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 +793 -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: DeepSeek-R1-Distill-Qwen-1.5B-GRPO
|
| 5 |
+
tags:
|
| 6 |
+
- generated_from_trainer
|
| 7 |
+
- trl
|
| 8 |
+
- grpo
|
| 9 |
+
licence: license
|
| 10 |
+
---
|
| 11 |
+
|
| 12 |
+
# Model Card for DeepSeek-R1-Distill-Qwen-1.5B-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="Jforeverss/DeepSeek-R1-Distill-Qwen-1.5B-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/yuzhao/huggingface/runs/a9v4dljj)
|
| 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: 0.17.0.dev0
|
| 38 |
+
- Transformers: 4.51.2
|
| 39 |
+
- Pytorch: 2.6.0
|
| 40 |
+
- Datasets: 3.5.0
|
| 41 |
+
- Tokenizers: 0.21.1
|
| 42 |
+
|
| 43 |
+
## Citations
|
| 44 |
+
|
| 45 |
+
Cite GRPO as:
|
| 46 |
+
|
| 47 |
+
```bibtex
|
| 48 |
+
@article{zhihong2024deepseekmath,
|
| 49 |
+
title = {{DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models}},
|
| 50 |
+
author = {Zhihong Shao and Peiyi Wang and Qihao Zhu and Runxin Xu and Junxiao Song and Mingchuan Zhang and Y. K. Li and Y. Wu and Daya Guo},
|
| 51 |
+
year = 2024,
|
| 52 |
+
eprint = {arXiv:2402.03300},
|
| 53 |
+
}
|
| 54 |
+
|
| 55 |
+
```
|
| 56 |
+
|
| 57 |
+
Cite TRL as:
|
| 58 |
+
|
| 59 |
+
```bibtex
|
| 60 |
+
@misc{vonwerra2022trl,
|
| 61 |
+
title = {{TRL: Transformer Reinforcement Learning}},
|
| 62 |
+
author = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang and Kashif Rasul and Quentin Gallou{\'e}dec},
|
| 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": 0.1677406519651413,
|
| 4 |
+
"train_runtime": 1369.1313,
|
| 5 |
+
"train_samples": 100,
|
| 6 |
+
"train_samples_per_second": 0.073,
|
| 7 |
+
"train_steps_per_second": 0.018
|
| 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": "4.51.2"
|
| 9 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"total_flos": 0.0,
|
| 3 |
+
"train_loss": 0.1677406519651413,
|
| 4 |
+
"train_runtime": 1369.1313,
|
| 5 |
+
"train_samples": 100,
|
| 6 |
+
"train_samples_per_second": 0.073,
|
| 7 |
+
"train_steps_per_second": 0.018
|
| 8 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,793 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_global_step": null,
|
| 3 |
+
"best_metric": null,
|
| 4 |
+
"best_model_checkpoint": null,
|
| 5 |
+
"epoch": 1.0,
|
| 6 |
+
"eval_steps": 500,
|
| 7 |
+
"global_step": 25,
|
| 8 |
+
"is_hyper_param_search": false,
|
| 9 |
+
"is_local_process_zero": true,
|
| 10 |
+
"is_world_process_zero": true,
|
| 11 |
+
"log_history": [
|
| 12 |
+
{
|
| 13 |
+
"clip_ratio/high_max": 0.0,
|
| 14 |
+
"clip_ratio/high_mean": 0.0,
|
| 15 |
+
"clip_ratio/low_mean": 0.0,
|
| 16 |
+
"clip_ratio/low_min": 0.0,
|
| 17 |
+
"clip_ratio/region_mean": 0.0,
|
| 18 |
+
"completions/clipped_ratio": 0.578125,
|
| 19 |
+
"completions/max_length": 2048.0,
|
| 20 |
+
"completions/max_terminated_length": 1791.0,
|
| 21 |
+
"completions/mean_length": 1737.125,
|
| 22 |
+
"completions/mean_terminated_length": 1440.3125305175781,
|
| 23 |
+
"completions/min_length": 970.75,
|
| 24 |
+
"completions/min_terminated_length": 970.75,
|
| 25 |
+
"epoch": 0.04,
|
| 26 |
+
"grad_norm": 0.2609442174434662,
|
| 27 |
+
"learning_rate": 0.0,
|
| 28 |
+
"length_penalty/max": 1.2887113094329834,
|
| 29 |
+
"length_penalty/mean": 0.030646182829514146,
|
| 30 |
+
"loss": 0.1256,
|
| 31 |
+
"num_tokens": 121000.0,
|
| 32 |
+
"reward": 0.47265625,
|
| 33 |
+
"reward_std": 0.2607314772903919,
|
| 34 |
+
"rewards/accuracy_reward/mean": 0.0625,
|
| 35 |
+
"rewards/accuracy_reward/std": 0.1632782220840454,
|
| 36 |
+
"rewards/format_reward/mean": 0.0,
|
| 37 |
+
"rewards/format_reward/std": 0.0,
|
| 38 |
+
"rewards/tag_count_reward/mean": 0.41015625,
|
| 39 |
+
"rewards/tag_count_reward/std": 0.13350078836083412,
|
| 40 |
+
"step": 1
|
| 41 |
+
},
|
| 42 |
+
{
|
| 43 |
+
"clip_ratio/high_max": 0.0,
|
| 44 |
+
"clip_ratio/high_mean": 0.0,
|
| 45 |
+
"clip_ratio/low_mean": 0.0,
|
| 46 |
+
"clip_ratio/low_min": 0.0,
|
| 47 |
+
"clip_ratio/region_mean": 0.0,
|
| 48 |
+
"completions/clipped_ratio": 0.484375,
|
| 49 |
+
"completions/max_length": 1792.5,
|
| 50 |
+
"completions/max_terminated_length": 1450.0,
|
| 51 |
+
"completions/mean_length": 1600.515625,
|
| 52 |
+
"completions/mean_terminated_length": 1249.0848388671875,
|
| 53 |
+
"completions/min_length": 1017.75,
|
| 54 |
+
"completions/min_terminated_length": 1017.75,
|
| 55 |
+
"epoch": 0.08,
|
| 56 |
+
"grad_norm": 0.2399621158838272,
|
| 57 |
+
"learning_rate": 3.333333333333333e-07,
|
| 58 |
+
"length_penalty/max": 1.7182817459106445,
|
| 59 |
+
"length_penalty/mean": 0.06100043409969658,
|
| 60 |
+
"loss": 0.1339,
|
| 61 |
+
"num_tokens": 233273.0,
|
| 62 |
+
"reward": 0.44140625,
|
| 63 |
+
"reward_std": 0.17154236882925034,
|
| 64 |
+
"rewards/accuracy_reward/mean": 0.046875,
|
| 65 |
+
"rewards/accuracy_reward/std": 0.10077822208404541,
|
| 66 |
+
"rewards/format_reward/mean": 0.0,
|
| 67 |
+
"rewards/format_reward/std": 0.0,
|
| 68 |
+
"rewards/tag_count_reward/mean": 0.39453125,
|
| 69 |
+
"rewards/tag_count_reward/std": 0.07449322193861008,
|
| 70 |
+
"step": 2
|
| 71 |
+
},
|
| 72 |
+
{
|
| 73 |
+
"clip_ratio/high_max": 0.0,
|
| 74 |
+
"clip_ratio/high_mean": 0.0,
|
| 75 |
+
"clip_ratio/low_mean": 0.0,
|
| 76 |
+
"clip_ratio/low_min": 0.0,
|
| 77 |
+
"clip_ratio/region_mean": 0.0,
|
| 78 |
+
"completions/clipped_ratio": 0.5625,
|
| 79 |
+
"completions/max_length": 1790.25,
|
| 80 |
+
"completions/max_terminated_length": 1688.75,
|
| 81 |
+
"completions/mean_length": 1620.703125,
|
| 82 |
+
"completions/mean_terminated_length": 1240.828125,
|
| 83 |
+
"completions/min_length": 713.5,
|
| 84 |
+
"completions/min_terminated_length": 713.5,
|
| 85 |
+
"epoch": 0.12,
|
| 86 |
+
"grad_norm": 0.2947452664375305,
|
| 87 |
+
"learning_rate": 6.666666666666666e-07,
|
| 88 |
+
"length_penalty/max": 1.1383908987045288,
|
| 89 |
+
"length_penalty/mean": 0.0120134013632196,
|
| 90 |
+
"loss": 0.137,
|
| 91 |
+
"num_tokens": 344086.0,
|
| 92 |
+
"reward": 0.421875,
|
| 93 |
+
"reward_std": 0.22617439925670624,
|
| 94 |
+
"rewards/accuracy_reward/mean": 0.015625,
|
| 95 |
+
"rewards/accuracy_reward/std": 0.0625,
|
| 96 |
+
"rewards/format_reward/mean": 0.0,
|
| 97 |
+
"rewards/format_reward/std": 0.0,
|
| 98 |
+
"rewards/tag_count_reward/mean": 0.40625,
|
| 99 |
+
"rewards/tag_count_reward/std": 0.17714006640017033,
|
| 100 |
+
"step": 3
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"clip_ratio/high_max": 0.0,
|
| 104 |
+
"clip_ratio/high_mean": 0.0,
|
| 105 |
+
"clip_ratio/low_mean": 0.0,
|
| 106 |
+
"clip_ratio/low_min": 0.0,
|
| 107 |
+
"clip_ratio/region_mean": 0.0,
|
| 108 |
+
"completions/clipped_ratio": 0.578125,
|
| 109 |
+
"completions/max_length": 2048.0,
|
| 110 |
+
"completions/max_terminated_length": 1288.0,
|
| 111 |
+
"completions/mean_length": 1807.453125,
|
| 112 |
+
"completions/mean_terminated_length": 1056.0083618164062,
|
| 113 |
+
"completions/min_length": 1228.25,
|
| 114 |
+
"completions/min_terminated_length": 716.25,
|
| 115 |
+
"epoch": 0.16,
|
| 116 |
+
"grad_norm": 0.19154201447963715,
|
| 117 |
+
"learning_rate": 1e-06,
|
| 118 |
+
"length_penalty/max": 0.8591408729553223,
|
| 119 |
+
"length_penalty/mean": 0.05627863667905331,
|
| 120 |
+
"loss": 0.1524,
|
| 121 |
+
"num_tokens": 467811.0,
|
| 122 |
+
"reward": 0.39453125,
|
| 123 |
+
"reward_std": 0.11663510650396347,
|
| 124 |
+
"rewards/accuracy_reward/mean": 0.0,
|
| 125 |
+
"rewards/accuracy_reward/std": 0.0,
|
| 126 |
+
"rewards/format_reward/mean": 0.0,
|
| 127 |
+
"rewards/format_reward/std": 0.0,
|
| 128 |
+
"rewards/tag_count_reward/mean": 0.39453125,
|
| 129 |
+
"rewards/tag_count_reward/std": 0.11663510836660862,
|
| 130 |
+
"step": 4
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"clip_ratio/high_max": 0.0,
|
| 134 |
+
"clip_ratio/high_mean": 0.0,
|
| 135 |
+
"clip_ratio/low_mean": 0.0,
|
| 136 |
+
"clip_ratio/low_min": 0.0,
|
| 137 |
+
"clip_ratio/region_mean": 0.0,
|
| 138 |
+
"completions/clipped_ratio": 0.515625,
|
| 139 |
+
"completions/max_length": 2048.0,
|
| 140 |
+
"completions/max_terminated_length": 1685.25,
|
| 141 |
+
"completions/mean_length": 1689.171875,
|
| 142 |
+
"completions/mean_terminated_length": 1288.3262329101562,
|
| 143 |
+
"completions/min_length": 909.0,
|
| 144 |
+
"completions/min_terminated_length": 909.0,
|
| 145 |
+
"epoch": 0.2,
|
| 146 |
+
"grad_norm": 0.2524236738681793,
|
| 147 |
+
"learning_rate": 9.954196488464196e-07,
|
| 148 |
+
"length_penalty/max": 1.2887113094329834,
|
| 149 |
+
"length_penalty/mean": 0.016135167446918786,
|
| 150 |
+
"loss": 0.171,
|
| 151 |
+
"num_tokens": 584430.0,
|
| 152 |
+
"reward": 0.515625,
|
| 153 |
+
"reward_std": 0.3025277405977249,
|
| 154 |
+
"rewards/accuracy_reward/mean": 0.140625,
|
| 155 |
+
"rewards/accuracy_reward/std": 0.2257782220840454,
|
| 156 |
+
"rewards/format_reward/mean": 0.0,
|
| 157 |
+
"rewards/format_reward/std": 0.0,
|
| 158 |
+
"rewards/tag_count_reward/mean": 0.375,
|
| 159 |
+
"rewards/tag_count_reward/std": 0.09814241342246532,
|
| 160 |
+
"step": 5
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"clip_ratio/high_max": 0.0,
|
| 164 |
+
"clip_ratio/high_mean": 0.0,
|
| 165 |
+
"clip_ratio/low_mean": 0.0,
|
| 166 |
+
"clip_ratio/low_min": 0.0,
|
| 167 |
+
"clip_ratio/region_mean": 0.0,
|
| 168 |
+
"completions/clipped_ratio": 0.265625,
|
| 169 |
+
"completions/max_length": 1830.75,
|
| 170 |
+
"completions/max_terminated_length": 1529.75,
|
| 171 |
+
"completions/mean_length": 1199.828125,
|
| 172 |
+
"completions/mean_terminated_length": 1011.0104217529297,
|
| 173 |
+
"completions/min_length": 390.0,
|
| 174 |
+
"completions/min_terminated_length": 390.0,
|
| 175 |
+
"epoch": 0.24,
|
| 176 |
+
"grad_norm": 0.40206748247146606,
|
| 177 |
+
"learning_rate": 9.817718381265238e-07,
|
| 178 |
+
"length_penalty/max": 1.7182817459106445,
|
| 179 |
+
"length_penalty/mean": 0.31160077918320894,
|
| 180 |
+
"loss": 0.4075,
|
| 181 |
+
"num_tokens": 670611.0,
|
| 182 |
+
"reward": 0.73046875,
|
| 183 |
+
"reward_std": 0.38131191954016685,
|
| 184 |
+
"rewards/accuracy_reward/mean": 0.234375,
|
| 185 |
+
"rewards/accuracy_reward/std": 0.3384781554341316,
|
| 186 |
+
"rewards/format_reward/mean": 0.0,
|
| 187 |
+
"rewards/format_reward/std": 0.0,
|
| 188 |
+
"rewards/tag_count_reward/mean": 0.49609375,
|
| 189 |
+
"rewards/tag_count_reward/std": 0.14783130772411823,
|
| 190 |
+
"step": 6
|
| 191 |
+
},
|
| 192 |
+
{
|
| 193 |
+
"clip_ratio/high_max": 0.0,
|
| 194 |
+
"clip_ratio/high_mean": 0.0,
|
| 195 |
+
"clip_ratio/low_mean": 0.0,
|
| 196 |
+
"clip_ratio/low_min": 0.0,
|
| 197 |
+
"clip_ratio/region_mean": 0.0,
|
| 198 |
+
"completions/clipped_ratio": 0.484375,
|
| 199 |
+
"completions/max_length": 2048.0,
|
| 200 |
+
"completions/max_terminated_length": 1228.5,
|
| 201 |
+
"completions/mean_length": 1517.125,
|
| 202 |
+
"completions/mean_terminated_length": 801.0875244140625,
|
| 203 |
+
"completions/min_length": 1055.75,
|
| 204 |
+
"completions/min_terminated_length": 543.75,
|
| 205 |
+
"epoch": 0.28,
|
| 206 |
+
"grad_norm": 0.28752532601356506,
|
| 207 |
+
"learning_rate": 9.593343979095332e-07,
|
| 208 |
+
"length_penalty/max": 1.2887113094329834,
|
| 209 |
+
"length_penalty/mean": 0.029227381804957986,
|
| 210 |
+
"loss": 0.1864,
|
| 211 |
+
"num_tokens": 775643.0,
|
| 212 |
+
"reward": 0.47265625,
|
| 213 |
+
"reward_std": 0.19874459877610207,
|
| 214 |
+
"rewards/accuracy_reward/mean": 0.078125,
|
| 215 |
+
"rewards/accuracy_reward/std": 0.11967839300632477,
|
| 216 |
+
"rewards/format_reward/mean": 0.0,
|
| 217 |
+
"rewards/format_reward/std": 0.0,
|
| 218 |
+
"rewards/tag_count_reward/mean": 0.39453125,
|
| 219 |
+
"rewards/tag_count_reward/std": 0.11168401688337326,
|
| 220 |
+
"step": 7
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"clip_ratio/high_max": 0.0,
|
| 224 |
+
"clip_ratio/high_mean": 0.0,
|
| 225 |
+
"clip_ratio/low_mean": 0.0,
|
| 226 |
+
"clip_ratio/low_min": 0.0,
|
| 227 |
+
"clip_ratio/region_mean": 0.0,
|
| 228 |
+
"completions/clipped_ratio": 0.375,
|
| 229 |
+
"completions/max_length": 1775.25,
|
| 230 |
+
"completions/max_terminated_length": 1460.75,
|
| 231 |
+
"completions/mean_length": 1364.109375,
|
| 232 |
+
"completions/mean_terminated_length": 1070.6136474609375,
|
| 233 |
+
"completions/min_length": 745.25,
|
| 234 |
+
"completions/min_terminated_length": 745.25,
|
| 235 |
+
"epoch": 0.32,
|
| 236 |
+
"grad_norm": 0.40461164712905884,
|
| 237 |
+
"learning_rate": 9.285640897740315e-07,
|
| 238 |
+
"length_penalty/max": 1.7182817459106445,
|
| 239 |
+
"length_penalty/mean": 0.14408211456611753,
|
| 240 |
+
"loss": 0.3547,
|
| 241 |
+
"num_tokens": 872850.0,
|
| 242 |
+
"reward": 0.46875,
|
| 243 |
+
"reward_std": 0.1377599686384201,
|
| 244 |
+
"rewards/accuracy_reward/mean": 0.0,
|
| 245 |
+
"rewards/accuracy_reward/std": 0.0,
|
| 246 |
+
"rewards/format_reward/mean": 0.0,
|
| 247 |
+
"rewards/format_reward/std": 0.0,
|
| 248 |
+
"rewards/tag_count_reward/mean": 0.46875,
|
| 249 |
+
"rewards/tag_count_reward/std": 0.1377599686384201,
|
| 250 |
+
"step": 8
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"clip_ratio/high_max": 0.0,
|
| 254 |
+
"clip_ratio/high_mean": 0.0,
|
| 255 |
+
"clip_ratio/low_mean": 0.0,
|
| 256 |
+
"clip_ratio/low_min": 0.0,
|
| 257 |
+
"clip_ratio/region_mean": 0.0,
|
| 258 |
+
"completions/clipped_ratio": 0.65625,
|
| 259 |
+
"completions/max_length": 1775.75,
|
| 260 |
+
"completions/max_terminated_length": 1338.25,
|
| 261 |
+
"completions/mean_length": 1638.125,
|
| 262 |
+
"completions/mean_terminated_length": 1121.9375,
|
| 263 |
+
"completions/min_length": 759.5,
|
| 264 |
+
"completions/min_terminated_length": 759.5,
|
| 265 |
+
"epoch": 0.36,
|
| 266 |
+
"grad_norm": 0.3729335069656372,
|
| 267 |
+
"learning_rate": 8.900873084594162e-07,
|
| 268 |
+
"length_penalty/max": 0.8591408729553223,
|
| 269 |
+
"length_penalty/mean": 0.09117699135094881,
|
| 270 |
+
"loss": 0.2097,
|
| 271 |
+
"num_tokens": 987002.0,
|
| 272 |
+
"reward": 0.44921875,
|
| 273 |
+
"reward_std": 0.2101699709892273,
|
| 274 |
+
"rewards/accuracy_reward/mean": 0.078125,
|
| 275 |
+
"rewards/accuracy_reward/std": 0.11967839300632477,
|
| 276 |
+
"rewards/format_reward/mean": 0.0,
|
| 277 |
+
"rewards/format_reward/std": 0.0,
|
| 278 |
+
"rewards/tag_count_reward/mean": 0.37109375,
|
| 279 |
+
"rewards/tag_count_reward/std": 0.13251042738556862,
|
| 280 |
+
"step": 9
|
| 281 |
+
},
|
| 282 |
+
{
|
| 283 |
+
"clip_ratio/high_max": 0.0,
|
| 284 |
+
"clip_ratio/high_mean": 0.0,
|
| 285 |
+
"clip_ratio/low_mean": 0.0,
|
| 286 |
+
"clip_ratio/low_min": 0.0,
|
| 287 |
+
"clip_ratio/region_mean": 0.0,
|
| 288 |
+
"completions/clipped_ratio": 0.71875,
|
| 289 |
+
"completions/max_length": 2048.0,
|
| 290 |
+
"completions/max_terminated_length": 1423.0,
|
| 291 |
+
"completions/mean_length": 1841.125,
|
| 292 |
+
"completions/mean_terminated_length": 1111.2361145019531,
|
| 293 |
+
"completions/min_length": 1294.0,
|
| 294 |
+
"completions/min_terminated_length": 782.0,
|
| 295 |
+
"epoch": 0.4,
|
| 296 |
+
"grad_norm": 0.28423377871513367,
|
| 297 |
+
"learning_rate": 8.446873302753783e-07,
|
| 298 |
+
"length_penalty/max": 1.2887113094329834,
|
| 299 |
+
"length_penalty/mean": 0.0555590707808733,
|
| 300 |
+
"loss": 0.1336,
|
| 301 |
+
"num_tokens": 1115938.0,
|
| 302 |
+
"reward": 0.36328125,
|
| 303 |
+
"reward_std": 0.1533849686384201,
|
| 304 |
+
"rewards/accuracy_reward/mean": 0.015625,
|
| 305 |
+
"rewards/accuracy_reward/std": 0.0625,
|
| 306 |
+
"rewards/format_reward/mean": 0.0,
|
| 307 |
+
"rewards/format_reward/std": 0.0,
|
| 308 |
+
"rewards/tag_count_reward/mean": 0.34765625,
|
| 309 |
+
"rewards/tag_count_reward/std": 0.109375,
|
| 310 |
+
"step": 10
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"clip_ratio/high_max": 0.0,
|
| 314 |
+
"clip_ratio/high_mean": 0.0,
|
| 315 |
+
"clip_ratio/low_mean": 0.0,
|
| 316 |
+
"clip_ratio/low_min": 0.0,
|
| 317 |
+
"clip_ratio/region_mean": 0.0,
|
| 318 |
+
"completions/clipped_ratio": 0.78125,
|
| 319 |
+
"completions/max_length": 2048.0,
|
| 320 |
+
"completions/max_terminated_length": 1434.25,
|
| 321 |
+
"completions/mean_length": 1883.8125,
|
| 322 |
+
"completions/mean_terminated_length": 1268.5,
|
| 323 |
+
"completions/min_length": 1151.5,
|
| 324 |
+
"completions/min_terminated_length": 1151.5,
|
| 325 |
+
"epoch": 0.44,
|
| 326 |
+
"grad_norm": 0.22027868032455444,
|
| 327 |
+
"learning_rate": 7.932883678550189e-07,
|
| 328 |
+
"length_penalty/max": 0.8591408729553223,
|
| 329 |
+
"length_penalty/mean": 0.03302514273673296,
|
| 330 |
+
"loss": 0.1644,
|
| 331 |
+
"num_tokens": 1245622.0,
|
| 332 |
+
"reward": 0.32421875,
|
| 333 |
+
"reward_std": 0.08912044763565063,
|
| 334 |
+
"rewards/accuracy_reward/mean": 0.0,
|
| 335 |
+
"rewards/accuracy_reward/std": 0.0,
|
| 336 |
+
"rewards/format_reward/mean": 0.0,
|
| 337 |
+
"rewards/format_reward/std": 0.0,
|
| 338 |
+
"rewards/tag_count_reward/mean": 0.32421875,
|
| 339 |
+
"rewards/tag_count_reward/std": 0.08912044949829578,
|
| 340 |
+
"step": 11
|
| 341 |
+
},
|
| 342 |
+
{
|
| 343 |
+
"clip_ratio/high_max": 0.0,
|
| 344 |
+
"clip_ratio/high_mean": 0.0,
|
| 345 |
+
"clip_ratio/low_mean": 0.0,
|
| 346 |
+
"clip_ratio/low_min": 0.0,
|
| 347 |
+
"clip_ratio/region_mean": 0.0,
|
| 348 |
+
"completions/clipped_ratio": 0.1875,
|
| 349 |
+
"completions/max_length": 1946.75,
|
| 350 |
+
"completions/max_terminated_length": 1911.5,
|
| 351 |
+
"completions/mean_length": 1276.859375,
|
| 352 |
+
"completions/mean_terminated_length": 1185.3010559082031,
|
| 353 |
+
"completions/min_length": 700.0,
|
| 354 |
+
"completions/min_terminated_length": 700.0,
|
| 355 |
+
"epoch": 0.48,
|
| 356 |
+
"grad_norm": 0.39662182331085205,
|
| 357 |
+
"learning_rate": 7.36936755850849e-07,
|
| 358 |
+
"length_penalty/max": 1.7182817459106445,
|
| 359 |
+
"length_penalty/mean": 0.12480469699949026,
|
| 360 |
+
"loss": 0.2638,
|
| 361 |
+
"num_tokens": 1337741.0,
|
| 362 |
+
"reward": 0.80078125,
|
| 363 |
+
"reward_std": 0.31010416708886623,
|
| 364 |
+
"rewards/accuracy_reward/mean": 0.28125,
|
| 365 |
+
"rewards/accuracy_reward/std": 0.2561737895011902,
|
| 366 |
+
"rewards/format_reward/mean": 0.0,
|
| 367 |
+
"rewards/format_reward/std": 0.0,
|
| 368 |
+
"rewards/tag_count_reward/mean": 0.51953125,
|
| 369 |
+
"rewards/tag_count_reward/std": 0.11239240504801273,
|
| 370 |
+
"step": 12
|
| 371 |
+
},
|
| 372 |
+
{
|
| 373 |
+
"clip_ratio/high_max": 0.0,
|
| 374 |
+
"clip_ratio/high_mean": 0.0,
|
| 375 |
+
"clip_ratio/low_mean": 0.0,
|
| 376 |
+
"clip_ratio/low_min": 0.0,
|
| 377 |
+
"clip_ratio/region_mean": 0.0,
|
| 378 |
+
"completions/clipped_ratio": 0.828125,
|
| 379 |
+
"completions/max_length": 2048.0,
|
| 380 |
+
"completions/max_terminated_length": 1780.25,
|
| 381 |
+
"completions/mean_length": 1930.375,
|
| 382 |
+
"completions/mean_terminated_length": 1433.2708435058594,
|
| 383 |
+
"completions/min_length": 1187.75,
|
| 384 |
+
"completions/min_terminated_length": 1187.75,
|
| 385 |
+
"epoch": 0.52,
|
| 386 |
+
"grad_norm": 0.2315625697374344,
|
| 387 |
+
"learning_rate": 6.767796505786434e-07,
|
| 388 |
+
"length_penalty/max": 0.8591408729553223,
|
| 389 |
+
"length_penalty/mean": 0.014636571402661502,
|
| 390 |
+
"loss": 0.1469,
|
| 391 |
+
"num_tokens": 1470789.0,
|
| 392 |
+
"reward": 0.4140625,
|
| 393 |
+
"reward_std": 0.3433516249060631,
|
| 394 |
+
"rewards/accuracy_reward/mean": 0.109375,
|
| 395 |
+
"rewards/accuracy_reward/std": 0.2596946656703949,
|
| 396 |
+
"rewards/format_reward/mean": 0.0,
|
| 397 |
+
"rewards/format_reward/std": 0.0,
|
| 398 |
+
"rewards/tag_count_reward/mean": 0.3046875,
|
| 399 |
+
"rewards/tag_count_reward/std": 0.10298692621290684,
|
| 400 |
+
"step": 13
|
| 401 |
+
},
|
| 402 |
+
{
|
| 403 |
+
"clip_ratio/high_max": 0.0,
|
| 404 |
+
"clip_ratio/high_mean": 0.0,
|
| 405 |
+
"clip_ratio/low_mean": 0.0,
|
| 406 |
+
"clip_ratio/low_min": 0.0,
|
| 407 |
+
"clip_ratio/region_mean": 0.0,
|
| 408 |
+
"completions/clipped_ratio": 0.40625,
|
| 409 |
+
"completions/max_length": 1753.25,
|
| 410 |
+
"completions/max_terminated_length": 1625.0,
|
| 411 |
+
"completions/mean_length": 1386.890625,
|
| 412 |
+
"completions/mean_terminated_length": 1197.8125,
|
| 413 |
+
"completions/min_length": 921.0,
|
| 414 |
+
"completions/min_terminated_length": 921.0,
|
| 415 |
+
"epoch": 0.56,
|
| 416 |
+
"grad_norm": 0.3378031253814697,
|
| 417 |
+
"learning_rate": 6.140416772229784e-07,
|
| 418 |
+
"length_penalty/max": 1.2887113094329834,
|
| 419 |
+
"length_penalty/mean": 0.03282552631571889,
|
| 420 |
+
"loss": 0.0967,
|
| 421 |
+
"num_tokens": 1568638.0,
|
| 422 |
+
"reward": 0.53125,
|
| 423 |
+
"reward_std": 0.2592369243502617,
|
| 424 |
+
"rewards/accuracy_reward/mean": 0.09375,
|
| 425 |
+
"rewards/accuracy_reward/std": 0.125,
|
| 426 |
+
"rewards/format_reward/mean": 0.0,
|
| 427 |
+
"rewards/format_reward/std": 0.0,
|
| 428 |
+
"rewards/tag_count_reward/mean": 0.4375,
|
| 429 |
+
"rewards/tag_count_reward/std": 0.13423692621290684,
|
| 430 |
+
"step": 14
|
| 431 |
+
},
|
| 432 |
+
{
|
| 433 |
+
"clip_ratio/high_max": 0.0,
|
| 434 |
+
"clip_ratio/high_mean": 0.0,
|
| 435 |
+
"clip_ratio/low_mean": 0.0,
|
| 436 |
+
"clip_ratio/low_min": 0.0,
|
| 437 |
+
"clip_ratio/region_mean": 0.0,
|
| 438 |
+
"completions/clipped_ratio": 0.203125,
|
| 439 |
+
"completions/max_length": 1816.25,
|
| 440 |
+
"completions/max_terminated_length": 1311.75,
|
| 441 |
+
"completions/mean_length": 1082.921875,
|
| 442 |
+
"completions/mean_terminated_length": 881.4708404541016,
|
| 443 |
+
"completions/min_length": 589.75,
|
| 444 |
+
"completions/min_terminated_length": 589.75,
|
| 445 |
+
"epoch": 0.6,
|
| 446 |
+
"grad_norm": 0.3406469523906708,
|
| 447 |
+
"learning_rate": 5.5e-07,
|
| 448 |
+
"length_penalty/max": 1.7182817459106445,
|
| 449 |
+
"length_penalty/mean": 0.08080398238962516,
|
| 450 |
+
"loss": 0.2032,
|
| 451 |
+
"num_tokens": 1650633.0,
|
| 452 |
+
"reward": 0.82421875,
|
| 453 |
+
"reward_std": 0.44111213833093643,
|
| 454 |
+
"rewards/accuracy_reward/mean": 0.3125,
|
| 455 |
+
"rewards/accuracy_reward/std": 0.36797719448804855,
|
| 456 |
+
"rewards/format_reward/mean": 0.0,
|
| 457 |
+
"rewards/format_reward/std": 0.0,
|
| 458 |
+
"rewards/tag_count_reward/mean": 0.51171875,
|
| 459 |
+
"rewards/tag_count_reward/std": 0.17121704295277596,
|
| 460 |
+
"step": 15
|
| 461 |
+
},
|
| 462 |
+
{
|
| 463 |
+
"clip_ratio/high_max": 0.0,
|
| 464 |
+
"clip_ratio/high_mean": 0.0,
|
| 465 |
+
"clip_ratio/low_mean": 0.0,
|
| 466 |
+
"clip_ratio/low_min": 0.0,
|
| 467 |
+
"clip_ratio/region_mean": 0.0,
|
| 468 |
+
"completions/clipped_ratio": 0.65625,
|
| 469 |
+
"completions/max_length": 2048.0,
|
| 470 |
+
"completions/max_terminated_length": 1934.25,
|
| 471 |
+
"completions/mean_length": 1840.953125,
|
| 472 |
+
"completions/mean_terminated_length": 1465.6999969482422,
|
| 473 |
+
"completions/min_length": 1048.5,
|
| 474 |
+
"completions/min_terminated_length": 1048.5,
|
| 475 |
+
"epoch": 0.64,
|
| 476 |
+
"grad_norm": 0.2429068237543106,
|
| 477 |
+
"learning_rate": 4.859583227770217e-07,
|
| 478 |
+
"length_penalty/max": 1.5175965428352356,
|
| 479 |
+
"length_penalty/mean": 0.034873521068220725,
|
| 480 |
+
"loss": 0.2006,
|
| 481 |
+
"num_tokens": 1777382.0,
|
| 482 |
+
"reward": 0.37890625,
|
| 483 |
+
"reward_std": 0.17128954827785492,
|
| 484 |
+
"rewards/accuracy_reward/mean": 0.0,
|
| 485 |
+
"rewards/accuracy_reward/std": 0.0,
|
| 486 |
+
"rewards/format_reward/mean": 0.0,
|
| 487 |
+
"rewards/format_reward/std": 0.0,
|
| 488 |
+
"rewards/tag_count_reward/mean": 0.37890625,
|
| 489 |
+
"rewards/tag_count_reward/std": 0.17128955200314522,
|
| 490 |
+
"step": 16
|
| 491 |
+
},
|
| 492 |
+
{
|
| 493 |
+
"clip_ratio/high_max": 0.0,
|
| 494 |
+
"clip_ratio/high_mean": 0.0,
|
| 495 |
+
"clip_ratio/low_mean": 0.0,
|
| 496 |
+
"clip_ratio/low_min": 0.0,
|
| 497 |
+
"clip_ratio/region_mean": 0.0,
|
| 498 |
+
"completions/clipped_ratio": 0.6875,
|
| 499 |
+
"completions/max_length": 1944.75,
|
| 500 |
+
"completions/max_terminated_length": 1414.25,
|
| 501 |
+
"completions/mean_length": 1839.71875,
|
| 502 |
+
"completions/mean_terminated_length": 1185.09375,
|
| 503 |
+
"completions/min_length": 1461.0,
|
| 504 |
+
"completions/min_terminated_length": 949.0,
|
| 505 |
+
"epoch": 0.68,
|
| 506 |
+
"grad_norm": 0.23434214293956757,
|
| 507 |
+
"learning_rate": 4.2322034942135664e-07,
|
| 508 |
+
"length_penalty/max": 0.8591408729553223,
|
| 509 |
+
"length_penalty/mean": 0.036483183183008805,
|
| 510 |
+
"loss": 0.0793,
|
| 511 |
+
"num_tokens": 1909060.0,
|
| 512 |
+
"reward": 0.359375,
|
| 513 |
+
"reward_std": 0.131654754281044,
|
| 514 |
+
"rewards/accuracy_reward/mean": 0.015625,
|
| 515 |
+
"rewards/accuracy_reward/std": 0.0625,
|
| 516 |
+
"rewards/format_reward/mean": 0.0,
|
| 517 |
+
"rewards/format_reward/std": 0.0,
|
| 518 |
+
"rewards/tag_count_reward/mean": 0.34375,
|
| 519 |
+
"rewards/tag_count_reward/std": 0.08054866641759872,
|
| 520 |
+
"step": 17
|
| 521 |
+
},
|
| 522 |
+
{
|
| 523 |
+
"clip_ratio/high_max": 0.0,
|
| 524 |
+
"clip_ratio/high_mean": 0.0,
|
| 525 |
+
"clip_ratio/low_mean": 0.0,
|
| 526 |
+
"clip_ratio/low_min": 0.0,
|
| 527 |
+
"clip_ratio/region_mean": 0.0,
|
| 528 |
+
"completions/clipped_ratio": 0.515625,
|
| 529 |
+
"completions/max_length": 2048.0,
|
| 530 |
+
"completions/max_terminated_length": 1832.0,
|
| 531 |
+
"completions/mean_length": 1656.09375,
|
| 532 |
+
"completions/mean_terminated_length": 1481.7916870117188,
|
| 533 |
+
"completions/min_length": 1184.75,
|
| 534 |
+
"completions/min_terminated_length": 1184.75,
|
| 535 |
+
"epoch": 0.72,
|
| 536 |
+
"grad_norm": 0.25593411922454834,
|
| 537 |
+
"learning_rate": 3.630632441491511e-07,
|
| 538 |
+
"length_penalty/max": 1.7182817459106445,
|
| 539 |
+
"length_penalty/mean": 0.09959548246115446,
|
| 540 |
+
"loss": 0.1676,
|
| 541 |
+
"num_tokens": 2024506.0,
|
| 542 |
+
"reward": 0.4375,
|
| 543 |
+
"reward_std": 0.21411163359880447,
|
| 544 |
+
"rewards/accuracy_reward/mean": 0.03125,
|
| 545 |
+
"rewards/accuracy_reward/std": 0.125,
|
| 546 |
+
"rewards/format_reward/mean": 0.0,
|
| 547 |
+
"rewards/format_reward/std": 0.0,
|
| 548 |
+
"rewards/tag_count_reward/mean": 0.40625,
|
| 549 |
+
"rewards/tag_count_reward/std": 0.12007714062929153,
|
| 550 |
+
"step": 18
|
| 551 |
+
},
|
| 552 |
+
{
|
| 553 |
+
"clip_ratio/high_max": 0.0,
|
| 554 |
+
"clip_ratio/high_mean": 0.0,
|
| 555 |
+
"clip_ratio/low_mean": 0.0,
|
| 556 |
+
"clip_ratio/low_min": 0.0,
|
| 557 |
+
"clip_ratio/region_mean": 0.0,
|
| 558 |
+
"completions/clipped_ratio": 0.578125,
|
| 559 |
+
"completions/max_length": 2048.0,
|
| 560 |
+
"completions/max_terminated_length": 1655.5,
|
| 561 |
+
"completions/mean_length": 1796.53125,
|
| 562 |
+
"completions/mean_terminated_length": 1361.3659362792969,
|
| 563 |
+
"completions/min_length": 1045.25,
|
| 564 |
+
"completions/min_terminated_length": 1045.25,
|
| 565 |
+
"epoch": 0.76,
|
| 566 |
+
"grad_norm": 0.22864650189876556,
|
| 567 |
+
"learning_rate": 3.0671163214498125e-07,
|
| 568 |
+
"length_penalty/max": 1.2887113094329834,
|
| 569 |
+
"length_penalty/mean": 0.042454172391444445,
|
| 570 |
+
"loss": 0.1844,
|
| 571 |
+
"num_tokens": 2147164.0,
|
| 572 |
+
"reward": 0.5,
|
| 573 |
+
"reward_std": 0.24606626480817795,
|
| 574 |
+
"rewards/accuracy_reward/mean": 0.125,
|
| 575 |
+
"rewards/accuracy_reward/std": 0.12909944355487823,
|
| 576 |
+
"rewards/format_reward/mean": 0.0,
|
| 577 |
+
"rewards/format_reward/std": 0.0,
|
| 578 |
+
"rewards/tag_count_reward/mean": 0.375,
|
| 579 |
+
"rewards/tag_count_reward/std": 0.12002557702362537,
|
| 580 |
+
"step": 19
|
| 581 |
+
},
|
| 582 |
+
{
|
| 583 |
+
"clip_ratio/high_max": 0.0,
|
| 584 |
+
"clip_ratio/high_mean": 0.0,
|
| 585 |
+
"clip_ratio/low_mean": 0.0,
|
| 586 |
+
"clip_ratio/low_min": 0.0,
|
| 587 |
+
"clip_ratio/region_mean": 0.0,
|
| 588 |
+
"completions/clipped_ratio": 0.578125,
|
| 589 |
+
"completions/max_length": 2048.0,
|
| 590 |
+
"completions/max_terminated_length": 1936.75,
|
| 591 |
+
"completions/mean_length": 1776.328125,
|
| 592 |
+
"completions/mean_terminated_length": 1501.357177734375,
|
| 593 |
+
"completions/min_length": 1152.0,
|
| 594 |
+
"completions/min_terminated_length": 1152.0,
|
| 595 |
+
"epoch": 0.8,
|
| 596 |
+
"grad_norm": 0.23410044610500336,
|
| 597 |
+
"learning_rate": 2.553126697246217e-07,
|
| 598 |
+
"length_penalty/max": 0.8591408729553223,
|
| 599 |
+
"length_penalty/mean": 0.03306913282722235,
|
| 600 |
+
"loss": 0.1182,
|
| 601 |
+
"num_tokens": 2270321.0,
|
| 602 |
+
"reward": 0.3984375,
|
| 603 |
+
"reward_std": 0.18846620619297028,
|
| 604 |
+
"rewards/accuracy_reward/mean": 0.03125,
|
| 605 |
+
"rewards/accuracy_reward/std": 0.08539126068353653,
|
| 606 |
+
"rewards/format_reward/mean": 0.0,
|
| 607 |
+
"rewards/format_reward/std": 0.0,
|
| 608 |
+
"rewards/tag_count_reward/mean": 0.3671875,
|
| 609 |
+
"rewards/tag_count_reward/std": 0.11579498276114464,
|
| 610 |
+
"step": 20
|
| 611 |
+
},
|
| 612 |
+
{
|
| 613 |
+
"clip_ratio/high_max": 0.0,
|
| 614 |
+
"clip_ratio/high_mean": 0.0,
|
| 615 |
+
"clip_ratio/low_mean": 0.0,
|
| 616 |
+
"clip_ratio/low_min": 0.0,
|
| 617 |
+
"clip_ratio/region_mean": 0.0,
|
| 618 |
+
"completions/clipped_ratio": 0.890625,
|
| 619 |
+
"completions/max_length": 2048.0,
|
| 620 |
+
"completions/max_terminated_length": 1299.75,
|
| 621 |
+
"completions/mean_length": 1982.984375,
|
| 622 |
+
"completions/mean_terminated_length": 1205.1499938964844,
|
| 623 |
+
"completions/min_length": 1604.5,
|
| 624 |
+
"completions/min_terminated_length": 1092.5,
|
| 625 |
+
"epoch": 0.84,
|
| 626 |
+
"grad_norm": 0.23763558268547058,
|
| 627 |
+
"learning_rate": 2.0991269154058384e-07,
|
| 628 |
+
"length_penalty/max": 1.2887113094329834,
|
| 629 |
+
"length_penalty/mean": 0.009122955147176981,
|
| 630 |
+
"loss": 0.0829,
|
| 631 |
+
"num_tokens": 2406048.0,
|
| 632 |
+
"reward": 0.30078125,
|
| 633 |
+
"reward_std": 0.0843954049050808,
|
| 634 |
+
"rewards/accuracy_reward/mean": 0.0,
|
| 635 |
+
"rewards/accuracy_reward/std": 0.0,
|
| 636 |
+
"rewards/format_reward/mean": 0.0,
|
| 637 |
+
"rewards/format_reward/std": 0.0,
|
| 638 |
+
"rewards/tag_count_reward/mean": 0.30078125,
|
| 639 |
+
"rewards/tag_count_reward/std": 0.08439540676772594,
|
| 640 |
+
"step": 21
|
| 641 |
+
},
|
| 642 |
+
{
|
| 643 |
+
"clip_ratio/high_max": 0.0,
|
| 644 |
+
"clip_ratio/high_mean": 0.0,
|
| 645 |
+
"clip_ratio/low_mean": 0.0,
|
| 646 |
+
"clip_ratio/low_min": 0.0,
|
| 647 |
+
"clip_ratio/region_mean": 0.0,
|
| 648 |
+
"completions/clipped_ratio": 0.203125,
|
| 649 |
+
"completions/max_length": 2048.0,
|
| 650 |
+
"completions/max_terminated_length": 1631.25,
|
| 651 |
+
"completions/mean_length": 1375.09375,
|
| 652 |
+
"completions/mean_terminated_length": 1211.906997680664,
|
| 653 |
+
"completions/min_length": 609.75,
|
| 654 |
+
"completions/min_terminated_length": 609.75,
|
| 655 |
+
"epoch": 0.88,
|
| 656 |
+
"grad_norm": 0.32974064350128174,
|
| 657 |
+
"learning_rate": 1.7143591022596842e-07,
|
| 658 |
+
"length_penalty/max": 1.7182817459106445,
|
| 659 |
+
"length_penalty/mean": 0.04439200600609183,
|
| 660 |
+
"loss": 0.1938,
|
| 661 |
+
"num_tokens": 2502486.0,
|
| 662 |
+
"reward": 0.484375,
|
| 663 |
+
"reward_std": 0.15886800736188889,
|
| 664 |
+
"rewards/accuracy_reward/mean": NaN,
|
| 665 |
+
"rewards/accuracy_reward/std": NaN,
|
| 666 |
+
"rewards/format_reward/mean": 0.0,
|
| 667 |
+
"rewards/format_reward/std": 0.0,
|
| 668 |
+
"rewards/tag_count_reward/mean": 0.484375,
|
| 669 |
+
"rewards/tag_count_reward/std": 0.15886801108717918,
|
| 670 |
+
"step": 22
|
| 671 |
+
},
|
| 672 |
+
{
|
| 673 |
+
"clip_ratio/high_max": 0.0,
|
| 674 |
+
"clip_ratio/high_mean": 0.0,
|
| 675 |
+
"clip_ratio/low_mean": 0.0,
|
| 676 |
+
"clip_ratio/low_min": 0.0,
|
| 677 |
+
"clip_ratio/region_mean": 0.0,
|
| 678 |
+
"completions/clipped_ratio": 0.625,
|
| 679 |
+
"completions/max_length": 2048.0,
|
| 680 |
+
"completions/max_terminated_length": 1727.75,
|
| 681 |
+
"completions/mean_length": 1790.875,
|
| 682 |
+
"completions/mean_terminated_length": 1463.1750183105469,
|
| 683 |
+
"completions/min_length": 1102.5,
|
| 684 |
+
"completions/min_terminated_length": 1102.5,
|
| 685 |
+
"epoch": 0.92,
|
| 686 |
+
"grad_norm": 0.26998037099838257,
|
| 687 |
+
"learning_rate": 1.406656020904667e-07,
|
| 688 |
+
"length_penalty/max": 1.2887113094329834,
|
| 689 |
+
"length_penalty/mean": 0.04420536628458649,
|
| 690 |
+
"loss": 0.1293,
|
| 691 |
+
"num_tokens": 2627998.0,
|
| 692 |
+
"reward": 0.3671875,
|
| 693 |
+
"reward_std": 0.08904074877500534,
|
| 694 |
+
"rewards/accuracy_reward/mean": 0.0,
|
| 695 |
+
"rewards/accuracy_reward/std": 0.0,
|
| 696 |
+
"rewards/format_reward/mean": 0.0,
|
| 697 |
+
"rewards/format_reward/std": 0.0,
|
| 698 |
+
"rewards/tag_count_reward/mean": 0.3671875,
|
| 699 |
+
"rewards/tag_count_reward/std": 0.08904075436294079,
|
| 700 |
+
"step": 23
|
| 701 |
+
},
|
| 702 |
+
{
|
| 703 |
+
"clip_ratio/high_max": 0.0,
|
| 704 |
+
"clip_ratio/high_mean": 0.0,
|
| 705 |
+
"clip_ratio/low_mean": 0.0,
|
| 706 |
+
"clip_ratio/low_min": 0.0,
|
| 707 |
+
"clip_ratio/region_mean": 0.0,
|
| 708 |
+
"completions/clipped_ratio": 0.984375,
|
| 709 |
+
"completions/max_length": 2048.0,
|
| 710 |
+
"completions/max_terminated_length": 319.0,
|
| 711 |
+
"completions/mean_length": 2035.9375,
|
| 712 |
+
"completions/mean_terminated_length": 319.0,
|
| 713 |
+
"completions/min_length": 1855.0,
|
| 714 |
+
"completions/min_terminated_length": 319.0,
|
| 715 |
+
"epoch": 0.96,
|
| 716 |
+
"grad_norm": 0.13524992763996124,
|
| 717 |
+
"learning_rate": 1.1822816187347622e-07,
|
| 718 |
+
"length_penalty/max": 0.0,
|
| 719 |
+
"length_penalty/mean": 0.0,
|
| 720 |
+
"loss": 0.0531,
|
| 721 |
+
"num_tokens": 2769834.0,
|
| 722 |
+
"reward": 0.2578125,
|
| 723 |
+
"reward_std": 0.021347813308238983,
|
| 724 |
+
"rewards/accuracy_reward/mean": 0.0,
|
| 725 |
+
"rewards/accuracy_reward/std": 0.0,
|
| 726 |
+
"rewards/format_reward/mean": 0.0,
|
| 727 |
+
"rewards/format_reward/std": 0.0,
|
| 728 |
+
"rewards/tag_count_reward/mean": 0.2578125,
|
| 729 |
+
"rewards/tag_count_reward/std": 0.021347815170884132,
|
| 730 |
+
"step": 24
|
| 731 |
+
},
|
| 732 |
+
{
|
| 733 |
+
"clip_ratio/high_max": 0.0,
|
| 734 |
+
"clip_ratio/high_mean": 0.0,
|
| 735 |
+
"clip_ratio/low_mean": 0.0,
|
| 736 |
+
"clip_ratio/low_min": 0.0,
|
| 737 |
+
"clip_ratio/region_mean": 0.0,
|
| 738 |
+
"completions/clipped_ratio": 0.625,
|
| 739 |
+
"completions/max_length": 2048.0,
|
| 740 |
+
"completions/max_terminated_length": 1416.5,
|
| 741 |
+
"completions/mean_length": 1802.1875,
|
| 742 |
+
"completions/mean_terminated_length": 1391.0833435058594,
|
| 743 |
+
"completions/min_length": 1364.75,
|
| 744 |
+
"completions/min_terminated_length": 1364.75,
|
| 745 |
+
"epoch": 1.0,
|
| 746 |
+
"grad_norm": 0.26286277174949646,
|
| 747 |
+
"learning_rate": 1.045803511535803e-07,
|
| 748 |
+
"length_penalty/max": 1.2887113094329834,
|
| 749 |
+
"length_penalty/mean": 0.023291549878194928,
|
| 750 |
+
"loss": 0.0976,
|
| 751 |
+
"num_tokens": 2897030.0,
|
| 752 |
+
"reward": 0.484375,
|
| 753 |
+
"reward_std": 0.31899941712617874,
|
| 754 |
+
"rewards/accuracy_reward/mean": 0.09375,
|
| 755 |
+
"rewards/accuracy_reward/std": 0.1971946656703949,
|
| 756 |
+
"rewards/format_reward/mean": 0.0,
|
| 757 |
+
"rewards/format_reward/std": 0.0,
|
| 758 |
+
"rewards/tag_count_reward/mean": 0.390625,
|
| 759 |
+
"rewards/tag_count_reward/std": 0.16548692621290684,
|
| 760 |
+
"step": 25
|
| 761 |
+
},
|
| 762 |
+
{
|
| 763 |
+
"epoch": 1.0,
|
| 764 |
+
"step": 25,
|
| 765 |
+
"total_flos": 0.0,
|
| 766 |
+
"train_loss": 0.1677406519651413,
|
| 767 |
+
"train_runtime": 1369.1313,
|
| 768 |
+
"train_samples_per_second": 0.073,
|
| 769 |
+
"train_steps_per_second": 0.018
|
| 770 |
+
}
|
| 771 |
+
],
|
| 772 |
+
"logging_steps": 1,
|
| 773 |
+
"max_steps": 25,
|
| 774 |
+
"num_input_tokens_seen": 2897030,
|
| 775 |
+
"num_train_epochs": 1,
|
| 776 |
+
"save_steps": 1000,
|
| 777 |
+
"stateful_callbacks": {
|
| 778 |
+
"TrainerControl": {
|
| 779 |
+
"args": {
|
| 780 |
+
"should_epoch_stop": false,
|
| 781 |
+
"should_evaluate": false,
|
| 782 |
+
"should_log": false,
|
| 783 |
+
"should_save": true,
|
| 784 |
+
"should_training_stop": true
|
| 785 |
+
},
|
| 786 |
+
"attributes": {}
|
| 787 |
+
}
|
| 788 |
+
},
|
| 789 |
+
"total_flos": 0.0,
|
| 790 |
+
"train_batch_size": 16,
|
| 791 |
+
"trial_name": null,
|
| 792 |
+
"trial_params": null
|
| 793 |
+
}
|