 665 |
+
"reward": 1.9890625536441804,
|
| 666 |
+
"reward_std": 0.39320042356848717,
|
| 667 |
+
"rewards/accuracy_reward": 0.13333333656191826,
|
| 668 |
+
"rewards/format_reward": 0.9054687723517418,
|
| 669 |
+
"rewards/tag_count_reward": 0.9502604335546494,
|
| 670 |
+
"step": 215
|
| 671 |
+
},
|
| 672 |
+
{
|
| 673 |
+
"clip_ratio": 0.0,
|
| 674 |
+
"completion_length": 438.5679801940918,
|
| 675 |
+
"epoch": 3.544,
|
| 676 |
+
"grad_norm": 3.926846504211426,
|
| 677 |
+
"kl": 1.026171875,
|
| 678 |
+
"learning_rate": 1.3647164651317176e-06,
|
| 679 |
+
"loss": 0.0547,
|
| 680 |
+
"reward": 1.9758463978767395,
|
| 681 |
+
"reward_std": 0.4062122445553541,
|
| 682 |
+
"rewards/accuracy_reward": 0.12291666923556477,
|
| 683 |
+
"rewards/format_reward": 0.9041666865348816,
|
| 684 |
+
"rewards/tag_count_reward": 0.9487630411982536,
|
| 685 |
+
"step": 220
|
| 686 |
+
},
|
| 687 |
+
{
|
| 688 |
+
"clip_ratio": 0.0,
|
| 689 |
+
"completion_length": 444.0872512817383,
|
| 690 |
+
"epoch": 3.624,
|
| 691 |
+
"grad_norm": 3.1936042308807373,
|
| 692 |
+
"kl": 0.9034423828125,
|
| 693 |
+
"learning_rate": 1.3317033954233242e-06,
|
| 694 |
+
"loss": 0.048,
|
| 695 |
+
"reward": 1.9549479603767395,
|
| 696 |
+
"reward_std": 0.38959153592586515,
|
| 697 |
+
"rewards/accuracy_reward": 0.1049479195382446,
|
| 698 |
+
"rewards/format_reward": 0.9028646066784859,
|
| 699 |
+
"rewards/tag_count_reward": 0.9471354380249977,
|
| 700 |
+
"step": 225
|
| 701 |
+
},
|
| 702 |
+
{
|
| 703 |
+
"clip_ratio": 0.0,
|
| 704 |
+
"completion_length": 443.2856895446777,
|
| 705 |
+
"epoch": 3.7039999999999997,
|
| 706 |
+
"grad_norm": 3.45194149017334,
|
| 707 |
+
"kl": 3.086962890625,
|
| 708 |
+
"learning_rate": 1.2982789155092406e-06,
|
| 709 |
+
"loss": 0.1277,
|
| 710 |
+
"reward": 1.8746094167232514,
|
| 711 |
+
"reward_std": 0.5123490341007709,
|
| 712 |
+
"rewards/accuracy_reward": 0.10755208698101341,
|
| 713 |
+
"rewards/format_reward": 0.8559895992279053,
|
| 714 |
+
"rewards/tag_count_reward": 0.9110677286982536,
|
| 715 |
+
"step": 230
|
| 716 |
+
},
|
| 717 |
+
{
|
| 718 |
+
"clip_ratio": 0.0,
|
| 719 |
+
"completion_length": 459.6276145935059,
|
| 720 |
+
"epoch": 3.784,
|
| 721 |
+
"grad_norm": 2.9876277446746826,
|
| 722 |
+
"kl": 1.5169921875,
|
| 723 |
+
"learning_rate": 1.264484481619177e-06,
|
| 724 |
+
"loss": 0.0562,
|
| 725 |
+
"reward": 1.8753906697034837,
|
| 726 |
+
"reward_std": 0.47588179595768454,
|
| 727 |
+
"rewards/accuracy_reward": 0.08463541921228171,
|
| 728 |
+
"rewards/format_reward": 0.8697916865348816,
|
| 729 |
+
"rewards/tag_count_reward": 0.9209635615348816,
|
| 730 |
+
"step": 235
|
| 731 |
+
},
|
| 732 |
+
{
|
| 733 |
+
"clip_ratio": 0.0,
|
| 734 |
+
"completion_length": 453.0531379699707,
|
| 735 |
+
"epoch": 3.864,
|
| 736 |
+
"grad_norm": 2.683588981628418,
|
| 737 |
+
"kl": 1.91220703125,
|
| 738 |
+
"learning_rate": 1.2303620088350698e-06,
|
| 739 |
+
"loss": 0.0736,
|
| 740 |
+
"reward": 1.8893880665302276,
|
| 741 |
+
"reward_std": 0.4542847402393818,
|
| 742 |
+
"rewards/accuracy_reward": 0.09505208649206906,
|
| 743 |
+
"rewards/format_reward": 0.8731771022081375,
|
| 744 |
+
"rewards/tag_count_reward": 0.9211588725447655,
|
| 745 |
+
"step": 240
|
| 746 |
+
},
|
| 747 |
+
{
|
| 748 |
+
"clip_ratio": 0.0,
|
| 749 |
+
"completion_length": 476.33881454467775,
|
| 750 |
+
"epoch": 3.944,
|
| 751 |
+
"grad_norm": 1.4992015361785889,
|
| 752 |
+
"kl": 1.704296875,
|
| 753 |
+
"learning_rate": 1.1959538191039984e-06,
|
| 754 |
+
"loss": 0.0533,
|
| 755 |
+
"reward": 1.9165365010499955,
|
| 756 |
+
"reward_std": 0.44639485478401186,
|
| 757 |
+
"rewards/accuracy_reward": 0.10234375309664756,
|
| 758 |
+
"rewards/format_reward": 0.88619794100523,
|
| 759 |
+
"rewards/tag_count_reward": 0.9279948130249978,
|
| 760 |
+
"step": 245
|
| 761 |
+
},
|
| 762 |
+
{
|
| 763 |
+
"clip_ratio": 0.0,
|
| 764 |
+
"completion_length": 459.0494895935059,
|
| 765 |
+
"epoch": 4.032,
|
| 766 |
+
"grad_norm": 2.175201416015625,
|
| 767 |
+
"kl": 1.41162109375,
|
| 768 |
+
"learning_rate": 1.161302588746464e-06,
|
| 769 |
+
"loss": 0.0501,
|
| 770 |
+
"reward": 1.9514974415302277,
|
| 771 |
+
"reward_std": 0.38656867146492,
|
| 772 |
+
"rewards/accuracy_reward": 0.09531250256113707,
|
| 773 |
+
"rewards/format_reward": 0.9125000149011612,
|
| 774 |
+
"rewards/tag_count_reward": 0.9436849191784858,
|
| 775 |
+
"step": 250
|
| 776 |
+
},
|
| 777 |
+
{
|
| 778 |
+
"clip_ratio": 0.0,
|
| 779 |
+
"completion_length": 456.7244903564453,
|
| 780 |
+
"epoch": 4.112,
|
| 781 |
+
"grad_norm": 2.2169039249420166,
|
| 782 |
+
"kl": 1.56513671875,
|
| 783 |
+
"learning_rate": 1.1264512955251477e-06,
|
| 784 |
+
"loss": 0.0673,
|
| 785 |
+
"reward": 1.9486979573965073,
|
| 786 |
+
"reward_std": 0.3977412488311529,
|
| 787 |
+
"rewards/accuracy_reward": 0.0968750029336661,
|
| 788 |
+
"rewards/format_reward": 0.9091146051883697,
|
| 789 |
+
"rewards/tag_count_reward": 0.9427083522081375,
|
| 790 |
+
"step": 255
|
| 791 |
+
},
|
| 792 |
+
{
|
| 793 |
+
"clip_ratio": 0.0,
|
| 794 |
+
"completion_length": 438.6416763305664,
|
| 795 |
+
"epoch": 4.192,
|
| 796 |
+
"grad_norm": 1.7040432691574097,
|
| 797 |
+
"kl": 1.164892578125,
|
| 798 |
+
"learning_rate": 1.0914431653397856e-06,
|
| 799 |
+
"loss": 0.0609,
|
| 800 |
+
"reward": 1.9953776597976685,
|
| 801 |
+
"reward_std": 0.3568487804383039,
|
| 802 |
+
"rewards/accuracy_reward": 0.10859375339932739,
|
| 803 |
+
"rewards/format_reward": 0.93151044100523,
|
| 804 |
+
"rewards/tag_count_reward": 0.9552734643220901,
|
| 805 |
+
"step": 260
|
| 806 |
+
},
|
| 807 |
+
{
|
| 808 |
+
"clip_ratio": 0.0,
|
| 809 |
+
"completion_length": 421.86694107055666,
|
| 810 |
+
"epoch": 4.272,
|
| 811 |
+
"grad_norm": 1.1802079677581787,
|
| 812 |
+
"kl": 1.281884765625,
|
| 813 |
+
"learning_rate": 1.0563216186142838e-06,
|
| 814 |
+
"loss": 0.0606,
|
| 815 |
+
"reward": 1.9884766191244125,
|
| 816 |
+
"reward_std": 0.38088383749127386,
|
| 817 |
+
"rewards/accuracy_reward": 0.11223958663176745,
|
| 818 |
+
"rewards/format_reward": 0.9242187678813935,
|
| 819 |
+
"rewards/tag_count_reward": 0.9520182505249977,
|
| 820 |
+
"step": 265
|
| 821 |
+
},
|
| 822 |
+
{
|
| 823 |
+
"clip_ratio": 0.0,
|
| 824 |
+
"completion_length": 450.229695892334,
|
| 825 |
+
"epoch": 4.352,
|
| 826 |
+
"grad_norm": 0.8549647927284241,
|
| 827 |
+
"kl": 1.15458984375,
|
| 828 |
+
"learning_rate": 1.0211302164425654e-06,
|
| 829 |
+
"loss": 0.0333,
|
| 830 |
+
"reward": 2.0100261092185976,
|
| 831 |
+
"reward_std": 0.39018577747046945,
|
| 832 |
+
"rewards/accuracy_reward": 0.13671875335276126,
|
| 833 |
+
"rewards/format_reward": 0.9205729365348816,
|
| 834 |
+
"rewards/tag_count_reward": 0.952734398841858,
|
| 835 |
+
"step": 270
|
| 836 |
+
},
|
| 837 |
+
{
|
| 838 |
+
"clip_ratio": 0.0,
|
| 839 |
+
"completion_length": 473.3450653076172,
|
| 840 |
+
"epoch": 4.432,
|
| 841 |
+
"grad_norm": 1.360452651977539,
|
| 842 |
+
"kl": 0.996142578125,
|
| 843 |
+
"learning_rate": 9.859126065599434e-07,
|
| 844 |
+
"loss": 0.0378,
|
| 845 |
+
"reward": 1.9733073562383652,
|
| 846 |
+
"reward_std": 0.361860204860568,
|
| 847 |
+
"rewards/accuracy_reward": 0.10338541967794299,
|
| 848 |
+
"rewards/format_reward": 0.9190104350447654,
|
| 849 |
+
"rewards/tag_count_reward": 0.9509114801883698,
|
| 850 |
+
"step": 275
|
| 851 |
+
},
|
| 852 |
+
{
|
| 853 |
+
"clip_ratio": 0.0,
|
| 854 |
+
"completion_length": 456.2471473693848,
|
| 855 |
+
"epoch": 4.5120000000000005,
|
| 856 |
+
"grad_norm": 1.2597142457962036,
|
| 857 |
+
"kl": 1.180615234375,
|
| 858 |
+
"learning_rate": 9.507124692070355e-07,
|
| 859 |
+
"loss": 0.0429,
|
| 860 |
+
"reward": 2.0083985060453413,
|
| 861 |
+
"reward_std": 0.4000336788594723,
|
| 862 |
+
"rewards/accuracy_reward": 0.13541667046956718,
|
| 863 |
+
"rewards/format_reward": 0.9218750208616256,
|
| 864 |
+
"rewards/tag_count_reward": 0.9511067897081376,
|
| 865 |
+
"step": 280
|
| 866 |
+
},
|
| 867 |
+
{
|
| 868 |
+
"clip_ratio": 0.0,
|
| 869 |
+
"completion_length": 465.9554801940918,
|
| 870 |
+
"epoch": 4.592,
|
| 871 |
+
"grad_norm": 1.3279640674591064,
|
| 872 |
+
"kl": 0.97479248046875,
|
| 873 |
+
"learning_rate": 9.15573462953361e-07,
|
| 874 |
+
"loss": 0.0412,
|
| 875 |
+
"reward": 2.039453184604645,
|
| 876 |
+
"reward_std": 0.3459701970219612,
|
| 877 |
+
"rewards/accuracy_reward": 0.13671875298023223,
|
| 878 |
+
"rewards/format_reward": 0.9406250178813934,
|
| 879 |
+
"rewards/tag_count_reward": 0.9621093913912773,
|
| 880 |
+
"step": 285
|
| 881 |
+
},
|
| 882 |
+
{
|
| 883 |
+
"clip_ratio": 0.0,
|
| 884 |
+
"completion_length": 487.58568954467773,
|
| 885 |
+
"epoch": 4.672,
|
| 886 |
+
"grad_norm": 1.1848392486572266,
|
| 887 |
+
"kl": 0.62801513671875,
|
| 888 |
+
"learning_rate": 8.805391705478147e-07,
|
| 889 |
+
"loss": 0.0341,
|
| 890 |
+
"reward": 2.0368490278720857,
|
| 891 |
+
"reward_std": 0.34789668396115303,
|
| 892 |
+
"rewards/accuracy_reward": 0.14088542181998492,
|
| 893 |
+
"rewards/format_reward": 0.93697919100523,
|
| 894 |
+
"rewards/tag_count_reward": 0.958984388411045,
|
| 895 |
+
"step": 290
|
| 896 |
+
},
|
| 897 |
+
{
|
| 898 |
+
"clip_ratio": 0.0,
|
| 899 |
+
"completion_length": 462.78100204467773,
|
| 900 |
+
"epoch": 4.752,
|
| 901 |
+
"grad_norm": 1.9356491565704346,
|
| 902 |
+
"kl": 0.9264404296875,
|
| 903 |
+
"learning_rate": 8.456530448631855e-07,
|
| 904 |
+
"loss": 0.0383,
|
| 905 |
+
"reward": 1.9699219435453414,
|
| 906 |
+
"reward_std": 0.4028118785470724,
|
| 907 |
+
"rewards/accuracy_reward": 0.12838542023673655,
|
| 908 |
+
"rewards/format_reward": 0.9046875193715096,
|
| 909 |
+
"rewards/tag_count_reward": 0.9368489772081375,
|
| 910 |
+
"step": 295
|
| 911 |
+
},
|
| 912 |
+
{
|
| 913 |
+
"epoch": 4.832,
|
| 914 |
+
"grad_norm": 1.8601847887039185,
|
| 915 |
+
"learning_rate": 8.109583550017549e-07,
|
| 916 |
+
"loss": 0.0298,
|
| 917 |
+
"step": 300
|
| 918 |
+
},
|
| 919 |
+
{
|
| 920 |
+
"epoch": 4.832,
|
| 921 |
+
"eval_clip_ratio": 0.0,
|
| 922 |
+
"eval_completion_length": 448.06251408503607,
|
| 923 |
+
"eval_kl": 0.5677208533653846,
|
| 924 |
+
"eval_loss": 0.0052544595673680305,
|
| 925 |
+
"eval_reward": 2.0404648138926578,
|
| 926 |
+
"eval_reward_std": 0.3293027098362262,
|
| 927 |
+
"eval_rewards/accuracy_reward": 0.1250000040118511,
|
| 928 |
+
"eval_rewards/format_reward": 0.9503205281037551,
|
| 929 |
+
"eval_rewards/tag_count_reward": 0.9651442399391761,
|
| 930 |
+
"eval_runtime": 140.3516,
|
| 931 |
+
"eval_samples_per_second": 0.705,
|
| 932 |
+
"eval_steps_per_second": 0.021,
|
| 933 |
+
"step": 300
|
| 934 |
+
},
|
| 935 |
+
{
|
| 936 |
+
"clip_ratio": 0.0,
|
| 937 |
+
"completion_length": 473.3700660705566,
|
| 938 |
+
"epoch": 4.912,
|
| 939 |
+
"grad_norm": 1.6324626207351685,
|
| 940 |
+
"kl": 0.8969482421875,
|
| 941 |
+
"learning_rate": 7.764981326288272e-07,
|
| 942 |
+
"loss": 0.0462,
|
| 943 |
+
"reward": 2.001009176671505,
|
| 944 |
+
"reward_std": 0.38462002836167813,
|
| 945 |
+
"rewards/accuracy_reward": 0.1316406281432137,
|
| 946 |
+
"rewards/format_reward": 0.9210937708616257,
|
| 947 |
+
"rewards/tag_count_reward": 0.9482747592031956,
|
| 948 |
+
"step": 305
|
| 949 |
+
},
|
| 950 |
+
{
|
| 951 |
+
"clip_ratio": 0.0,
|
| 952 |
+
"completion_length": 476.5562614440918,
|
| 953 |
+
"epoch": 4.992,
|
| 954 |
+
"grad_norm": 1.3346195220947266,
|
| 955 |
+
"kl": 0.7723388671875,
|
| 956 |
+
"learning_rate": 7.423151186007526e-07,
|
| 957 |
+
"loss": 0.0428,
|
| 958 |
+
"reward": 2.023112043738365,
|
| 959 |
+
"reward_std": 0.3479595385491848,
|
| 960 |
+
"rewards/accuracy_reward": 0.12942708758637309,
|
| 961 |
+
"rewards/format_reward": 0.9343750178813934,
|
| 962 |
+
"rewards/tag_count_reward": 0.9593099161982537,
|
| 963 |
+
"step": 310
|
| 964 |
+
},
|
| 965 |
+
{
|
| 966 |
+
"clip_ratio": 0.0,
|
| 967 |
+
"completion_length": 474.70105209350584,
|
| 968 |
+
"epoch": 5.08,
|
| 969 |
+
"grad_norm": 3.4773247241973877,
|
| 970 |
+
"kl": 1.13505859375,
|
| 971 |
+
"learning_rate": 7.084517099536377e-07,
|
| 972 |
+
"loss": 0.0536,
|
| 973 |
+
"reward": 2.0288412243127825,
|
| 974 |
+
"reward_std": 0.37681432962417605,
|
| 975 |
+
"rewards/accuracy_reward": 0.1414062537252903,
|
| 976 |
+
"rewards/format_reward": 0.9312500193715095,
|
| 977 |
+
"rewards/tag_count_reward": 0.9561849161982536,
|
| 978 |
+
"step": 315
|
| 979 |
+
},
|
| 980 |
+
{
|
| 981 |
+
"clip_ratio": 0.0,
|
| 982 |
+
"completion_length": 449.5380317687988,
|
| 983 |
+
"epoch": 5.16,
|
| 984 |
+
"grad_norm": 3.157280683517456,
|
| 985 |
+
"kl": 0.80751953125,
|
| 986 |
+
"learning_rate": 6.749499073184957e-07,
|
| 987 |
+
"loss": 0.0222,
|
| 988 |
+
"reward": 2.015625074505806,
|
| 989 |
+
"reward_std": 0.36971147619187833,
|
| 990 |
+
"rewards/accuracy_reward": 0.13515625351574273,
|
| 991 |
+
"rewards/format_reward": 0.9281250193715096,
|
| 992 |
+
"rewards/tag_count_reward": 0.9523437649011612,
|
| 993 |
+
"step": 320
|
| 994 |
+
},
|
| 995 |
+
{
|
| 996 |
+
"clip_ratio": 0.0,
|
| 997 |
+
"completion_length": 453.49766845703124,
|
| 998 |
+
"epoch": 5.24,
|
| 999 |
+
"grad_norm": 2.6990413665771484,
|
| 1000 |
+
"kl": 1.14775390625,
|
| 1001 |
+
"learning_rate": 6.418512628280544e-07,
|
| 1002 |
+
"loss": 0.0459,
|
| 1003 |
+
"reward": 2.000455787777901,
|
| 1004 |
+
"reward_std": 0.3810987573117018,
|
| 1005 |
+
"rewards/accuracy_reward": 0.1320312530733645,
|
| 1006 |
+
"rewards/format_reward": 0.91979169100523,
|
| 1007 |
+
"rewards/tag_count_reward": 0.9486328348517418,
|
| 1008 |
+
"step": 325
|
| 1009 |
+
},
|
| 1010 |
+
{
|
| 1011 |
+
"clip_ratio": 0.0,
|
| 1012 |
+
"completion_length": 438.2487075805664,
|
| 1013 |
+
"epoch": 5.32,
|
| 1014 |
+
"grad_norm": 14.02304744720459,
|
| 1015 |
+
"kl": 1.029150390625,
|
| 1016 |
+
"learning_rate": 6.091968285798378e-07,
|
| 1017 |
+
"loss": 0.0235,
|
| 1018 |
+
"reward": 1.9692057698965073,
|
| 1019 |
+
"reward_std": 0.44279313534498216,
|
| 1020 |
+
"rewards/accuracy_reward": 0.13697916995733977,
|
| 1021 |
+
"rewards/format_reward": 0.9005208522081375,
|
| 1022 |
+
"rewards/tag_count_reward": 0.93170575350523,
|
| 1023 |
+
"step": 330
|
| 1024 |
+
},
|
| 1025 |
+
{
|
| 1026 |
+
"clip_ratio": 0.0,
|
| 1027 |
+
"completion_length": 442.1963684082031,
|
| 1028 |
+
"epoch": 5.4,
|
| 1029 |
+
"grad_norm": 2.4285733699798584,
|
| 1030 |
+
"kl": 0.85400390625,
|
| 1031 |
+
"learning_rate": 5.770271057194369e-07,
|
| 1032 |
+
"loss": 0.0206,
|
| 1033 |
+
"reward": 1.9694662004709245,
|
| 1034 |
+
"reward_std": 0.4486567251384258,
|
| 1035 |
+
"rewards/accuracy_reward": 0.14244791967794299,
|
| 1036 |
+
"rewards/format_reward": 0.8971354350447655,
|
| 1037 |
+
"rewards/tag_count_reward": 0.9298828288912773,
|
| 1038 |
+
"step": 335
|
| 1039 |
+
},
|
| 1040 |
+
{
|
| 1041 |
+
"clip_ratio": 0.0,
|
| 1042 |
+
"completion_length": 434.9536582946777,
|
| 1043 |
+
"epoch": 5.48,
|
| 1044 |
+
"grad_norm": 1.456663966178894,
|
| 1045 |
+
"kl": 1.02734375,
|
| 1046 |
+
"learning_rate": 5.453819942071211e-07,
|
| 1047 |
+
"loss": 0.0171,
|
| 1048 |
+
"reward": 1.986132875084877,
|
| 1049 |
+
"reward_std": 0.45432515181601046,
|
| 1050 |
+
"rewards/accuracy_reward": 0.15182292135432363,
|
| 1051 |
+
"rewards/format_reward": 0.9010416820645333,
|
| 1052 |
+
"rewards/tag_count_reward": 0.9332682505249977,
|
| 1053 |
+
"step": 340
|
| 1054 |
+
},
|
| 1055 |
+
{
|
| 1056 |
+
"clip_ratio": 0.0,
|
| 1057 |
+
"completion_length": 440.12631301879884,
|
| 1058 |
+
"epoch": 5.5600000000000005,
|
| 1059 |
+
"grad_norm": 6.417184352874756,
|
| 1060 |
+
"kl": 1.1162109375,
|
| 1061 |
+
"learning_rate": 5.143007433301034e-07,
|
| 1062 |
+
"loss": 0.0256,
|
| 1063 |
+
"reward": 1.994401106238365,
|
| 1064 |
+
"reward_std": 0.42431050203740595,
|
| 1065 |
+
"rewards/accuracy_reward": 0.1507812541909516,
|
| 1066 |
+
"rewards/format_reward": 0.9062500223517418,
|
| 1067 |
+
"rewards/tag_count_reward": 0.9373698100447655,
|
| 1068 |
+
"step": 345
|
| 1069 |
+
},
|
| 1070 |
+
{
|
| 1071 |
+
"clip_ratio": 0.0,
|
| 1072 |
+
"completion_length": 435.890633392334,
|
| 1073 |
+
"epoch": 5.64,
|
| 1074 |
+
"grad_norm": 1.4459128379821777,
|
| 1075 |
+
"kl": 1.022119140625,
|
| 1076 |
+
"learning_rate": 4.838219030218274e-07,
|
| 1077 |
+
"loss": 0.0165,
|
| 1078 |
+
"reward": 2.000455787777901,
|
| 1079 |
+
"reward_std": 0.441185887157917,
|
| 1080 |
+
"rewards/accuracy_reward": 0.15000000391155482,
|
| 1081 |
+
"rewards/format_reward": 0.9106771036982536,
|
| 1082 |
+
"rewards/tag_count_reward": 0.9397786661982537,
|
| 1083 |
+
"step": 350
|
| 1084 |
+
},
|
| 1085 |
+
{
|
| 1086 |
+
"clip_ratio": 0.0,
|
| 1087 |
+
"completion_length": 454.38621215820314,
|
| 1088 |
+
"epoch": 5.72,
|
| 1089 |
+
"grad_norm": 3.449855327606201,
|
| 1090 |
+
"kl": 2.3652099609375,
|
| 1091 |
+
"learning_rate": 4.5398327604866047e-07,
|
| 1092 |
+
"loss": 0.1717,
|
| 1093 |
+
"reward": 2.000716209411621,
|
| 1094 |
+
"reward_std": 0.4119129028171301,
|
| 1095 |
+
"rewards/accuracy_reward": 0.13723958674818276,
|
| 1096 |
+
"rewards/format_reward": 0.9182291850447655,
|
| 1097 |
+
"rewards/tag_count_reward": 0.9452474132180214,
|
| 1098 |
+
"step": 355
|
| 1099 |
+
},
|
| 1100 |
+
{
|
| 1101 |
+
"clip_ratio": 0.0,
|
| 1102 |
+
"completion_length": 465.8750114440918,
|
| 1103 |
+
"epoch": 5.8,
|
| 1104 |
+
"grad_norm": 2.101736545562744,
|
| 1105 |
+
"kl": 1.1593994140625,
|
| 1106 |
+
"learning_rate": 4.2482187112329515e-07,
|
| 1107 |
+
"loss": 0.0262,
|
| 1108 |
+
"reward": 2.005924531817436,
|
| 1109 |
+
"reward_std": 0.40926145724952223,
|
| 1110 |
+
"rewards/accuracy_reward": 0.14505208674818276,
|
| 1111 |
+
"rewards/format_reward": 0.9166666850447655,
|
| 1112 |
+
"rewards/tag_count_reward": 0.9442057460546494,
|
| 1113 |
+
"step": 360
|
| 1114 |
+
},
|
| 1115 |
+
{
|
| 1116 |
+
"clip_ratio": 0.0,
|
| 1117 |
+
"completion_length": 431.95730514526366,
|
| 1118 |
+
"epoch": 5.88,
|
| 1119 |
+
"grad_norm": 1.7132809162139893,
|
| 1120 |
+
"kl": 0.95352783203125,
|
| 1121 |
+
"learning_rate": 3.963738570030134e-07,
|
| 1122 |
+
"loss": 0.0317,
|
| 1123 |
+
"reward": 2.0475261181592943,
|
| 1124 |
+
"reward_std": 0.38608854487538335,
|
| 1125 |
+
"rewards/accuracy_reward": 0.1666666718199849,
|
| 1126 |
+
"rewards/format_reward": 0.9283854395151139,
|
| 1127 |
+
"rewards/tag_count_reward": 0.9524739846587181,
|
| 1128 |
+
"step": 365
|
| 1129 |
+
},
|
| 1130 |
+
{
|
| 1131 |
+
"clip_ratio": 0.0,
|
| 1132 |
+
"completion_length": 446.39740829467775,
|
| 1133 |
+
"epoch": 5.96,
|
| 1134 |
+
"grad_norm": 1.1850680112838745,
|
| 1135 |
+
"kl": 0.68291015625,
|
| 1136 |
+
"learning_rate": 3.686745176297411e-07,
|
| 1137 |
+
"loss": 0.0102,
|
| 1138 |
+
"reward": 2.046158942580223,
|
| 1139 |
+
"reward_std": 0.3435225561261177,
|
| 1140 |
+
"rewards/accuracy_reward": 0.14583333767950535,
|
| 1141 |
+
"rewards/format_reward": 0.9406250163912773,
|
| 1142 |
+
"rewards/tag_count_reward": 0.9597005382180214,
|
| 1143 |
+
"step": 370
|
| 1144 |
+
},
|
| 1145 |
+
{
|
| 1146 |
+
"clip_ratio": 0.0,
|
| 1147 |
+
"completion_length": 435.18698806762694,
|
| 1148 |
+
"epoch": 6.048,
|
| 1149 |
+
"grad_norm": 1.8223094940185547,
|
| 1150 |
+
"kl": 0.9990966796875,
|
| 1151 |
+
"learning_rate": 3.4175820836753646e-07,
|
| 1152 |
+
"loss": 0.0236,
|
| 1153 |
+
"reward": 2.0619141340255736,
|
| 1154 |
+
"reward_std": 0.34180368296802044,
|
| 1155 |
+
"rewards/accuracy_reward": 0.1572916701436043,
|
| 1156 |
+
"rewards/format_reward": 0.9429687693715095,
|
| 1157 |
+
"rewards/tag_count_reward": 0.9616536617279052,
|
| 1158 |
+
"step": 375
|
| 1159 |
+
},
|
| 1160 |
+
{
|
| 1161 |
+
"clip_ratio": 0.0,
|
| 1162 |
+
"completion_length": 453.92188491821287,
|
| 1163 |
+
"epoch": 6.128,
|
| 1164 |
+
"grad_norm": 1.4606250524520874,
|
| 1165 |
+
"kl": 1.0483154296875,
|
| 1166 |
+
"learning_rate": 3.156583133917884e-07,
|
| 1167 |
+
"loss": 0.0246,
|
| 1168 |
+
"reward": 2.0753255993127824,
|
| 1169 |
+
"reward_std": 0.3737676966935396,
|
| 1170 |
+
"rewards/accuracy_reward": 0.1804687550291419,
|
| 1171 |
+
"rewards/format_reward": 0.9361979395151139,
|
| 1172 |
+
"rewards/tag_count_reward": 0.9586588710546493,
|
| 1173 |
+
"step": 380
|
| 1174 |
+
},
|
| 1175 |
+
{
|
| 1176 |
+
"clip_ratio": 0.0,
|
| 1177 |
+
"completion_length": 443.9789176940918,
|
| 1178 |
+
"epoch": 6.208,
|
| 1179 |
+
"grad_norm": 1.4942998886108398,
|
| 1180 |
+
"kl": 0.8615966796875,
|
| 1181 |
+
"learning_rate": 2.904072042829775e-07,
|
| 1182 |
+
"loss": 0.0161,
|
| 1183 |
+
"reward": 2.043815162777901,
|
| 1184 |
+
"reward_std": 0.3786072336137295,
|
| 1185 |
+
"rewards/accuracy_reward": 0.15390625409781933,
|
| 1186 |
+
"rewards/format_reward": 0.9343750208616257,
|
| 1187 |
+
"rewards/tag_count_reward": 0.9555338725447655,
|
| 1188 |
+
"step": 385
|
| 1189 |
+
},
|
| 1190 |
+
{
|
| 1191 |
+
"clip_ratio": 0.0,
|
| 1192 |
+
"completion_length": 448.456778717041,
|
| 1193 |
+
"epoch": 6.288,
|
| 1194 |
+
"grad_norm": 3.1579020023345947,
|
| 1195 |
+
"kl": 0.98740234375,
|
| 1196 |
+
"learning_rate": 2.660361998763508e-07,
|
| 1197 |
+
"loss": 0.0237,
|
| 1198 |
+
"reward": 2.04114590883255,
|
| 1199 |
+
"reward_std": 0.3731150545179844,
|
| 1200 |
+
"rewards/accuracy_reward": 0.15442708749324083,
|
| 1201 |
+
"rewards/format_reward": 0.9322916850447655,
|
| 1202 |
+
"rewards/tag_count_reward": 0.9544271036982537,
|
| 1203 |
+
"step": 390
|
| 1204 |
+
},
|
| 1205 |
+
{
|
| 1206 |
+
"clip_ratio": 0.0,
|
| 1207 |
+
"completion_length": 449.2880340576172,
|
| 1208 |
+
"epoch": 6.368,
|
| 1209 |
+
"grad_norm": 1.9755905866622925,
|
| 1210 |
+
"kl": 0.92861328125,
|
| 1211 |
+
"learning_rate": 2.425755274173159e-07,
|
| 1212 |
+
"loss": 0.0252,
|
| 1213 |
+
"reward": 2.0747396677732466,
|
| 1214 |
+
"reward_std": 0.39273047521710397,
|
| 1215 |
+
"rewards/accuracy_reward": 0.18593750540167092,
|
| 1216 |
+
"rewards/format_reward": 0.9341146007180214,
|
| 1217 |
+
"rewards/tag_count_reward": 0.9546875223517418,
|
| 1218 |
+
"step": 395
|
| 1219 |
+
},
|
| 1220 |
+
{
|
| 1221 |
+
"epoch": 6.448,
|
| 1222 |
+
"grad_norm": 1.7979745864868164,
|
| 1223 |
+
"learning_rate": 2.2005428507072465e-07,
|
| 1224 |
+
"loss": 0.0208,
|
| 1225 |
+
"step": 400
|
| 1226 |
+
},
|
| 1227 |
+
{
|
| 1228 |
+
"epoch": 6.448,
|
| 1229 |
+
"eval_clip_ratio": 0.0,
|
| 1230 |
+
"eval_completion_length": 427.849369929387,
|
| 1231 |
+
"eval_kl": 0.7789963942307693,
|
| 1232 |
+
"eval_loss": 0.026709111407399178,
|
| 1233 |
+
"eval_reward": 2.0749199298711924,
|
| 1234 |
+
"eval_reward_std": 0.36127966069258177,
|
| 1235 |
+
"eval_rewards/accuracy_reward": 0.17147436250860876,
|
| 1236 |
+
"eval_rewards/format_reward": 0.9407051434883704,
|
| 1237 |
+
"eval_rewards/tag_count_reward": 0.9627403983703027,
|
| 1238 |
+
"eval_runtime": 137.2381,
|
| 1239 |
+
"eval_samples_per_second": 0.721,
|
| 1240 |
+
"eval_steps_per_second": 0.022,
|
| 1241 |
+
"step": 400
|
| 1242 |
+
},
|
| 1243 |
+
{
|
| 1244 |
+
"clip_ratio": 0.0,
|
| 1245 |
+
"completion_length": 450.2595161437988,
|
| 1246 |
+
"epoch": 6.5280000000000005,
|
| 1247 |
+
"grad_norm": 3.3182427883148193,
|
| 1248 |
+
"kl": 0.9768798828125,
|
| 1249 |
+
"learning_rate": 1.985004058305535e-07,
|
| 1250 |
+
"loss": 0.0313,
|
| 1251 |
+
"reward": 2.052799554169178,
|
| 1252 |
+
"reward_std": 0.38064534645527603,
|
| 1253 |
+
"rewards/accuracy_reward": 0.16210937928408384,
|
| 1254 |
+
"rewards/format_reward": 0.9345052309334279,
|
| 1255 |
+
"rewards/tag_count_reward": 0.9561849169433116,
|
| 1256 |
+
"step": 405
|
| 1257 |
+
},
|
| 1258 |
+
{
|
| 1259 |
+
"clip_ratio": 0.0,
|
| 1260 |
+
"completion_length": 448.1330825805664,
|
| 1261 |
+
"epoch": 6.608,
|
| 1262 |
+
"grad_norm": 3.9416732788085938,
|
| 1263 |
+
"kl": 1.0939453125,
|
| 1264 |
+
"learning_rate": 1.7794062287473733e-07,
|
| 1265 |
+
"loss": 0.0289,
|
| 1266 |
+
"reward": 2.0518229812383653,
|
| 1267 |
+
"reward_std": 0.35645177885890006,
|
| 1268 |
+
"rewards/accuracy_reward": 0.15468750335276127,
|
| 1269 |
+
"rewards/format_reward": 0.9377604350447655,
|
| 1270 |
+
"rewards/tag_count_reward": 0.9593750149011612,
|
| 1271 |
+
"step": 410
|
| 1272 |
+
},
|
| 1273 |
+
{
|
| 1274 |
+
"clip_ratio": 0.0,
|
| 1275 |
+
"completion_length": 436.55521926879885,
|
| 1276 |
+
"epoch": 6.688,
|
| 1277 |
+
"grad_norm": 2.4022459983825684,
|
| 1278 |
+
"kl": 1.0167724609375,
|
| 1279 |
+
"learning_rate": 1.5840043640813272e-07,
|
| 1280 |
+
"loss": 0.0213,
|
| 1281 |
+
"reward": 2.047461000084877,
|
| 1282 |
+
"reward_std": 0.3820289634168148,
|
| 1283 |
+
"rewards/accuracy_reward": 0.16432292014360428,
|
| 1284 |
+
"rewards/format_reward": 0.9296875223517418,
|
| 1285 |
+
"rewards/tag_count_reward": 0.9534505411982537,
|
| 1286 |
+
"step": 415
|
| 1287 |
+
},
|
| 1288 |
+
{
|
| 1289 |
+
"clip_ratio": 0.0,
|
| 1290 |
+
"completion_length": 449.48621063232423,
|
| 1291 |
+
"epoch": 6.768,
|
| 1292 |
+
"grad_norm": 2.1802432537078857,
|
| 1293 |
+
"kl": 0.945361328125,
|
| 1294 |
+
"learning_rate": 1.3990408203472937e-07,
|
| 1295 |
+
"loss": 0.0146,
|
| 1296 |
+
"reward": 2.051237052679062,
|
| 1297 |
+
"reward_std": 0.3935232628136873,
|
| 1298 |
+
"rewards/accuracy_reward": 0.16536458812188357,
|
| 1299 |
+
"rewards/format_reward": 0.932031263411045,
|
| 1300 |
+
"rewards/tag_count_reward": 0.9538411647081375,
|
| 1301 |
+
"step": 420
|
| 1302 |
+
},
|
| 1303 |
+
{
|
| 1304 |
+
"clip_ratio": 0.0,
|
| 1305 |
+
"completion_length": 448.03881378173827,
|
| 1306 |
+
"epoch": 6.848,
|
| 1307 |
+
"grad_norm": 1.2726682424545288,
|
| 1308 |
+
"kl": 1.0002685546875,
|
| 1309 |
+
"learning_rate": 1.2247450069834076e-07,
|
| 1310 |
+
"loss": 0.0156,
|
| 1311 |
+
"reward": 2.064778706431389,
|
| 1312 |
+
"reward_std": 0.3714164044708014,
|
| 1313 |
+
"rewards/accuracy_reward": 0.16979167144745588,
|
| 1314 |
+
"rewards/format_reward": 0.9377604365348816,
|
| 1315 |
+
"rewards/tag_count_reward": 0.9572265848517418,
|
| 1316 |
+
"step": 425
|
| 1317 |
+
},
|
| 1318 |
+
{
|
| 1319 |
+
"clip_ratio": 0.0,
|
| 1320 |
+
"completion_length": 447.07500915527345,
|
| 1321 |
+
"epoch": 6.928,
|
| 1322 |
+
"grad_norm": 2.3944952487945557,
|
| 1323 |
+
"kl": 0.728662109375,
|
| 1324 |
+
"learning_rate": 1.0613331022905758e-07,
|
| 1325 |
+
"loss": 0.0221,
|
| 1326 |
+
"reward": 2.0807943403720857,
|
| 1327 |
+
"reward_std": 0.35667028427124026,
|
| 1328 |
+
"rewards/accuracy_reward": 0.17161458870396018,
|
| 1329 |
+
"rewards/format_reward": 0.9460937738418579,
|
| 1330 |
+
"rewards/tag_count_reward": 0.9630859613418579,
|
| 1331 |
+
"step": 430
|
| 1332 |
+
},
|
| 1333 |
+
{
|
| 1334 |
+
"clip_ratio": 0.0,
|
| 1335 |
+
"completion_length": 1584.4791870117188,
|
| 1336 |
+
"epoch": 7.016,
|
| 1337 |
+
"grad_norm": 3.8696494102478027,
|
| 1338 |
+
"kl": 0.512451171875,
|
| 1339 |
+
"learning_rate": 9.090077853075118e-08,
|
| 1340 |
+
"loss": 0.3338,
|
| 1341 |
+
"reward": 0.6214192882180214,
|
| 1342 |
+
"reward_std": 0.8211026787757874,
|
| 1343 |
+
"rewards/accuracy_reward": 0.06380208488553762,
|
| 1344 |
+
"rewards/format_reward": 0.25130208767950535,
|
| 1345 |
+
"rewards/tag_count_reward": 0.3063151128590107,
|
| 1346 |
+
"step": 435
|
| 1347 |
+
},
|
| 1348 |
+
{
|
| 1349 |
+
"clip_ratio": 0.0,
|
| 1350 |
+
"completion_length": 1574.7263488769531,
|
| 1351 |
+
"epoch": 7.096,
|
| 1352 |
+
"grad_norm": 5.592410087585449,
|
| 1353 |
+
"kl": 0.470166015625,
|
| 1354 |
+
"learning_rate": 7.679579844288509e-08,
|
| 1355 |
+
"loss": 0.3414,
|
| 1356 |
+
"reward": 0.6222656443715096,
|
| 1357 |
+
"reward_std": 0.8386320501565934,
|
| 1358 |
+
"rewards/accuracy_reward": 0.06223958514165133,
|
| 1359 |
+
"rewards/format_reward": 0.2518229246139526,
|
| 1360 |
+
"rewards/tag_count_reward": 0.3082031346857548,
|
| 1361 |
+
"step": 440
|
| 1362 |
+
},
|
| 1363 |
+
{
|
| 1364 |
+
"clip_ratio": 0.0,
|
| 1365 |
+
"completion_length": 1548.7713928222656,
|
| 1366 |
+
"epoch": 7.176,
|
| 1367 |
+
"grad_norm": 6.609188079833984,
|
| 1368 |
+
"kl": 0.45009765625,
|
| 1369 |
+
"learning_rate": 6.383586430781196e-08,
|
| 1370 |
+
"loss": 0.3274,
|
| 1371 |
+
"reward": 0.6048828318715096,
|
| 1372 |
+
"reward_std": 0.8250703603029251,
|
| 1373 |
+
"rewards/accuracy_reward": 0.06588541865348815,
|
| 1374 |
+
"rewards/format_reward": 0.23958334065973758,
|
| 1375 |
+
"rewards/tag_count_reward": 0.2994140714406967,
|
| 1376 |
+
"step": 445
|
| 1377 |
+
},
|
| 1378 |
+
{
|
| 1379 |
+
"clip_ratio": 0.0,
|
| 1380 |
+
"completion_length": 1514.1794677734374,
|
| 1381 |
+
"epoch": 7.256,
|
| 1382 |
+
"grad_norm": 9.236469268798828,
|
| 1383 |
+
"kl": 0.4421875,
|
| 1384 |
+
"learning_rate": 5.203705027262184e-08,
|
| 1385 |
+
"loss": 0.3091,
|
| 1386 |
+
"reward": 0.6100911624729634,
|
| 1387 |
+
"reward_std": 0.7998277023434639,
|
| 1388 |
+
"rewards/accuracy_reward": 0.06171875186264515,
|
| 1389 |
+
"rewards/format_reward": 0.24322917498648167,
|
| 1390 |
+
"rewards/tag_count_reward": 0.3051432378590107,
|
| 1391 |
+
"step": 450
|
| 1392 |
+
},
|
| 1393 |
+
{
|
| 1394 |
+
"clip_ratio": 0.0,
|
| 1395 |
+
"completion_length": 1422.5672241210937,
|
| 1396 |
+
"epoch": 7.336,
|
| 1397 |
+
"grad_norm": 14.974526405334473,
|
| 1398 |
+
"kl": 0.447216796875,
|
| 1399 |
+
"learning_rate": 4.141399035245052e-08,
|
| 1400 |
+
"loss": 0.332,
|
| 1401 |
+
"reward": 0.6535156451165676,
|
| 1402 |
+
"reward_std": 0.8434477031230927,
|
| 1403 |
+
"rewards/accuracy_reward": 0.06458333530463278,
|
| 1404 |
+
"rewards/format_reward": 0.2614583432674408,
|
| 1405 |
+
"rewards/tag_count_reward": 0.3274739678949118,
|
| 1406 |
+
"step": 455
|
| 1407 |
+
},
|
| 1408 |
+
{
|
| 1409 |
+
"clip_ratio": 0.0,
|
| 1410 |
+
"completion_length": 1339.1979614257812,
|
| 1411 |
+
"epoch": 7.416,
|
| 1412 |
+
"grad_norm": 19.16950035095215,
|
| 1413 |
+
"kl": 0.4712890625,
|
| 1414 |
+
"learning_rate": 3.1979860279976564e-08,
|
| 1415 |
+
"loss": 0.3387,
|
| 1416 |
+
"reward": 0.710091170668602,
|
| 1417 |
+
"reward_std": 0.8780084311962127,
|
| 1418 |
+
"rewards/accuracy_reward": 0.07421875204890967,
|
| 1419 |
+
"rewards/format_reward": 0.28307292610406876,
|
| 1420 |
+
"rewards/tag_count_reward": 0.3527994900941849,
|
| 1421 |
+
"step": 460
|
| 1422 |
+
},
|
| 1423 |
+
{
|
| 1424 |
+
"clip_ratio": 0.0,
|
| 1425 |
+
"completion_length": 1225.218276977539,
|
| 1426 |
+
"epoch": 7.496,
|
| 1427 |
+
"grad_norm": 22.52800750732422,
|
| 1428 |
+
"kl": 0.523876953125,
|
| 1429 |
+
"learning_rate": 2.374636116362172e-08,
|
| 1430 |
+
"loss": 0.3504,
|
| 1431 |
+
"reward": 0.7427083536982536,
|
| 1432 |
+
"reward_std": 0.8980668410658836,
|
| 1433 |
+
"rewards/accuracy_reward": 0.07317708597984166,
|
| 1434 |
+
"rewards/format_reward": 0.30130209159106014,
|
| 1435 |
+
"rewards/tag_count_reward": 0.36822917237877845,
|
| 1436 |
+
"step": 465
|
| 1437 |
+
},
|
| 1438 |
+
{
|
| 1439 |
+
"clip_ratio": 0.0,
|
| 1440 |
+
"completion_length": 1114.6648742675782,
|
| 1441 |
+
"epoch": 7.576,
|
| 1442 |
+
"grad_norm": 25.05874252319336,
|
| 1443 |
+
"kl": 0.5748046875,
|
| 1444 |
+
"learning_rate": 1.6723704974718757e-08,
|
| 1445 |
+
"loss": 0.3561,
|
| 1446 |
+
"reward": 0.7968099161982536,
|
| 1447 |
+
"reward_std": 0.9075167685747146,
|
| 1448 |
+
"rewards/accuracy_reward": 0.08072916930541396,
|
| 1449 |
+
"rewards/format_reward": 0.3229166749864817,
|
| 1450 |
+
"rewards/tag_count_reward": 0.3931640740483999,
|
| 1451 |
+
"step": 470
|
| 1452 |
+
},
|
| 1453 |
+
{
|
| 1454 |
+
"clip_ratio": 0.0,
|
| 1455 |
+
"completion_length": 1007.0466384887695,
|
| 1456 |
+
"epoch": 7.656,
|
| 1457 |
+
"grad_norm": 24.392803192138672,
|
| 1458 |
+
"kl": 0.617919921875,
|
| 1459 |
+
"learning_rate": 1.0920601881650005e-08,
|
| 1460 |
+
"loss": 0.3497,
|
| 1461 |
+
"reward": 0.8455729380249977,
|
| 1462 |
+
"reward_std": 0.9318716049194335,
|
| 1463 |
+
"rewards/accuracy_reward": 0.08203125311993062,
|
| 1464 |
+
"rewards/format_reward": 0.3481770932674408,
|
| 1465 |
+
"rewards/tag_count_reward": 0.41536459475755694,
|
| 1466 |
+
"step": 475
|
| 1467 |
+
},
|
| 1468 |
+
{
|
| 1469 |
+
"clip_ratio": 0.0,
|
| 1470 |
+
"completion_length": 946.3583541870117,
|
| 1471 |
+
"epoch": 7.736,
|
| 1472 |
+
"grad_norm": 22.616287231445312,
|
| 1473 |
+
"kl": 0.68662109375,
|
| 1474 |
+
"learning_rate": 6.344249446665673e-09,
|
| 1475 |
+
"loss": 0.3505,
|
| 1476 |
+
"reward": 0.8694010645151138,
|
| 1477 |
+
"reward_std": 0.9415230333805085,
|
| 1478 |
+
"rewards/accuracy_reward": 0.0838541688863188,
|
| 1479 |
+
"rewards/format_reward": 0.3583333432674408,
|
| 1480 |
+
"rewards/tag_count_reward": 0.42721355259418486,
|
| 1481 |
+
"step": 480
|
| 1482 |
+
},
|
| 1483 |
+
{
|
| 1484 |
+
"clip_ratio": 0.0,
|
| 1485 |
+
"completion_length": 865.5632995605469,
|
| 1486 |
+
"epoch": 7.816,
|
| 1487 |
+
"grad_norm": 24.148319244384766,
|
| 1488 |
+
"kl": 0.79765625,
|
| 1489 |
+
"learning_rate": 3.0003236987802272e-09,
|
| 1490 |
+
"loss": 0.3502,
|
| 1491 |
+
"reward": 0.9250651299953461,
|
| 1492 |
+
"reward_std": 0.9858234718441963,
|
| 1493 |
+
"rewards/accuracy_reward": 0.09817708579357713,
|
| 1494 |
+
"rewards/format_reward": 0.3794270932674408,
|
| 1495 |
+
"rewards/tag_count_reward": 0.4474609471857548,
|
| 1496 |
+
"step": 485
|
| 1497 |
+
},
|
| 1498 |
+
{
|
| 1499 |
+
"clip_ratio": 0.0,
|
| 1500 |
+
"completion_length": 807.2963745117188,
|
| 1501 |
+
"epoch": 7.896,
|
| 1502 |
+
"grad_norm": 19.522138595581055,
|
| 1503 |
+
"kl": 0.9572265625,
|
| 1504 |
+
"learning_rate": 8.929720938193331e-10,
|
| 1505 |
+
"loss": 0.3453,
|
| 1506 |
+
"reward": 0.9965495049953461,
|
| 1507 |
+
"reward_std": 0.9948423892259598,
|
| 1508 |
+
"rewards/accuracy_reward": 0.10338541963137686,
|
| 1509 |
+
"rewards/format_reward": 0.41250001043081286,
|
| 1510 |
+
"rewards/tag_count_reward": 0.4806640759110451,
|
| 1511 |
+
"step": 490
|
| 1512 |
+
},
|
| 1513 |
+
{
|
| 1514 |
+
"clip_ratio": 0.0,
|
| 1515 |
+
"completion_length": 747.9367416381835,
|
| 1516 |
+
"epoch": 7.976,
|
| 1517 |
+
"grad_norm": 17.482969284057617,
|
| 1518 |
+
"kl": 1.2998046875,
|
| 1519 |
+
"learning_rate": 2.4808370349460595e-11,
|
| 1520 |
+
"loss": 0.3647,
|
| 1521 |
+
"reward": 1.0485026329755782,
|
| 1522 |
+
"reward_std": 1.0057428479194641,
|
| 1523 |
+
"rewards/accuracy_reward": 0.10468750277068466,
|
| 1524 |
+
"rewards/format_reward": 0.43854167610406875,
|
| 1525 |
+
"rewards/tag_count_reward": 0.5052734553813935,
|
| 1526 |
+
"step": 495
|
| 1527 |
+
},
|
| 1528 |
+
{
|
| 1529 |
+
"clip_ratio": 0.0,
|
| 1530 |
+
"completion_length": 788.9921875,
|
| 1531 |
+
"epoch": 7.992,
|
| 1532 |
+
"kl": 1.58984375,
|
| 1533 |
+
"reward": 1.0631510764360428,
|
| 1534 |
+
"reward_std": 0.9985880479216576,
|
| 1535 |
+
"rewards/accuracy_reward": 0.09635416883975267,
|
| 1536 |
+
"rewards/format_reward": 0.4479166753590107,
|
| 1537 |
+
"rewards/tag_count_reward": 0.5188802257180214,
|
| 1538 |
+
"step": 496,
|
| 1539 |
+
"total_flos": 0.0,
|
| 1540 |
+
"train_loss": 0.04277147677156233,
|
| 1541 |
+
"train_runtime": 16581.2814,
|
| 1542 |
+
"train_samples_per_second": 3.86,
|
| 1543 |
+
"train_steps_per_second": 0.03
|
| 1544 |
+
}
|
| 1545 |
+
],
|
| 1546 |
+
"logging_steps": 5,
|
| 1547 |
+
"max_steps": 496,
|
| 1548 |
+
"num_input_tokens_seen": 0,
|
| 1549 |
+
"num_train_epochs": 8,
|
| 1550 |
+
"save_steps": 500,
|
| 1551 |
+
"stateful_callbacks": {
|
| 1552 |
+
"TrainerControl": {
|
| 1553 |
+
"args": {
|
| 1554 |
+
"should_epoch_stop": false,
|
| 1555 |
+
"should_evaluate": false,
|
| 1556 |
+
"should_log": false,
|
| 1557 |
+
"should_save": true,
|
| 1558 |
+
"should_training_stop": true
|
| 1559 |
+
},
|
| 1560 |
+
"attributes": {}
|
| 1561 |
+
}
|
| 1562 |
+
},
|
| 1563 |
+
"total_flos": 0.0,
|
| 1564 |
+
"train_batch_size": 12,
|
| 1565 |
+
"trial_name": null,
|
| 1566 |
+
"trial_params": null
|
| 1567 |
+
}
|