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 +2283 -0
README.md
ADDED
|
@@ -0,0 +1,68 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
base_model: Qwen/Qwen2.5-7B
|
| 3 |
+
library_name: transformers
|
| 4 |
+
model_name: Qwen-2.5-Base-7B-gen8-math3to5-ghpo-cold20-3Dhint-prompt1-epoch5-cosine0515-v2
|
| 5 |
+
tags:
|
| 6 |
+
- generated_from_trainer
|
| 7 |
+
- trl
|
| 8 |
+
- grpo
|
| 9 |
+
licence: license
|
| 10 |
+
---
|
| 11 |
+
|
| 12 |
+
# Model Card for Qwen-2.5-Base-7B-gen8-math3to5-ghpo-cold20-3Dhint-prompt1-epoch5-cosine0515-v2
|
| 13 |
+
|
| 14 |
+
This model is a fine-tuned version of [Qwen/Qwen2.5-7B](https://huggingface.co/Qwen/Qwen2.5-7B).
|
| 15 |
+
It has been trained using [TRL](https://github.com/huggingface/trl).
|
| 16 |
+
|
| 17 |
+
## Quick start
|
| 18 |
+
|
| 19 |
+
```python
|
| 20 |
+
from transformers import pipeline
|
| 21 |
+
|
| 22 |
+
question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
|
| 23 |
+
generator = pipeline("text-generation", model="Lansechen/Qwen-2.5-Base-7B-gen8-math3to5-ghpo-cold20-3Dhint-prompt1-epoch5-cosine0515-v2", device="cuda")
|
| 24 |
+
output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
|
| 25 |
+
print(output["generated_text"])
|
| 26 |
+
```
|
| 27 |
+
|
| 28 |
+
## Training procedure
|
| 29 |
+
|
| 30 |
+
[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/chenran1995-the-chinese-university-of-hong-kong/huggingface/runs/sho5xj5s)
|
| 31 |
+
|
| 32 |
+
|
| 33 |
+
This model was trained with GRPO, a method introduced in [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://huggingface.co/papers/2402.03300).
|
| 34 |
+
|
| 35 |
+
### Framework versions
|
| 36 |
+
|
| 37 |
+
- TRL: 0.18.0.dev0
|
| 38 |
+
- Transformers: 4.52.0.dev0
|
| 39 |
+
- Pytorch: 2.6.0
|
| 40 |
+
- Datasets: 3.5.1
|
| 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": 0.012358659607228598,
|
| 4 |
+
"train_runtime": 80233.6203,
|
| 5 |
+
"train_samples": 8888,
|
| 6 |
+
"train_samples_per_second": 0.554,
|
| 7 |
+
"train_steps_per_second": 0.005
|
| 8 |
+
}
|
generation_config.json
ADDED
|
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bos_token_id": 151643,
|
| 3 |
+
"eos_token_id": 151643,
|
| 4 |
+
"max_new_tokens": 2048,
|
| 5 |
+
"transformers_version": "4.52.0.dev0"
|
| 6 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"total_flos": 0.0,
|
| 3 |
+
"train_loss": 0.012358659607228598,
|
| 4 |
+
"train_runtime": 80233.6203,
|
| 5 |
+
"train_samples": 8888,
|
| 6 |
+
"train_samples_per_second": 0.554,
|
| 7 |
+
"train_steps_per_second": 0.005
|
| 8 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,2283 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_global_step": null,
|
| 3 |
+
"best_metric": null,
|
| 4 |
+
"best_model_checkpoint": null,
|
| 5 |
+
"epoch": 4.995499549954996,
|
| 6 |
+
"eval_steps": 15,
|
| 7 |
+
"global_step": 395,
|
| 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": 0.030133928571428603,
|
| 19 |
+
"completions/max_length": 2048.0,
|
| 20 |
+
"completions/max_terminated_length": 2017.0,
|
| 21 |
+
"completions/mean_length": 559.7366333007812,
|
| 22 |
+
"completions/mean_terminated_length": 513.4959716796875,
|
| 23 |
+
"completions/min_length": 4.0,
|
| 24 |
+
"completions/min_terminated_length": 4.0,
|
| 25 |
+
"epoch": 0.012601260126012601,
|
| 26 |
+
"grad_norm": 0.40923070907592773,
|
| 27 |
+
"learning_rate": 0.0,
|
| 28 |
+
"loss": 0.0084,
|
| 29 |
+
"num_tokens": 629108.0,
|
| 30 |
+
"reward": 0.263950914144516,
|
| 31 |
+
"reward_std": 0.3971380889415741,
|
| 32 |
+
"rewards/accuracy_reward/mean": 0.1607142835855484,
|
| 33 |
+
"rewards/accuracy_reward/std": 0.3674723207950592,
|
| 34 |
+
"rewards/cosine_scaled_reward/mean": 0.1607142835855484,
|
| 35 |
+
"rewards/cosine_scaled_reward/std": 0.3674723207950592,
|
| 36 |
+
"rewards/format_reward/mean": 0.0457589291036129,
|
| 37 |
+
"rewards/format_reward/std": 0.20907855033874512,
|
| 38 |
+
"step": 1
|
| 39 |
+
},
|
| 40 |
+
{
|
| 41 |
+
"clip_ratio/high_max": 0.0,
|
| 42 |
+
"clip_ratio/high_mean": 0.0,
|
| 43 |
+
"clip_ratio/low_mean": 0.0,
|
| 44 |
+
"clip_ratio/low_min": 0.0,
|
| 45 |
+
"clip_ratio/region_mean": 0.0,
|
| 46 |
+
"completions/clipped_ratio": 0.01981026785714285,
|
| 47 |
+
"completions/max_length": 2048.0,
|
| 48 |
+
"completions/max_terminated_length": 1937.5,
|
| 49 |
+
"completions/mean_length": 523.0039215087891,
|
| 50 |
+
"completions/mean_terminated_length": 492.1688919067383,
|
| 51 |
+
"completions/min_length": 3.5,
|
| 52 |
+
"completions/min_terminated_length": 3.5,
|
| 53 |
+
"epoch": 0.063006300630063,
|
| 54 |
+
"grad_norm": 0.43882623314857483,
|
| 55 |
+
"learning_rate": 1e-07,
|
| 56 |
+
"loss": -0.0058,
|
| 57 |
+
"num_tokens": 3051026.0,
|
| 58 |
+
"reward": 0.25034814700484276,
|
| 59 |
+
"reward_std": 0.38232600688934326,
|
| 60 |
+
"rewards/accuracy_reward/mean": 0.1526227667927742,
|
| 61 |
+
"rewards/accuracy_reward/std": 0.35816748440265656,
|
| 62 |
+
"rewards/cosine_scaled_reward/mean": 0.15248198807239532,
|
| 63 |
+
"rewards/cosine_scaled_reward/std": 0.3578503802418709,
|
| 64 |
+
"rewards/format_reward/mean": 0.04296875,
|
| 65 |
+
"rewards/format_reward/std": 0.2016594335436821,
|
| 66 |
+
"step": 5
|
| 67 |
+
},
|
| 68 |
+
{
|
| 69 |
+
"clip_ratio/high_max": 0.0,
|
| 70 |
+
"clip_ratio/high_mean": 0.0,
|
| 71 |
+
"clip_ratio/low_mean": 0.0,
|
| 72 |
+
"clip_ratio/low_min": 0.0,
|
| 73 |
+
"clip_ratio/region_mean": 0.0,
|
| 74 |
+
"completions/clipped_ratio": 0.02522321428571428,
|
| 75 |
+
"completions/max_length": 2048.0,
|
| 76 |
+
"completions/max_terminated_length": 2008.8,
|
| 77 |
+
"completions/mean_length": 549.8663208007813,
|
| 78 |
+
"completions/mean_terminated_length": 511.09259643554685,
|
| 79 |
+
"completions/min_length": 4.8,
|
| 80 |
+
"completions/min_terminated_length": 4.8,
|
| 81 |
+
"epoch": 0.126012601260126,
|
| 82 |
+
"grad_norm": 0.3167170286178589,
|
| 83 |
+
"learning_rate": 2.25e-07,
|
| 84 |
+
"loss": -0.0017,
|
| 85 |
+
"num_tokens": 6199283.0,
|
| 86 |
+
"reward": 0.2494979292154312,
|
| 87 |
+
"reward_std": 0.3860060155391693,
|
| 88 |
+
"rewards/accuracy_reward/mean": 0.15089285969734192,
|
| 89 |
+
"rewards/accuracy_reward/std": 0.3568773686885834,
|
| 90 |
+
"rewards/cosine_scaled_reward/mean": 0.1507815644145012,
|
| 91 |
+
"rewards/cosine_scaled_reward/std": 0.3566217184066772,
|
| 92 |
+
"rewards/format_reward/mean": 0.04642857164144516,
|
| 93 |
+
"rewards/format_reward/std": 0.21049095690250397,
|
| 94 |
+
"step": 10
|
| 95 |
+
},
|
| 96 |
+
{
|
| 97 |
+
"clip_ratio/high_max": 0.0,
|
| 98 |
+
"clip_ratio/high_mean": 0.0,
|
| 99 |
+
"clip_ratio/low_mean": 0.0,
|
| 100 |
+
"clip_ratio/low_min": 0.0,
|
| 101 |
+
"clip_ratio/region_mean": 0.0,
|
| 102 |
+
"completions/clipped_ratio": 0.022767857142857163,
|
| 103 |
+
"completions/max_length": 2048.0,
|
| 104 |
+
"completions/max_terminated_length": 1946.6,
|
| 105 |
+
"completions/mean_length": 550.6145263671875,
|
| 106 |
+
"completions/mean_terminated_length": 515.6882019042969,
|
| 107 |
+
"completions/min_length": 8.6,
|
| 108 |
+
"completions/min_terminated_length": 8.6,
|
| 109 |
+
"epoch": 0.18901890189018902,
|
| 110 |
+
"grad_norm": 0.34566730260849,
|
| 111 |
+
"learning_rate": 3.5e-07,
|
| 112 |
+
"loss": 0.0061,
|
| 113 |
+
"num_tokens": 9310292.0,
|
| 114 |
+
"reward": 0.2928151488304138,
|
| 115 |
+
"reward_std": 0.40939453840255735,
|
| 116 |
+
"rewards/accuracy_reward/mean": 0.17410714328289031,
|
| 117 |
+
"rewards/accuracy_reward/std": 0.378982412815094,
|
| 118 |
+
"rewards/cosine_scaled_reward/mean": 0.1740231305360794,
|
| 119 |
+
"rewards/cosine_scaled_reward/std": 0.37881343960762026,
|
| 120 |
+
"rewards/format_reward/mean": 0.0633928582072258,
|
| 121 |
+
"rewards/format_reward/std": 0.24350899159908296,
|
| 122 |
+
"step": 15
|
| 123 |
+
},
|
| 124 |
+
{
|
| 125 |
+
"clip_ratio/high_max": 0.0,
|
| 126 |
+
"clip_ratio/high_mean": 0.0,
|
| 127 |
+
"clip_ratio/low_mean": 0.0,
|
| 128 |
+
"clip_ratio/low_min": 0.0,
|
| 129 |
+
"clip_ratio/region_mean": 0.0,
|
| 130 |
+
"completions/clipped_ratio": 0.02901785714285714,
|
| 131 |
+
"completions/max_length": 2048.0,
|
| 132 |
+
"completions/max_terminated_length": 1996.8,
|
| 133 |
+
"completions/mean_length": 563.8605102539062,
|
| 134 |
+
"completions/mean_terminated_length": 519.6845397949219,
|
| 135 |
+
"completions/min_length": 3.6,
|
| 136 |
+
"completions/min_terminated_length": 3.6,
|
| 137 |
+
"epoch": 0.252025202520252,
|
| 138 |
+
"grad_norm": 6.714093208312988,
|
| 139 |
+
"learning_rate": 4.7499999999999995e-07,
|
| 140 |
+
"loss": -0.0035,
|
| 141 |
+
"num_tokens": 12491619.0,
|
| 142 |
+
"reward": 0.3050673246383667,
|
| 143 |
+
"reward_std": 0.425398576259613,
|
| 144 |
+
"rewards/accuracy_reward/mean": 0.16316964328289033,
|
| 145 |
+
"rewards/accuracy_reward/std": 0.36818013787269593,
|
| 146 |
+
"rewards/cosine_scaled_reward/mean": 0.16303642094135284,
|
| 147 |
+
"rewards/cosine_scaled_reward/std": 0.3678862988948822,
|
| 148 |
+
"rewards/format_reward/mean": 0.12075892686843873,
|
| 149 |
+
"rewards/format_reward/std": 0.3249303221702576,
|
| 150 |
+
"step": 20
|
| 151 |
+
},
|
| 152 |
+
{
|
| 153 |
+
"clip_ratio/high_max": 0.0,
|
| 154 |
+
"clip_ratio/high_mean": 0.0,
|
| 155 |
+
"clip_ratio/low_mean": 0.0,
|
| 156 |
+
"clip_ratio/low_min": 0.0,
|
| 157 |
+
"clip_ratio/region_mean": 0.0,
|
| 158 |
+
"completions/clipped_ratio": 0.021428571428571443,
|
| 159 |
+
"completions/max_length": 2048.0,
|
| 160 |
+
"completions/max_terminated_length": 1998.8,
|
| 161 |
+
"completions/mean_length": 553.5821655273437,
|
| 162 |
+
"completions/mean_terminated_length": 520.8716064453125,
|
| 163 |
+
"completions/min_length": 4.4,
|
| 164 |
+
"completions/min_terminated_length": 4.4,
|
| 165 |
+
"epoch": 0.31503150315031503,
|
| 166 |
+
"grad_norm": 0.3570958375930786,
|
| 167 |
+
"learning_rate": 6e-07,
|
| 168 |
+
"loss": -0.0081,
|
| 169 |
+
"num_tokens": 15812147.0,
|
| 170 |
+
"reward": 0.4677644371986389,
|
| 171 |
+
"reward_std": 0.6031840801239013,
|
| 172 |
+
"rewards/accuracy_reward/mean": 0.25223416090011597,
|
| 173 |
+
"rewards/accuracy_reward/std": 0.4337872326374054,
|
| 174 |
+
"rewards/cosine_scaled_reward/mean": 0.25069683194160464,
|
| 175 |
+
"rewards/cosine_scaled_reward/std": 0.43269214034080505,
|
| 176 |
+
"rewards/format_reward/mean": 0.1816964328289032,
|
| 177 |
+
"rewards/format_reward/std": 0.38075574636459353,
|
| 178 |
+
"step": 25
|
| 179 |
+
},
|
| 180 |
+
{
|
| 181 |
+
"clip_ratio/high_max": 0.0,
|
| 182 |
+
"clip_ratio/high_mean": 0.0,
|
| 183 |
+
"clip_ratio/low_mean": 0.0,
|
| 184 |
+
"clip_ratio/low_min": 0.0,
|
| 185 |
+
"clip_ratio/region_mean": 0.0,
|
| 186 |
+
"completions/clipped_ratio": 0.01696428571428572,
|
| 187 |
+
"completions/max_length": 2048.0,
|
| 188 |
+
"completions/max_terminated_length": 2001.6,
|
| 189 |
+
"completions/mean_length": 557.9491333007812,
|
| 190 |
+
"completions/mean_terminated_length": 532.2440673828125,
|
| 191 |
+
"completions/min_length": 6.6,
|
| 192 |
+
"completions/min_terminated_length": 6.6,
|
| 193 |
+
"epoch": 0.37803780378037805,
|
| 194 |
+
"grad_norm": 0.4451054632663727,
|
| 195 |
+
"learning_rate": 7.249999999999999e-07,
|
| 196 |
+
"loss": 0.0022,
|
| 197 |
+
"num_tokens": 19200175.0,
|
| 198 |
+
"reward": 0.6430311799049377,
|
| 199 |
+
"reward_std": 0.6464242577552796,
|
| 200 |
+
"rewards/accuracy_reward/mean": 0.3037946462631226,
|
| 201 |
+
"rewards/accuracy_reward/std": 0.4588939487934113,
|
| 202 |
+
"rewards/cosine_scaled_reward/mean": 0.3034730076789856,
|
| 203 |
+
"rewards/cosine_scaled_reward/std": 0.45845308899879456,
|
| 204 |
+
"rewards/format_reward/mean": 0.37499999403953554,
|
| 205 |
+
"rewards/format_reward/std": 0.47941267490386963,
|
| 206 |
+
"step": 30
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"clip_ratio/high_max": 0.0,
|
| 210 |
+
"clip_ratio/high_mean": 0.0,
|
| 211 |
+
"clip_ratio/low_mean": 0.0,
|
| 212 |
+
"clip_ratio/low_min": 0.0,
|
| 213 |
+
"clip_ratio/region_mean": 0.0,
|
| 214 |
+
"completions/clipped_ratio": 0.018080357142857117,
|
| 215 |
+
"completions/max_length": 2048.0,
|
| 216 |
+
"completions/max_terminated_length": 2006.4,
|
| 217 |
+
"completions/mean_length": 611.4062866210937,
|
| 218 |
+
"completions/mean_terminated_length": 584.9557250976562,
|
| 219 |
+
"completions/min_length": 39.0,
|
| 220 |
+
"completions/min_terminated_length": 39.0,
|
| 221 |
+
"epoch": 0.44104410441044106,
|
| 222 |
+
"grad_norm": 0.29937848448753357,
|
| 223 |
+
"learning_rate": 8.499999999999999e-07,
|
| 224 |
+
"loss": 0.0221,
|
| 225 |
+
"num_tokens": 22807531.0,
|
| 226 |
+
"reward": 1.002221155166626,
|
| 227 |
+
"reward_std": 0.6414472222328186,
|
| 228 |
+
"rewards/accuracy_reward/mean": 0.4667410671710968,
|
| 229 |
+
"rewards/accuracy_reward/std": 0.49585614800453187,
|
| 230 |
+
"rewards/cosine_scaled_reward/mean": 0.4662725567817688,
|
| 231 |
+
"rewards/cosine_scaled_reward/std": 0.49544560313224795,
|
| 232 |
+
"rewards/format_reward/mean": 0.604687488079071,
|
| 233 |
+
"rewards/format_reward/std": 0.4856111824512482,
|
| 234 |
+
"step": 35
|
| 235 |
+
},
|
| 236 |
+
{
|
| 237 |
+
"clip_ratio/high_max": 0.0,
|
| 238 |
+
"clip_ratio/high_mean": 0.0,
|
| 239 |
+
"clip_ratio/low_mean": 0.0,
|
| 240 |
+
"clip_ratio/low_min": 0.0,
|
| 241 |
+
"clip_ratio/region_mean": 0.0,
|
| 242 |
+
"completions/clipped_ratio": 0.01540178571428572,
|
| 243 |
+
"completions/max_length": 2048.0,
|
| 244 |
+
"completions/max_terminated_length": 2004.0,
|
| 245 |
+
"completions/mean_length": 666.3100830078125,
|
| 246 |
+
"completions/mean_terminated_length": 644.7192993164062,
|
| 247 |
+
"completions/min_length": 40.2,
|
| 248 |
+
"completions/min_terminated_length": 40.2,
|
| 249 |
+
"epoch": 0.504050405040504,
|
| 250 |
+
"grad_norm": 0.20907825231552124,
|
| 251 |
+
"learning_rate": 9.75e-07,
|
| 252 |
+
"loss": 0.0284,
|
| 253 |
+
"num_tokens": 26643040.0,
|
| 254 |
+
"reward": 1.2907849550247192,
|
| 255 |
+
"reward_std": 0.5618968069553375,
|
| 256 |
+
"rewards/accuracy_reward/mean": 0.5892857074737549,
|
| 257 |
+
"rewards/accuracy_reward/std": 0.49005323052406313,
|
| 258 |
+
"rewards/cosine_scaled_reward/mean": 0.5887125849723815,
|
| 259 |
+
"rewards/cosine_scaled_reward/std": 0.48966326117515563,
|
| 260 |
+
"rewards/format_reward/mean": 0.8142857074737548,
|
| 261 |
+
"rewards/format_reward/std": 0.37804563641548156,
|
| 262 |
+
"step": 40
|
| 263 |
+
},
|
| 264 |
+
{
|
| 265 |
+
"clip_ratio/high_max": 0.0,
|
| 266 |
+
"clip_ratio/high_mean": 0.0,
|
| 267 |
+
"clip_ratio/low_mean": 0.0,
|
| 268 |
+
"clip_ratio/low_min": 0.0,
|
| 269 |
+
"clip_ratio/region_mean": 0.0,
|
| 270 |
+
"completions/clipped_ratio": 0.011383928571428581,
|
| 271 |
+
"completions/max_length": 2048.0,
|
| 272 |
+
"completions/max_terminated_length": 1926.8,
|
| 273 |
+
"completions/mean_length": 675.0848388671875,
|
| 274 |
+
"completions/mean_terminated_length": 659.2405029296875,
|
| 275 |
+
"completions/min_length": 74.4,
|
| 276 |
+
"completions/min_terminated_length": 74.4,
|
| 277 |
+
"epoch": 0.5670567056705671,
|
| 278 |
+
"grad_norm": 0.13796494901180267,
|
| 279 |
+
"learning_rate": 9.99686774088805e-07,
|
| 280 |
+
"loss": 0.0281,
|
| 281 |
+
"num_tokens": 30549836.0,
|
| 282 |
+
"reward": 1.4507519245147704,
|
| 283 |
+
"reward_std": 0.4350401759147644,
|
| 284 |
+
"rewards/accuracy_reward/mean": 0.6517857074737549,
|
| 285 |
+
"rewards/accuracy_reward/std": 0.47552146911621096,
|
| 286 |
+
"rewards/cosine_scaled_reward/mean": 0.6508340954780578,
|
| 287 |
+
"rewards/cosine_scaled_reward/std": 0.47517149448394774,
|
| 288 |
+
"rewards/format_reward/mean": 0.9470982074737548,
|
| 289 |
+
"rewards/format_reward/std": 0.21898569464683532,
|
| 290 |
+
"step": 45
|
| 291 |
+
},
|
| 292 |
+
{
|
| 293 |
+
"clip_ratio/high_max": 0.0,
|
| 294 |
+
"clip_ratio/high_mean": 0.0,
|
| 295 |
+
"clip_ratio/low_mean": 0.0,
|
| 296 |
+
"clip_ratio/low_min": 0.0,
|
| 297 |
+
"clip_ratio/region_mean": 0.0,
|
| 298 |
+
"completions/clipped_ratio": 0.015625,
|
| 299 |
+
"completions/max_length": 2048.0,
|
| 300 |
+
"completions/max_terminated_length": 1992.2,
|
| 301 |
+
"completions/mean_length": 730.3631958007812,
|
| 302 |
+
"completions/mean_terminated_length": 709.4870239257813,
|
| 303 |
+
"completions/min_length": 188.6,
|
| 304 |
+
"completions/min_terminated_length": 188.6,
|
| 305 |
+
"epoch": 0.6300630063006301,
|
| 306 |
+
"grad_norm": 0.1085776761174202,
|
| 307 |
+
"learning_rate": 9.984149663879992e-07,
|
| 308 |
+
"loss": 0.0343,
|
| 309 |
+
"num_tokens": 34629415.0,
|
| 310 |
+
"reward": 1.4995574235916138,
|
| 311 |
+
"reward_std": 0.40187769532203677,
|
| 312 |
+
"rewards/accuracy_reward/mean": 0.6754464268684387,
|
| 313 |
+
"rewards/accuracy_reward/std": 0.4675646245479584,
|
| 314 |
+
"rewards/cosine_scaled_reward/mean": 0.6736683487892151,
|
| 315 |
+
"rewards/cosine_scaled_reward/std": 0.4671317100524902,
|
| 316 |
+
"rewards/format_reward/mean": 0.9745535850524902,
|
| 317 |
+
"rewards/format_reward/std": 0.15644133985042571,
|
| 318 |
+
"step": 50
|
| 319 |
+
},
|
| 320 |
+
{
|
| 321 |
+
"clip_ratio/high_max": 0.0,
|
| 322 |
+
"clip_ratio/high_mean": 0.0,
|
| 323 |
+
"clip_ratio/low_mean": 0.0,
|
| 324 |
+
"clip_ratio/low_min": 0.0,
|
| 325 |
+
"clip_ratio/region_mean": 0.0,
|
| 326 |
+
"completions/clipped_ratio": 0.01026785714285714,
|
| 327 |
+
"completions/max_length": 2048.0,
|
| 328 |
+
"completions/max_terminated_length": 1918.8,
|
| 329 |
+
"completions/mean_length": 718.3185546875,
|
| 330 |
+
"completions/mean_terminated_length": 704.5169189453125,
|
| 331 |
+
"completions/min_length": 187.0,
|
| 332 |
+
"completions/min_terminated_length": 187.0,
|
| 333 |
+
"epoch": 0.693069306930693,
|
| 334 |
+
"grad_norm": 0.1311427652835846,
|
| 335 |
+
"learning_rate": 9.961674879848956e-07,
|
| 336 |
+
"loss": 0.0267,
|
| 337 |
+
"num_tokens": 38633034.0,
|
| 338 |
+
"reward": 1.5256200790405274,
|
| 339 |
+
"reward_std": 0.38292278051376344,
|
| 340 |
+
"rewards/accuracy_reward/mean": 0.6888392925262451,
|
| 341 |
+
"rewards/accuracy_reward/std": 0.46227556467056274,
|
| 342 |
+
"rewards/cosine_scaled_reward/mean": 0.6876238822937012,
|
| 343 |
+
"rewards/cosine_scaled_reward/std": 0.4619812607765198,
|
| 344 |
+
"rewards/format_reward/mean": 0.985937488079071,
|
| 345 |
+
"rewards/format_reward/std": 0.11638441979885102,
|
| 346 |
+
"step": 55
|
| 347 |
+
},
|
| 348 |
+
{
|
| 349 |
+
"clip_ratio/high_max": 0.0,
|
| 350 |
+
"clip_ratio/high_mean": 0.0,
|
| 351 |
+
"clip_ratio/low_mean": 0.0,
|
| 352 |
+
"clip_ratio/low_min": 0.0,
|
| 353 |
+
"clip_ratio/region_mean": 0.0,
|
| 354 |
+
"completions/clipped_ratio": 0.01026785714285714,
|
| 355 |
+
"completions/max_length": 2048.0,
|
| 356 |
+
"completions/max_terminated_length": 1962.6,
|
| 357 |
+
"completions/mean_length": 726.8395385742188,
|
| 358 |
+
"completions/mean_terminated_length": 713.1712280273438,
|
| 359 |
+
"completions/min_length": 185.8,
|
| 360 |
+
"completions/min_terminated_length": 185.8,
|
| 361 |
+
"epoch": 0.7560756075607561,
|
| 362 |
+
"grad_norm": 0.10231342166662216,
|
| 363 |
+
"learning_rate": 9.929487384240101e-07,
|
| 364 |
+
"loss": 0.0236,
|
| 365 |
+
"num_tokens": 42664491.0,
|
| 366 |
+
"reward": 1.5491954326629638,
|
| 367 |
+
"reward_std": 0.3630220055580139,
|
| 368 |
+
"rewards/accuracy_reward/mean": 0.7042410612106323,
|
| 369 |
+
"rewards/accuracy_reward/std": 0.4564294159412384,
|
| 370 |
+
"rewards/cosine_scaled_reward/mean": 0.7012924790382385,
|
| 371 |
+
"rewards/cosine_scaled_reward/std": 0.4559535324573517,
|
| 372 |
+
"rewards/format_reward/mean": 0.9886160731315613,
|
| 373 |
+
"rewards/format_reward/std": 0.10382185429334641,
|
| 374 |
+
"step": 60
|
| 375 |
+
},
|
| 376 |
+
{
|
| 377 |
+
"clip_ratio/high_max": 0.0,
|
| 378 |
+
"clip_ratio/high_mean": 0.0,
|
| 379 |
+
"clip_ratio/low_mean": 0.0,
|
| 380 |
+
"clip_ratio/low_min": 0.0,
|
| 381 |
+
"clip_ratio/region_mean": 0.0,
|
| 382 |
+
"completions/clipped_ratio": 0.01450892857142858,
|
| 383 |
+
"completions/max_length": 2048.0,
|
| 384 |
+
"completions/max_terminated_length": 1976.2,
|
| 385 |
+
"completions/mean_length": 732.7027099609375,
|
| 386 |
+
"completions/mean_terminated_length": 713.31240234375,
|
| 387 |
+
"completions/min_length": 193.4,
|
| 388 |
+
"completions/min_terminated_length": 193.4,
|
| 389 |
+
"epoch": 0.819081908190819,
|
| 390 |
+
"grad_norm": 0.10937383025884628,
|
| 391 |
+
"learning_rate": 9.88765018558761e-07,
|
| 392 |
+
"loss": 0.0175,
|
| 393 |
+
"num_tokens": 46741551.0,
|
| 394 |
+
"reward": 1.5278964996337892,
|
| 395 |
+
"reward_std": 0.35590446591377256,
|
| 396 |
+
"rewards/accuracy_reward/mean": 0.6912946343421936,
|
| 397 |
+
"rewards/accuracy_reward/std": 0.46131653189659116,
|
| 398 |
+
"rewards/cosine_scaled_reward/mean": 0.6897214889526367,
|
| 399 |
+
"rewards/cosine_scaled_reward/std": 0.46095741391181944,
|
| 400 |
+
"rewards/format_reward/mean": 0.9834821462631226,
|
| 401 |
+
"rewards/format_reward/std": 0.12617166191339493,
|
| 402 |
+
"step": 65
|
| 403 |
+
},
|
| 404 |
+
{
|
| 405 |
+
"clip_ratio/high_max": 0.0,
|
| 406 |
+
"clip_ratio/high_mean": 0.0,
|
| 407 |
+
"clip_ratio/low_mean": 0.0,
|
| 408 |
+
"clip_ratio/low_min": 0.0,
|
| 409 |
+
"clip_ratio/region_mean": 0.0,
|
| 410 |
+
"completions/clipped_ratio": 0.00892857142857142,
|
| 411 |
+
"completions/max_length": 2048.0,
|
| 412 |
+
"completions/max_terminated_length": 1979.8,
|
| 413 |
+
"completions/mean_length": 721.2656494140625,
|
| 414 |
+
"completions/mean_terminated_length": 709.3232177734375,
|
| 415 |
+
"completions/min_length": 211.8,
|
| 416 |
+
"completions/min_terminated_length": 211.8,
|
| 417 |
+
"epoch": 0.8820882088208821,
|
| 418 |
+
"grad_norm": 0.10150983184576035,
|
| 419 |
+
"learning_rate": 9.836245182172519e-07,
|
| 420 |
+
"loss": 0.0195,
|
| 421 |
+
"num_tokens": 50729557.0,
|
| 422 |
+
"reward": 1.5760997772216796,
|
| 423 |
+
"reward_std": 0.35591434240341185,
|
| 424 |
+
"rewards/accuracy_reward/mean": 0.7227155566215515,
|
| 425 |
+
"rewards/accuracy_reward/std": 0.4474908530712128,
|
| 426 |
+
"rewards/cosine_scaled_reward/mean": 0.7200031995773315,
|
| 427 |
+
"rewards/cosine_scaled_reward/std": 0.44707669615745543,
|
| 428 |
+
"rewards/format_reward/mean": 0.990625,
|
| 429 |
+
"rewards/format_reward/std": 0.09542139172554016,
|
| 430 |
+
"step": 70
|
| 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": 0.011830357142857139,
|
| 439 |
+
"completions/max_length": 2048.0,
|
| 440 |
+
"completions/max_terminated_length": 2008.8,
|
| 441 |
+
"completions/mean_length": 731.1739135742188,
|
| 442 |
+
"completions/mean_terminated_length": 715.3806396484375,
|
| 443 |
+
"completions/min_length": 151.2,
|
| 444 |
+
"completions/min_terminated_length": 151.2,
|
| 445 |
+
"epoch": 0.9450945094509451,
|
| 446 |
+
"grad_norm": 0.10765000432729721,
|
| 447 |
+
"learning_rate": 9.775373001702997e-07,
|
| 448 |
+
"loss": 0.0213,
|
| 449 |
+
"num_tokens": 54802448.0,
|
| 450 |
+
"reward": 1.5546066045761109,
|
| 451 |
+
"reward_std": 0.32919522523880007,
|
| 452 |
+
"rewards/accuracy_reward/mean": 0.7082589507102967,
|
| 453 |
+
"rewards/accuracy_reward/std": 0.4541184425354004,
|
| 454 |
+
"rewards/cosine_scaled_reward/mean": 0.7047487974166871,
|
| 455 |
+
"rewards/cosine_scaled_reward/std": 0.45348618030548093,
|
| 456 |
+
"rewards/format_reward/mean": 0.9879464268684387,
|
| 457 |
+
"rewards/format_reward/std": 0.10762715935707093,
|
| 458 |
+
"step": 75
|
| 459 |
+
},
|
| 460 |
+
{
|
| 461 |
+
"clip_ratio/high_max": 0.0,
|
| 462 |
+
"clip_ratio/high_mean": 0.0,
|
| 463 |
+
"clip_ratio/low_mean": 0.0,
|
| 464 |
+
"clip_ratio/low_min": 0.0,
|
| 465 |
+
"clip_ratio/region_mean": 0.0,
|
| 466 |
+
"completions/clipped_ratio": 0.013587898751733695,
|
| 467 |
+
"completions/max_length": 2048.0,
|
| 468 |
+
"completions/max_terminated_length": 2006.6,
|
| 469 |
+
"completions/mean_length": 748.0619995117188,
|
| 470 |
+
"completions/mean_terminated_length": 730.115478515625,
|
| 471 |
+
"completions/min_length": 196.2,
|
| 472 |
+
"completions/min_terminated_length": 196.2,
|
| 473 |
+
"epoch": 1.0126012601260126,
|
| 474 |
+
"grad_norm": 0.10716311633586884,
|
| 475 |
+
"learning_rate": 9.705152804330871e-07,
|
| 476 |
+
"loss": 0.015,
|
| 477 |
+
"num_tokens": 58996685.0,
|
| 478 |
+
"reward": 1.5426273107528687,
|
| 479 |
+
"reward_std": 0.36739041805267336,
|
| 480 |
+
"rewards/accuracy_reward/mean": 0.7006696462631226,
|
| 481 |
+
"rewards/accuracy_reward/std": 0.45738084316253663,
|
| 482 |
+
"rewards/cosine_scaled_reward/mean": 0.6977543830871582,
|
| 483 |
+
"rewards/cosine_scaled_reward/std": 0.45684285163879396,
|
| 484 |
+
"rewards/format_reward/mean": 0.9861607193946839,
|
| 485 |
+
"rewards/format_reward/std": 0.11394104063510894,
|
| 486 |
+
"step": 80
|
| 487 |
+
},
|
| 488 |
+
{
|
| 489 |
+
"clip_ratio/high_max": 0.0,
|
| 490 |
+
"clip_ratio/high_mean": 0.0,
|
| 491 |
+
"clip_ratio/low_mean": 0.0,
|
| 492 |
+
"clip_ratio/low_min": 0.0,
|
| 493 |
+
"clip_ratio/region_mean": 0.0,
|
| 494 |
+
"completions/clipped_ratio": 0.011607142857142861,
|
| 495 |
+
"completions/max_length": 2048.0,
|
| 496 |
+
"completions/max_terminated_length": 2002.6,
|
| 497 |
+
"completions/mean_length": 723.7663330078125,
|
| 498 |
+
"completions/mean_terminated_length": 708.2584350585937,
|
| 499 |
+
"completions/min_length": 195.8,
|
| 500 |
+
"completions/min_terminated_length": 195.8,
|
| 501 |
+
"epoch": 1.0756075607560756,
|
| 502 |
+
"grad_norm": 0.11011335998773575,
|
| 503 |
+
"learning_rate": 9.625722049390045e-07,
|
| 504 |
+
"loss": 0.0214,
|
| 505 |
+
"num_tokens": 63042006.0,
|
| 506 |
+
"reward": 1.5647123813629151,
|
| 507 |
+
"reward_std": 0.3331080436706543,
|
| 508 |
+
"rewards/accuracy_reward/mean": 0.7145089268684387,
|
| 509 |
+
"rewards/accuracy_reward/std": 0.4507896304130554,
|
| 510 |
+
"rewards/cosine_scaled_reward/mean": 0.7122372031211853,
|
| 511 |
+
"rewards/cosine_scaled_reward/std": 0.4505117475986481,
|
| 512 |
+
"rewards/format_reward/mean": 0.9881696581840516,
|
| 513 |
+
"rewards/format_reward/std": 0.10608766973018646,
|
| 514 |
+
"step": 85
|
| 515 |
+
},
|
| 516 |
+
{
|
| 517 |
+
"clip_ratio/high_max": 0.0,
|
| 518 |
+
"clip_ratio/high_mean": 0.0,
|
| 519 |
+
"clip_ratio/low_mean": 0.0,
|
| 520 |
+
"clip_ratio/low_min": 0.0,
|
| 521 |
+
"clip_ratio/region_mean": 0.0,
|
| 522 |
+
"completions/clipped_ratio": 0.013839285714285698,
|
| 523 |
+
"completions/max_length": 2048.0,
|
| 524 |
+
"completions/max_terminated_length": 2014.6,
|
| 525 |
+
"completions/mean_length": 756.4529418945312,
|
| 526 |
+
"completions/mean_terminated_length": 738.1903930664063,
|
| 527 |
+
"completions/min_length": 226.4,
|
| 528 |
+
"completions/min_terminated_length": 226.4,
|
| 529 |
+
"epoch": 1.1386138613861387,
|
| 530 |
+
"grad_norm": 0.09067358821630478,
|
| 531 |
+
"learning_rate": 9.537236226313388e-07,
|
| 532 |
+
"loss": 0.0201,
|
| 533 |
+
"num_tokens": 67245803.0,
|
| 534 |
+
"reward": 1.5548530101776123,
|
| 535 |
+
"reward_std": 0.3273438990116119,
|
| 536 |
+
"rewards/accuracy_reward/mean": 0.7091517925262452,
|
| 537 |
+
"rewards/accuracy_reward/std": 0.45312303900718687,
|
| 538 |
+
"rewards/cosine_scaled_reward/mean": 0.7056880831718445,
|
| 539 |
+
"rewards/cosine_scaled_reward/std": 0.45253998041152954,
|
| 540 |
+
"rewards/format_reward/mean": 0.9857142806053162,
|
| 541 |
+
"rewards/format_reward/std": 0.11536658853292465,
|
| 542 |
+
"step": 90
|
| 543 |
+
},
|
| 544 |
+
{
|
| 545 |
+
"clip_ratio/high_max": 0.0,
|
| 546 |
+
"clip_ratio/high_mean": 0.0,
|
| 547 |
+
"clip_ratio/low_mean": 0.0,
|
| 548 |
+
"clip_ratio/low_min": 0.0,
|
| 549 |
+
"clip_ratio/region_mean": 0.0,
|
| 550 |
+
"completions/clipped_ratio": 0.009821428571428559,
|
| 551 |
+
"completions/max_length": 2048.0,
|
| 552 |
+
"completions/max_terminated_length": 2014.4,
|
| 553 |
+
"completions/mean_length": 730.8464721679687,
|
| 554 |
+
"completions/mean_terminated_length": 717.7846435546875,
|
| 555 |
+
"completions/min_length": 215.2,
|
| 556 |
+
"completions/min_terminated_length": 215.2,
|
| 557 |
+
"epoch": 1.2016201620162017,
|
| 558 |
+
"grad_norm": 0.11546584218740463,
|
| 559 |
+
"learning_rate": 9.439868550254868e-07,
|
| 560 |
+
"loss": 0.0172,
|
| 561 |
+
"num_tokens": 71285419.0,
|
| 562 |
+
"reward": 1.5703669548034669,
|
| 563 |
+
"reward_std": 0.34961515069007876,
|
| 564 |
+
"rewards/accuracy_reward/mean": 0.7183035731315612,
|
| 565 |
+
"rewards/accuracy_reward/std": 0.4488598704338074,
|
| 566 |
+
"rewards/cosine_scaled_reward/mean": 0.7148409485816956,
|
| 567 |
+
"rewards/cosine_scaled_reward/std": 0.4484779477119446,
|
| 568 |
+
"rewards/format_reward/mean": 0.9892857074737549,
|
| 569 |
+
"rewards/format_reward/std": 0.0997354432940483,
|
| 570 |
+
"step": 95
|
| 571 |
+
},
|
| 572 |
+
{
|
| 573 |
+
"clip_ratio/high_max": 0.0,
|
| 574 |
+
"clip_ratio/high_mean": 0.0,
|
| 575 |
+
"clip_ratio/low_mean": 0.0,
|
| 576 |
+
"clip_ratio/low_min": 0.0,
|
| 577 |
+
"clip_ratio/region_mean": 0.0,
|
| 578 |
+
"completions/clipped_ratio": 0.007366071428571419,
|
| 579 |
+
"completions/max_length": 2048.0,
|
| 580 |
+
"completions/max_terminated_length": 1957.2,
|
| 581 |
+
"completions/mean_length": 717.281494140625,
|
| 582 |
+
"completions/mean_terminated_length": 707.4199096679688,
|
| 583 |
+
"completions/min_length": 215.6,
|
| 584 |
+
"completions/min_terminated_length": 215.6,
|
| 585 |
+
"epoch": 1.2646264626462647,
|
| 586 |
+
"grad_norm": 0.10343993455171585,
|
| 587 |
+
"learning_rate": 9.333809623012763e-07,
|
| 588 |
+
"loss": 0.0156,
|
| 589 |
+
"num_tokens": 75264080.0,
|
| 590 |
+
"reward": 1.6102480173110962,
|
| 591 |
+
"reward_std": 0.30375452637672423,
|
| 592 |
+
"rewards/accuracy_reward/mean": 0.74375,
|
| 593 |
+
"rewards/accuracy_reward/std": 0.43397698402404783,
|
| 594 |
+
"rewards/cosine_scaled_reward/mean": 0.7412548184394836,
|
| 595 |
+
"rewards/cosine_scaled_reward/std": 0.4337835073471069,
|
| 596 |
+
"rewards/format_reward/mean": 0.9917410731315612,
|
| 597 |
+
"rewards/format_reward/std": 0.09024428129196167,
|
| 598 |
+
"step": 100
|
| 599 |
+
},
|
| 600 |
+
{
|
| 601 |
+
"clip_ratio/high_max": 0.0,
|
| 602 |
+
"clip_ratio/high_mean": 0.0,
|
| 603 |
+
"clip_ratio/low_mean": 0.0,
|
| 604 |
+
"clip_ratio/low_min": 0.0,
|
| 605 |
+
"clip_ratio/region_mean": 0.0,
|
| 606 |
+
"completions/clipped_ratio": 0.009821428571428537,
|
| 607 |
+
"completions/max_length": 2048.0,
|
| 608 |
+
"completions/max_terminated_length": 2011.0,
|
| 609 |
+
"completions/mean_length": 733.489111328125,
|
| 610 |
+
"completions/mean_terminated_length": 720.5335571289063,
|
| 611 |
+
"completions/min_length": 193.0,
|
| 612 |
+
"completions/min_terminated_length": 193.0,
|
| 613 |
+
"epoch": 1.3276327632763276,
|
| 614 |
+
"grad_norm": 0.1060086041688919,
|
| 615 |
+
"learning_rate": 9.219267059917681e-07,
|
| 616 |
+
"loss": 0.0199,
|
| 617 |
+
"num_tokens": 79353743.0,
|
| 618 |
+
"reward": 1.5742143154144288,
|
| 619 |
+
"reward_std": 0.3207855045795441,
|
| 620 |
+
"rewards/accuracy_reward/mean": 0.7203125,
|
| 621 |
+
"rewards/accuracy_reward/std": 0.44847235083580017,
|
| 622 |
+
"rewards/cosine_scaled_reward/mean": 0.7178481221199036,
|
| 623 |
+
"rewards/cosine_scaled_reward/std": 0.44811044335365297,
|
| 624 |
+
"rewards/format_reward/mean": 0.9899553537368775,
|
| 625 |
+
"rewards/format_reward/std": 0.09872472882270814,
|
| 626 |
+
"step": 105
|
| 627 |
+
},
|
| 628 |
+
{
|
| 629 |
+
"clip_ratio/high_max": 0.0,
|
| 630 |
+
"clip_ratio/high_mean": 0.0,
|
| 631 |
+
"clip_ratio/low_mean": 0.0,
|
| 632 |
+
"clip_ratio/low_min": 0.0,
|
| 633 |
+
"clip_ratio/region_mean": 0.0,
|
| 634 |
+
"completions/clipped_ratio": 0.006026785714285721,
|
| 635 |
+
"completions/max_length": 2048.0,
|
| 636 |
+
"completions/max_terminated_length": 1975.8,
|
| 637 |
+
"completions/mean_length": 731.3013671875,
|
| 638 |
+
"completions/mean_terminated_length": 723.3466918945312,
|
| 639 |
+
"completions/min_length": 218.6,
|
| 640 |
+
"completions/min_terminated_length": 218.6,
|
| 641 |
+
"epoch": 1.3906390639063906,
|
| 642 |
+
"grad_norm": 0.12040749192237854,
|
| 643 |
+
"learning_rate": 9.096465083415807e-07,
|
| 644 |
+
"loss": 0.0203,
|
| 645 |
+
"num_tokens": 83426645.0,
|
| 646 |
+
"reward": 1.6153923511505126,
|
| 647 |
+
"reward_std": 0.3091603219509125,
|
| 648 |
+
"rewards/accuracy_reward/mean": 0.7464285731315613,
|
| 649 |
+
"rewards/accuracy_reward/std": 0.43417892456054685,
|
| 650 |
+
"rewards/cosine_scaled_reward/mean": 0.7441774845123291,
|
| 651 |
+
"rewards/cosine_scaled_reward/std": 0.43402957916259766,
|
| 652 |
+
"rewards/format_reward/mean": 0.99375,
|
| 653 |
+
"rewards/format_reward/std": 0.07496543601155281,
|
| 654 |
+
"step": 110
|
| 655 |
+
},
|
| 656 |
+
{
|
| 657 |
+
"clip_ratio/high_max": 0.0,
|
| 658 |
+
"clip_ratio/high_mean": 0.0,
|
| 659 |
+
"clip_ratio/low_mean": 0.0,
|
| 660 |
+
"clip_ratio/low_min": 0.0,
|
| 661 |
+
"clip_ratio/region_mean": 0.0,
|
| 662 |
+
"completions/clipped_ratio": 0.012053571428571441,
|
| 663 |
+
"completions/max_length": 2048.0,
|
| 664 |
+
"completions/max_terminated_length": 1982.8,
|
| 665 |
+
"completions/mean_length": 744.0600830078125,
|
| 666 |
+
"completions/mean_terminated_length": 728.1871215820313,
|
| 667 |
+
"completions/min_length": 205.4,
|
| 668 |
+
"completions/min_terminated_length": 205.4,
|
| 669 |
+
"epoch": 1.4536453645364538,
|
| 670 |
+
"grad_norm": 0.12004546821117401,
|
| 671 |
+
"learning_rate": 8.96564408414293e-07,
|
| 672 |
+
"loss": 0.0196,
|
| 673 |
+
"num_tokens": 87563834.0,
|
| 674 |
+
"reward": 1.5543434858322143,
|
| 675 |
+
"reward_std": 0.31985793113708494,
|
| 676 |
+
"rewards/accuracy_reward/mean": 0.7119892239570618,
|
| 677 |
+
"rewards/accuracy_reward/std": 0.45167325139045716,
|
| 678 |
+
"rewards/cosine_scaled_reward/mean": 0.7037603855133057,
|
| 679 |
+
"rewards/cosine_scaled_reward/std": 0.45347819328308103,
|
| 680 |
+
"rewards/format_reward/mean": 0.987500011920929,
|
| 681 |
+
"rewards/format_reward/std": 0.10970796942710877,
|
| 682 |
+
"step": 115
|
| 683 |
+
},
|
| 684 |
+
{
|
| 685 |
+
"clip_ratio/high_max": 0.0,
|
| 686 |
+
"clip_ratio/high_mean": 0.0,
|
| 687 |
+
"clip_ratio/low_mean": 0.0,
|
| 688 |
+
"clip_ratio/low_min": 0.0,
|
| 689 |
+
"clip_ratio/region_mean": 0.0,
|
| 690 |
+
"completions/clipped_ratio": 0.00714285714285714,
|
| 691 |
+
"completions/max_length": 2048.0,
|
| 692 |
+
"completions/max_terminated_length": 2009.0,
|
| 693 |
+
"completions/mean_length": 722.8154296875,
|
| 694 |
+
"completions/mean_terminated_length": 713.2384399414062,
|
| 695 |
+
"completions/min_length": 201.4,
|
| 696 |
+
"completions/min_terminated_length": 201.4,
|
| 697 |
+
"epoch": 1.5166516651665165,
|
| 698 |
+
"grad_norm": 0.09482675045728683,
|
| 699 |
+
"learning_rate": 8.827060150348489e-07,
|
| 700 |
+
"loss": 0.0093,
|
| 701 |
+
"num_tokens": 91570751.0,
|
| 702 |
+
"reward": 1.604229950904846,
|
| 703 |
+
"reward_std": 0.2770054042339325,
|
| 704 |
+
"rewards/accuracy_reward/mean": 0.7392857074737549,
|
| 705 |
+
"rewards/accuracy_reward/std": 0.438806426525116,
|
| 706 |
+
"rewards/cosine_scaled_reward/mean": 0.7370312094688416,
|
| 707 |
+
"rewards/cosine_scaled_reward/std": 0.43855146169662473,
|
| 708 |
+
"rewards/format_reward/mean": 0.9928571343421936,
|
| 709 |
+
"rewards/format_reward/std": 0.08236530497670173,
|
| 710 |
+
"step": 120
|
| 711 |
+
},
|
| 712 |
+
{
|
| 713 |
+
"clip_ratio/high_max": 0.0,
|
| 714 |
+
"clip_ratio/high_mean": 0.0,
|
| 715 |
+
"clip_ratio/low_mean": 0.0,
|
| 716 |
+
"clip_ratio/low_min": 0.0,
|
| 717 |
+
"clip_ratio/region_mean": 0.0,
|
| 718 |
+
"completions/clipped_ratio": 0.012053571428571441,
|
| 719 |
+
"completions/max_length": 2048.0,
|
| 720 |
+
"completions/max_terminated_length": 2035.0,
|
| 721 |
+
"completions/mean_length": 761.4022583007812,
|
| 722 |
+
"completions/mean_terminated_length": 745.6811889648437,
|
| 723 |
+
"completions/min_length": 196.4,
|
| 724 |
+
"completions/min_terminated_length": 196.4,
|
| 725 |
+
"epoch": 1.5796579657965797,
|
| 726 |
+
"grad_norm": 0.10029105842113495,
|
| 727 |
+
"learning_rate": 8.680984566590792e-07,
|
| 728 |
+
"loss": 0.0132,
|
| 729 |
+
"num_tokens": 95783545.0,
|
| 730 |
+
"reward": 1.5971355199813844,
|
| 731 |
+
"reward_std": 0.28857868909835815,
|
| 732 |
+
"rewards/accuracy_reward/mean": 0.7368303656578064,
|
| 733 |
+
"rewards/accuracy_reward/std": 0.43986259698867797,
|
| 734 |
+
"rewards/cosine_scaled_reward/mean": 0.7335565090179443,
|
| 735 |
+
"rewards/cosine_scaled_reward/std": 0.4395099520683289,
|
| 736 |
+
"rewards/format_reward/mean": 0.9870535731315613,
|
| 737 |
+
"rewards/format_reward/std": 0.11282701939344406,
|
| 738 |
+
"step": 125
|
| 739 |
+
},
|
| 740 |
+
{
|
| 741 |
+
"clip_ratio/high_max": 0.0,
|
| 742 |
+
"clip_ratio/high_mean": 0.0,
|
| 743 |
+
"clip_ratio/low_mean": 0.0,
|
| 744 |
+
"clip_ratio/low_min": 0.0,
|
| 745 |
+
"clip_ratio/region_mean": 0.0,
|
| 746 |
+
"completions/clipped_ratio": 0.010267857142857117,
|
| 747 |
+
"completions/max_length": 2048.0,
|
| 748 |
+
"completions/max_terminated_length": 1965.6,
|
| 749 |
+
"completions/mean_length": 726.2194580078125,
|
| 750 |
+
"completions/mean_terminated_length": 712.602392578125,
|
| 751 |
+
"completions/min_length": 215.4,
|
| 752 |
+
"completions/min_terminated_length": 215.4,
|
| 753 |
+
"epoch": 1.6426642664266426,
|
| 754 |
+
"grad_norm": 0.11565880477428436,
|
| 755 |
+
"learning_rate": 8.527703282684765e-07,
|
| 756 |
+
"loss": 0.0187,
|
| 757 |
+
"num_tokens": 99800320.0,
|
| 758 |
+
"reward": 1.6029990196228028,
|
| 759 |
+
"reward_std": 0.29794790744781496,
|
| 760 |
+
"rewards/accuracy_reward/mean": 0.7397321462631226,
|
| 761 |
+
"rewards/accuracy_reward/std": 0.4377941250801086,
|
| 762 |
+
"rewards/cosine_scaled_reward/mean": 0.7374711394309997,
|
| 763 |
+
"rewards/cosine_scaled_reward/std": 0.43750142455101015,
|
| 764 |
+
"rewards/format_reward/mean": 0.989062488079071,
|
| 765 |
+
"rewards/format_reward/std": 0.10165355652570725,
|
| 766 |
+
"step": 130
|
| 767 |
+
},
|
| 768 |
+
{
|
| 769 |
+
"clip_ratio/high_max": 0.0,
|
| 770 |
+
"clip_ratio/high_mean": 0.0,
|
| 771 |
+
"clip_ratio/low_mean": 0.0,
|
| 772 |
+
"clip_ratio/low_min": 0.0,
|
| 773 |
+
"clip_ratio/region_mean": 0.0,
|
| 774 |
+
"completions/clipped_ratio": 0.006696428571428559,
|
| 775 |
+
"completions/max_length": 2048.0,
|
| 776 |
+
"completions/max_terminated_length": 2003.0,
|
| 777 |
+
"completions/mean_length": 733.4511474609375,
|
| 778 |
+
"completions/mean_terminated_length": 724.589501953125,
|
| 779 |
+
"completions/min_length": 170.4,
|
| 780 |
+
"completions/min_terminated_length": 170.4,
|
| 781 |
+
"epoch": 1.7056705670567056,
|
| 782 |
+
"grad_norm": 0.1487206071615219,
|
| 783 |
+
"learning_rate": 8.36751635394175e-07,
|
| 784 |
+
"loss": 0.0119,
|
| 785 |
+
"num_tokens": 103861941.0,
|
| 786 |
+
"reward": 1.6143394470214845,
|
| 787 |
+
"reward_std": 0.2868204951286316,
|
| 788 |
+
"rewards/accuracy_reward/mean": 0.746875,
|
| 789 |
+
"rewards/accuracy_reward/std": 0.43289305567741393,
|
| 790 |
+
"rewards/cosine_scaled_reward/mean": 0.7434109926223755,
|
| 791 |
+
"rewards/cosine_scaled_reward/std": 0.432673978805542,
|
| 792 |
+
"rewards/format_reward/mean": 0.9915178656578064,
|
| 793 |
+
"rewards/format_reward/std": 0.08637134805321693,
|
| 794 |
+
"step": 135
|
| 795 |
+
},
|
| 796 |
+
{
|
| 797 |
+
"clip_ratio/high_max": 0.0,
|
| 798 |
+
"clip_ratio/high_mean": 0.0,
|
| 799 |
+
"clip_ratio/low_mean": 0.0,
|
| 800 |
+
"clip_ratio/low_min": 0.0,
|
| 801 |
+
"clip_ratio/region_mean": 0.0,
|
| 802 |
+
"completions/clipped_ratio": 0.0049107142857142795,
|
| 803 |
+
"completions/max_length": 2048.0,
|
| 804 |
+
"completions/max_terminated_length": 1965.6,
|
| 805 |
+
"completions/mean_length": 711.4899780273438,
|
| 806 |
+
"completions/mean_terminated_length": 704.9090209960938,
|
| 807 |
+
"completions/min_length": 210.0,
|
| 808 |
+
"completions/min_terminated_length": 210.0,
|
| 809 |
+
"epoch": 1.7686768676867688,
|
| 810 |
+
"grad_norm": 0.11649075895547867,
|
| 811 |
+
"learning_rate": 8.200737353797148e-07,
|
| 812 |
+
"loss": 0.0127,
|
| 813 |
+
"num_tokens": 107860712.0,
|
| 814 |
+
"reward": 1.6414124965667725,
|
| 815 |
+
"reward_std": 0.27274273335933685,
|
| 816 |
+
"rewards/accuracy_reward/mean": 0.7636160612106323,
|
| 817 |
+
"rewards/accuracy_reward/std": 0.42368212938308714,
|
| 818 |
+
"rewards/cosine_scaled_reward/mean": 0.7611732006072998,
|
| 819 |
+
"rewards/cosine_scaled_reward/std": 0.4236971378326416,
|
| 820 |
+
"rewards/format_reward/mean": 0.9944196462631225,
|
| 821 |
+
"rewards/format_reward/std": 0.07310284897685052,
|
| 822 |
+
"step": 140
|
| 823 |
+
},
|
| 824 |
+
{
|
| 825 |
+
"clip_ratio/high_max": 0.0,
|
| 826 |
+
"clip_ratio/high_mean": 0.0,
|
| 827 |
+
"clip_ratio/low_mean": 0.0,
|
| 828 |
+
"clip_ratio/low_min": 0.0,
|
| 829 |
+
"clip_ratio/region_mean": 0.0,
|
| 830 |
+
"completions/clipped_ratio": 0.00758928571428572,
|
| 831 |
+
"completions/max_length": 2048.0,
|
| 832 |
+
"completions/max_terminated_length": 1938.4,
|
| 833 |
+
"completions/mean_length": 727.2859619140625,
|
| 834 |
+
"completions/mean_terminated_length": 717.28701171875,
|
| 835 |
+
"completions/min_length": 227.4,
|
| 836 |
+
"completions/min_terminated_length": 227.4,
|
| 837 |
+
"epoch": 1.8316831683168315,
|
| 838 |
+
"grad_norm": 0.1159307211637497,
|
| 839 |
+
"learning_rate": 8.027692759975693e-07,
|
| 840 |
+
"loss": 0.0154,
|
| 841 |
+
"num_tokens": 111983657.0,
|
| 842 |
+
"reward": 1.5885592937469482,
|
| 843 |
+
"reward_std": 0.29645509719848634,
|
| 844 |
+
"rewards/accuracy_reward/mean": 0.7290178418159485,
|
| 845 |
+
"rewards/accuracy_reward/std": 0.4415256381034851,
|
| 846 |
+
"rewards/cosine_scaled_reward/mean": 0.7266719102859497,
|
| 847 |
+
"rewards/cosine_scaled_reward/std": 0.4412645697593689,
|
| 848 |
+
"rewards/format_reward/mean": 0.9924107193946838,
|
| 849 |
+
"rewards/format_reward/std": 0.0823630191385746,
|
| 850 |
+
"step": 145
|
| 851 |
+
},
|
| 852 |
+
{
|
| 853 |
+
"clip_ratio/high_max": 0.0,
|
| 854 |
+
"clip_ratio/high_mean": 0.0,
|
| 855 |
+
"clip_ratio/low_mean": 0.0,
|
| 856 |
+
"clip_ratio/low_min": 0.0,
|
| 857 |
+
"clip_ratio/region_mean": 0.0,
|
| 858 |
+
"completions/clipped_ratio": 0.00758928571428572,
|
| 859 |
+
"completions/max_length": 2048.0,
|
| 860 |
+
"completions/max_terminated_length": 1896.8,
|
| 861 |
+
"completions/mean_length": 724.03017578125,
|
| 862 |
+
"completions/mean_terminated_length": 713.9320678710938,
|
| 863 |
+
"completions/min_length": 218.2,
|
| 864 |
+
"completions/min_terminated_length": 218.2,
|
| 865 |
+
"epoch": 1.8946894689468947,
|
| 866 |
+
"grad_norm": 0.13099786639213562,
|
| 867 |
+
"learning_rate": 7.848721315395974e-07,
|
| 868 |
+
"loss": 0.0131,
|
| 869 |
+
"num_tokens": 116046080.0,
|
| 870 |
+
"reward": 1.6239006757736205,
|
| 871 |
+
"reward_std": 0.29182094931602476,
|
| 872 |
+
"rewards/accuracy_reward/mean": 0.7533482193946839,
|
| 873 |
+
"rewards/accuracy_reward/std": 0.4308763206005096,
|
| 874 |
+
"rewards/cosine_scaled_reward/mean": 0.7493637323379516,
|
| 875 |
+
"rewards/cosine_scaled_reward/std": 0.43060003519058226,
|
| 876 |
+
"rewards/format_reward/mean": 0.9917410731315612,
|
| 877 |
+
"rewards/format_reward/std": 0.08932423293590545,
|
| 878 |
+
"step": 150
|
| 879 |
+
},
|
| 880 |
+
{
|
| 881 |
+
"clip_ratio/high_max": 0.0,
|
| 882 |
+
"clip_ratio/high_mean": 0.0,
|
| 883 |
+
"clip_ratio/low_mean": 0.0,
|
| 884 |
+
"clip_ratio/low_min": 0.0,
|
| 885 |
+
"clip_ratio/region_mean": 0.0,
|
| 886 |
+
"completions/clipped_ratio": 0.0046875,
|
| 887 |
+
"completions/max_length": 2048.0,
|
| 888 |
+
"completions/max_terminated_length": 1962.8,
|
| 889 |
+
"completions/mean_length": 708.4319702148438,
|
| 890 |
+
"completions/mean_terminated_length": 702.11376953125,
|
| 891 |
+
"completions/min_length": 197.4,
|
| 892 |
+
"completions/min_terminated_length": 197.4,
|
| 893 |
+
"epoch": 1.9576957695769577,
|
| 894 |
+
"grad_norm": 0.1514667421579361,
|
| 895 |
+
"learning_rate": 7.664173365065265e-07,
|
| 896 |
+
"loss": 0.0115,
|
| 897 |
+
"num_tokens": 120009631.0,
|
| 898 |
+
"reward": 1.6182810306549071,
|
| 899 |
+
"reward_std": 0.2975043416023254,
|
| 900 |
+
"rewards/accuracy_reward/mean": 0.7477678537368775,
|
| 901 |
+
"rewards/accuracy_reward/std": 0.4321995139122009,
|
| 902 |
+
"rewards/cosine_scaled_reward/mean": 0.7457136750221253,
|
| 903 |
+
"rewards/cosine_scaled_reward/std": 0.4320578873157501,
|
| 904 |
+
"rewards/format_reward/mean": 0.995312488079071,
|
| 905 |
+
"rewards/format_reward/std": 0.0615858756005764,
|
| 906 |
+
"step": 155
|
| 907 |
+
},
|
| 908 |
+
{
|
| 909 |
+
"clip_ratio/high_max": 0.0,
|
| 910 |
+
"clip_ratio/high_mean": 0.0,
|
| 911 |
+
"clip_ratio/low_mean": 0.0,
|
| 912 |
+
"clip_ratio/low_min": 0.0,
|
| 913 |
+
"clip_ratio/region_mean": 0.0,
|
| 914 |
+
"completions/clipped_ratio": 0.0064255374479889275,
|
| 915 |
+
"completions/max_length": 2048.0,
|
| 916 |
+
"completions/max_terminated_length": 1967.2,
|
| 917 |
+
"completions/mean_length": 701.7818725585937,
|
| 918 |
+
"completions/mean_terminated_length": 693.0792114257813,
|
| 919 |
+
"completions/min_length": 189.6,
|
| 920 |
+
"completions/min_terminated_length": 189.6,
|
| 921 |
+
"epoch": 2.025202520252025,
|
| 922 |
+
"grad_norm": 0.1373867690563202,
|
| 923 |
+
"learning_rate": 7.474410170262718e-07,
|
| 924 |
+
"loss": 0.0105,
|
| 925 |
+
"num_tokens": 123923226.0,
|
| 926 |
+
"reward": 1.6074599981307984,
|
| 927 |
+
"reward_std": 0.28035031259059906,
|
| 928 |
+
"rewards/accuracy_reward/mean": 0.7415178656578064,
|
| 929 |
+
"rewards/accuracy_reward/std": 0.43680434226989745,
|
| 930 |
+
"rewards/cosine_scaled_reward/mean": 0.7381341576576232,
|
| 931 |
+
"rewards/cosine_scaled_reward/std": 0.43655965924263,
|
| 932 |
+
"rewards/format_reward/mean": 0.99375,
|
| 933 |
+
"rewards/format_reward/std": 0.07582019194960594,
|
| 934 |
+
"step": 160
|
| 935 |
+
},
|
| 936 |
+
{
|
| 937 |
+
"clip_ratio/high_max": 0.0,
|
| 938 |
+
"clip_ratio/high_mean": 0.0,
|
| 939 |
+
"clip_ratio/low_mean": 0.0,
|
| 940 |
+
"clip_ratio/low_min": 0.0,
|
| 941 |
+
"clip_ratio/region_mean": 0.0,
|
| 942 |
+
"completions/clipped_ratio": 0.005357142857142861,
|
| 943 |
+
"completions/max_length": 2048.0,
|
| 944 |
+
"completions/max_terminated_length": 1936.6,
|
| 945 |
+
"completions/mean_length": 696.6839599609375,
|
| 946 |
+
"completions/mean_terminated_length": 689.4219970703125,
|
| 947 |
+
"completions/min_length": 212.8,
|
| 948 |
+
"completions/min_terminated_length": 212.8,
|
| 949 |
+
"epoch": 2.0882088208820884,
|
| 950 |
+
"grad_norm": 0.13395772874355316,
|
| 951 |
+
"learning_rate": 7.279803201353449e-07,
|
| 952 |
+
"loss": 0.0106,
|
| 953 |
+
"num_tokens": 127847306.0,
|
| 954 |
+
"reward": 1.6294495105743407,
|
| 955 |
+
"reward_std": 0.2632309556007385,
|
| 956 |
+
"rewards/accuracy_reward/mean": 0.7558035731315613,
|
| 957 |
+
"rewards/accuracy_reward/std": 0.4289921045303345,
|
| 958 |
+
"rewards/cosine_scaled_reward/mean": 0.7535416603088378,
|
| 959 |
+
"rewards/cosine_scaled_reward/std": 0.4288810133934021,
|
| 960 |
+
"rewards/format_reward/mean": 0.993749988079071,
|
| 961 |
+
"rewards/format_reward/std": 0.07680554389953613,
|
| 962 |
+
"step": 165
|
| 963 |
+
},
|
| 964 |
+
{
|
| 965 |
+
"clip_ratio/high_max": 0.0,
|
| 966 |
+
"clip_ratio/high_mean": 0.0,
|
| 967 |
+
"clip_ratio/low_mean": 0.0,
|
| 968 |
+
"clip_ratio/low_min": 0.0,
|
| 969 |
+
"clip_ratio/region_mean": 0.0,
|
| 970 |
+
"completions/clipped_ratio": 0.003571428571428581,
|
| 971 |
+
"completions/max_length": 2026.2,
|
| 972 |
+
"completions/max_terminated_length": 1965.2,
|
| 973 |
+
"completions/mean_length": 725.4308471679688,
|
| 974 |
+
"completions/mean_terminated_length": 720.7519287109375,
|
| 975 |
+
"completions/min_length": 225.8,
|
| 976 |
+
"completions/min_terminated_length": 225.8,
|
| 977 |
+
"epoch": 2.151215121512151,
|
| 978 |
+
"grad_norm": 0.14027050137519836,
|
| 979 |
+
"learning_rate": 7.080733410617851e-07,
|
| 980 |
+
"loss": 0.0119,
|
| 981 |
+
"num_tokens": 131877732.0,
|
| 982 |
+
"reward": 1.6384329319000244,
|
| 983 |
+
"reward_std": 0.2705275148153305,
|
| 984 |
+
"rewards/accuracy_reward/mean": 0.7616071462631225,
|
| 985 |
+
"rewards/accuracy_reward/std": 0.42367841601371764,
|
| 986 |
+
"rewards/cosine_scaled_reward/mean": 0.7578925013542175,
|
| 987 |
+
"rewards/cosine_scaled_reward/std": 0.4236116588115692,
|
| 988 |
+
"rewards/format_reward/mean": 0.9957589268684387,
|
| 989 |
+
"rewards/format_reward/std": 0.057058504223823546,
|
| 990 |
+
"step": 170
|
| 991 |
+
},
|
| 992 |
+
{
|
| 993 |
+
"clip_ratio/high_max": 0.0,
|
| 994 |
+
"clip_ratio/high_mean": 0.0,
|
| 995 |
+
"clip_ratio/low_mean": 0.0,
|
| 996 |
+
"clip_ratio/low_min": 0.0,
|
| 997 |
+
"clip_ratio/region_mean": 0.0,
|
| 998 |
+
"completions/clipped_ratio": 0.007366071428571441,
|
| 999 |
+
"completions/max_length": 2048.0,
|
| 1000 |
+
"completions/max_terminated_length": 1993.6,
|
| 1001 |
+
"completions/mean_length": 743.7538330078125,
|
| 1002 |
+
"completions/mean_terminated_length": 734.13173828125,
|
| 1003 |
+
"completions/min_length": 226.4,
|
| 1004 |
+
"completions/min_terminated_length": 226.4,
|
| 1005 |
+
"epoch": 2.2142214221422143,
|
| 1006 |
+
"grad_norm": 0.14795565605163574,
|
| 1007 |
+
"learning_rate": 6.87759048651962e-07,
|
| 1008 |
+
"loss": 0.0141,
|
| 1009 |
+
"num_tokens": 135973037.0,
|
| 1010 |
+
"reward": 1.5985393047332763,
|
| 1011 |
+
"reward_std": 0.2923765182495117,
|
| 1012 |
+
"rewards/accuracy_reward/mean": 0.7366071224212647,
|
| 1013 |
+
"rewards/accuracy_reward/std": 0.4388765335083008,
|
| 1014 |
+
"rewards/cosine_scaled_reward/mean": 0.7316766262054444,
|
| 1015 |
+
"rewards/cosine_scaled_reward/std": 0.43860033750534055,
|
| 1016 |
+
"rewards/format_reward/mean": 0.9921875,
|
| 1017 |
+
"rewards/format_reward/std": 0.08475830480456352,
|
| 1018 |
+
"step": 175
|
| 1019 |
+
},
|
| 1020 |
+
{
|
| 1021 |
+
"clip_ratio/high_max": 0.0,
|
| 1022 |
+
"clip_ratio/high_mean": 0.0,
|
| 1023 |
+
"clip_ratio/low_mean": 0.0,
|
| 1024 |
+
"clip_ratio/low_min": 0.0,
|
| 1025 |
+
"clip_ratio/region_mean": 0.0,
|
| 1026 |
+
"completions/clipped_ratio": 0.005357142857142861,
|
| 1027 |
+
"completions/max_length": 2018.0,
|
| 1028 |
+
"completions/max_terminated_length": 1964.4,
|
| 1029 |
+
"completions/mean_length": 717.5312866210937,
|
| 1030 |
+
"completions/mean_terminated_length": 710.5458862304688,
|
| 1031 |
+
"completions/min_length": 207.2,
|
| 1032 |
+
"completions/min_terminated_length": 207.2,
|
| 1033 |
+
"epoch": 2.2772277227722775,
|
| 1034 |
+
"grad_norm": 0.12406405061483383,
|
| 1035 |
+
"learning_rate": 6.670772090872278e-07,
|
| 1036 |
+
"loss": 0.0113,
|
| 1037 |
+
"num_tokens": 139942497.0,
|
| 1038 |
+
"reward": 1.6446252107620238,
|
| 1039 |
+
"reward_std": 0.26751722395420074,
|
| 1040 |
+
"rewards/accuracy_reward/mean": 0.7672940850257873,
|
| 1041 |
+
"rewards/accuracy_reward/std": 0.4188421845436096,
|
| 1042 |
+
"rewards/cosine_scaled_reward/mean": 0.7645792007446289,
|
| 1043 |
+
"rewards/cosine_scaled_reward/std": 0.4186910569667816,
|
| 1044 |
+
"rewards/format_reward/mean": 0.9944196343421936,
|
| 1045 |
+
"rewards/format_reward/std": 0.06284433305263519,
|
| 1046 |
+
"step": 180
|
| 1047 |
+
},
|
| 1048 |
+
{
|
| 1049 |
+
"clip_ratio/high_max": 0.0,
|
| 1050 |
+
"clip_ratio/high_mean": 0.0,
|
| 1051 |
+
"clip_ratio/low_mean": 0.0,
|
| 1052 |
+
"clip_ratio/low_min": 0.0,
|
| 1053 |
+
"clip_ratio/region_mean": 0.0,
|
| 1054 |
+
"completions/clipped_ratio": 0.003125,
|
| 1055 |
+
"completions/max_length": 2048.0,
|
| 1056 |
+
"completions/max_terminated_length": 1955.4,
|
| 1057 |
+
"completions/mean_length": 712.99423828125,
|
| 1058 |
+
"completions/mean_terminated_length": 708.8283447265625,
|
| 1059 |
+
"completions/min_length": 204.6,
|
| 1060 |
+
"completions/min_terminated_length": 204.6,
|
| 1061 |
+
"epoch": 2.34023402340234,
|
| 1062 |
+
"grad_norm": 0.1320047676563263,
|
| 1063 |
+
"learning_rate": 6.460683080397508e-07,
|
| 1064 |
+
"loss": 0.0134,
|
| 1065 |
+
"num_tokens": 143991319.0,
|
| 1066 |
+
"reward": 1.6394832611083985,
|
| 1067 |
+
"reward_std": 0.27438428401947024,
|
| 1068 |
+
"rewards/accuracy_reward/mean": 0.7629826188087463,
|
| 1069 |
+
"rewards/accuracy_reward/std": 0.4248282194137573,
|
| 1070 |
+
"rewards/cosine_scaled_reward/mean": 0.7604665756225586,
|
| 1071 |
+
"rewards/cosine_scaled_reward/std": 0.42470039129257203,
|
| 1072 |
+
"rewards/format_reward/mean": 0.9966517806053161,
|
| 1073 |
+
"rewards/format_reward/std": 0.05458408519625664,
|
| 1074 |
+
"step": 185
|
| 1075 |
+
},
|
| 1076 |
+
{
|
| 1077 |
+
"clip_ratio/high_max": 0.0,
|
| 1078 |
+
"clip_ratio/high_mean": 0.0,
|
| 1079 |
+
"clip_ratio/low_mean": 0.0,
|
| 1080 |
+
"clip_ratio/low_min": 0.0,
|
| 1081 |
+
"clip_ratio/region_mean": 0.0,
|
| 1082 |
+
"completions/clipped_ratio": 0.0049107142857142795,
|
| 1083 |
+
"completions/max_length": 2048.0,
|
| 1084 |
+
"completions/max_terminated_length": 2016.4,
|
| 1085 |
+
"completions/mean_length": 693.4261596679687,
|
| 1086 |
+
"completions/mean_terminated_length": 686.7283447265625,
|
| 1087 |
+
"completions/min_length": 220.8,
|
| 1088 |
+
"completions/min_terminated_length": 220.8,
|
| 1089 |
+
"epoch": 2.4032403240324034,
|
| 1090 |
+
"grad_norm": 0.14782555401325226,
|
| 1091 |
+
"learning_rate": 6.247734714199105e-07,
|
| 1092 |
+
"loss": 0.0151,
|
| 1093 |
+
"num_tokens": 147901684.0,
|
| 1094 |
+
"reward": 1.6312123775482177,
|
| 1095 |
+
"reward_std": 0.26806690990924836,
|
| 1096 |
+
"rewards/accuracy_reward/mean": 0.7569196462631226,
|
| 1097 |
+
"rewards/accuracy_reward/std": 0.428833419084549,
|
| 1098 |
+
"rewards/cosine_scaled_reward/mean": 0.7550585746765137,
|
| 1099 |
+
"rewards/cosine_scaled_reward/std": 0.42870454788208007,
|
| 1100 |
+
"rewards/format_reward/mean": 0.9935267925262451,
|
| 1101 |
+
"rewards/format_reward/std": 0.07879414483904838,
|
| 1102 |
+
"step": 190
|
| 1103 |
+
},
|
| 1104 |
+
{
|
| 1105 |
+
"clip_ratio/high_max": 0.0,
|
| 1106 |
+
"clip_ratio/high_mean": 0.0,
|
| 1107 |
+
"clip_ratio/low_mean": 0.0,
|
| 1108 |
+
"clip_ratio/low_min": 0.0,
|
| 1109 |
+
"clip_ratio/region_mean": 0.0,
|
| 1110 |
+
"completions/clipped_ratio": 0.005803571428571441,
|
| 1111 |
+
"completions/max_length": 2048.0,
|
| 1112 |
+
"completions/max_terminated_length": 1876.6,
|
| 1113 |
+
"completions/mean_length": 711.0444580078125,
|
| 1114 |
+
"completions/mean_terminated_length": 703.2611206054687,
|
| 1115 |
+
"completions/min_length": 180.4,
|
| 1116 |
+
"completions/min_terminated_length": 180.4,
|
| 1117 |
+
"epoch": 2.466246624662466,
|
| 1118 |
+
"grad_norm": 0.162477046251297,
|
| 1119 |
+
"learning_rate": 6.032343848703969e-07,
|
| 1120 |
+
"loss": 0.0154,
|
| 1121 |
+
"num_tokens": 151869315.0,
|
| 1122 |
+
"reward": 1.639928650856018,
|
| 1123 |
+
"reward_std": 0.273574697971344,
|
| 1124 |
+
"rewards/accuracy_reward/mean": 0.7631696343421936,
|
| 1125 |
+
"rewards/accuracy_reward/std": 0.42458475232124326,
|
| 1126 |
+
"rewards/cosine_scaled_reward/mean": 0.7599910497665405,
|
| 1127 |
+
"rewards/cosine_scaled_reward/std": 0.4245521605014801,
|
| 1128 |
+
"rewards/format_reward/mean": 0.9935267806053162,
|
| 1129 |
+
"rewards/format_reward/std": 0.07808763757348061,
|
| 1130 |
+
"step": 195
|
| 1131 |
+
},
|
| 1132 |
+
{
|
| 1133 |
+
"clip_ratio/high_max": 0.0,
|
| 1134 |
+
"clip_ratio/high_mean": 0.0,
|
| 1135 |
+
"clip_ratio/low_mean": 0.0,
|
| 1136 |
+
"clip_ratio/low_min": 0.0,
|
| 1137 |
+
"clip_ratio/region_mean": 0.0,
|
| 1138 |
+
"completions/clipped_ratio": 0.004241071428571441,
|
| 1139 |
+
"completions/max_length": 2048.0,
|
| 1140 |
+
"completions/max_terminated_length": 1941.8,
|
| 1141 |
+
"completions/mean_length": 718.8129760742188,
|
| 1142 |
+
"completions/mean_terminated_length": 713.2514038085938,
|
| 1143 |
+
"completions/min_length": 214.0,
|
| 1144 |
+
"completions/min_terminated_length": 214.0,
|
| 1145 |
+
"epoch": 2.5292529252925293,
|
| 1146 |
+
"grad_norm": 0.14557433128356934,
|
| 1147 |
+
"learning_rate": 5.814932121646073e-07,
|
| 1148 |
+
"loss": 0.0066,
|
| 1149 |
+
"num_tokens": 155857253.0,
|
| 1150 |
+
"reward": 1.6182387113571166,
|
| 1151 |
+
"reward_std": 0.28983315229415896,
|
| 1152 |
+
"rewards/accuracy_reward/mean": 0.7482142806053161,
|
| 1153 |
+
"rewards/accuracy_reward/std": 0.4310362696647644,
|
| 1154 |
+
"rewards/cosine_scaled_reward/mean": 0.7462987899780273,
|
| 1155 |
+
"rewards/cosine_scaled_reward/std": 0.43085583448410036,
|
| 1156 |
+
"rewards/format_reward/mean": 0.99375,
|
| 1157 |
+
"rewards/format_reward/std": 0.07288356423377991,
|
| 1158 |
+
"step": 200
|
| 1159 |
+
},
|
| 1160 |
+
{
|
| 1161 |
+
"clip_ratio/high_max": 0.0,
|
| 1162 |
+
"clip_ratio/high_mean": 0.0,
|
| 1163 |
+
"clip_ratio/low_mean": 0.0,
|
| 1164 |
+
"clip_ratio/low_min": 0.0,
|
| 1165 |
+
"clip_ratio/region_mean": 0.0,
|
| 1166 |
+
"completions/clipped_ratio": 0.005357142857142861,
|
| 1167 |
+
"completions/max_length": 2010.8,
|
| 1168 |
+
"completions/max_terminated_length": 1930.4,
|
| 1169 |
+
"completions/mean_length": 700.2652099609375,
|
| 1170 |
+
"completions/mean_terminated_length": 693.0643432617187,
|
| 1171 |
+
"completions/min_length": 226.6,
|
| 1172 |
+
"completions/min_terminated_length": 226.6,
|
| 1173 |
+
"epoch": 2.592259225922592,
|
| 1174 |
+
"grad_norm": 0.1486082673072815,
|
| 1175 |
+
"learning_rate": 5.595925126690801e-07,
|
| 1176 |
+
"loss": 0.013,
|
| 1177 |
+
"num_tokens": 159736985.0,
|
| 1178 |
+
"reward": 1.6559500694274902,
|
| 1179 |
+
"reward_std": 0.2599520206451416,
|
| 1180 |
+
"rewards/accuracy_reward/mean": 0.7738839268684388,
|
| 1181 |
+
"rewards/accuracy_reward/std": 0.41739049553871155,
|
| 1182 |
+
"rewards/cosine_scaled_reward/mean": 0.771275007724762,
|
| 1183 |
+
"rewards/cosine_scaled_reward/std": 0.41737368106842043,
|
| 1184 |
+
"rewards/format_reward/mean": 0.9928571343421936,
|
| 1185 |
+
"rewards/format_reward/std": 0.06702591478824615,
|
| 1186 |
+
"step": 205
|
| 1187 |
+
},
|
| 1188 |
+
{
|
| 1189 |
+
"clip_ratio/high_max": 0.0,
|
| 1190 |
+
"clip_ratio/high_mean": 0.0,
|
| 1191 |
+
"clip_ratio/low_mean": 0.0,
|
| 1192 |
+
"clip_ratio/low_min": 0.0,
|
| 1193 |
+
"clip_ratio/region_mean": 0.0,
|
| 1194 |
+
"completions/clipped_ratio": 0.0055803571428571395,
|
| 1195 |
+
"completions/max_length": 2048.0,
|
| 1196 |
+
"completions/max_terminated_length": 1987.4,
|
| 1197 |
+
"completions/mean_length": 712.8966796875,
|
| 1198 |
+
"completions/mean_terminated_length": 705.3931274414062,
|
| 1199 |
+
"completions/min_length": 189.0,
|
| 1200 |
+
"completions/min_terminated_length": 189.0,
|
| 1201 |
+
"epoch": 2.6552655265526552,
|
| 1202 |
+
"grad_norm": 0.12872540950775146,
|
| 1203 |
+
"learning_rate": 5.375751580315354e-07,
|
| 1204 |
+
"loss": 0.0121,
|
| 1205 |
+
"num_tokens": 163677482.0,
|
| 1206 |
+
"reward": 1.6398494243621826,
|
| 1207 |
+
"reward_std": 0.2575774312019348,
|
| 1208 |
+
"rewards/accuracy_reward/mean": 0.7622767925262451,
|
| 1209 |
+
"rewards/accuracy_reward/std": 0.42372516393661497,
|
| 1210 |
+
"rewards/cosine_scaled_reward/mean": 0.7609486818313599,
|
| 1211 |
+
"rewards/cosine_scaled_reward/std": 0.4236899495124817,
|
| 1212 |
+
"rewards/format_reward/mean": 0.9941964387893677,
|
| 1213 |
+
"rewards/format_reward/std": 0.07346804365515709,
|
| 1214 |
+
"step": 210
|
| 1215 |
+
},
|
| 1216 |
+
{
|
| 1217 |
+
"clip_ratio/high_max": 0.0,
|
| 1218 |
+
"clip_ratio/high_mean": 0.0,
|
| 1219 |
+
"clip_ratio/low_mean": 0.0,
|
| 1220 |
+
"clip_ratio/low_min": 0.0,
|
| 1221 |
+
"clip_ratio/region_mean": 0.0,
|
| 1222 |
+
"completions/clipped_ratio": 0.0022321428571428826,
|
| 1223 |
+
"completions/max_length": 2038.0,
|
| 1224 |
+
"completions/max_terminated_length": 1982.6,
|
| 1225 |
+
"completions/mean_length": 703.0440185546875,
|
| 1226 |
+
"completions/mean_terminated_length": 700.063623046875,
|
| 1227 |
+
"completions/min_length": 237.0,
|
| 1228 |
+
"completions/min_terminated_length": 237.0,
|
| 1229 |
+
"epoch": 2.7182718271827184,
|
| 1230 |
+
"grad_norm": 0.11875882744789124,
|
| 1231 |
+
"learning_rate": 5.154842482576126e-07,
|
| 1232 |
+
"loss": 0.0083,
|
| 1233 |
+
"num_tokens": 167635639.0,
|
| 1234 |
+
"reward": 1.6419362783432008,
|
| 1235 |
+
"reward_std": 0.2711367040872574,
|
| 1236 |
+
"rewards/accuracy_reward/mean": 0.7629464268684387,
|
| 1237 |
+
"rewards/accuracy_reward/std": 0.42230920791625975,
|
| 1238 |
+
"rewards/cosine_scaled_reward/mean": 0.7606581330299378,
|
| 1239 |
+
"rewards/cosine_scaled_reward/std": 0.422201132774353,
|
| 1240 |
+
"rewards/format_reward/mean": 0.9973214268684387,
|
| 1241 |
+
"rewards/format_reward/std": 0.05046965479850769,
|
| 1242 |
+
"step": 215
|
| 1243 |
+
},
|
| 1244 |
+
{
|
| 1245 |
+
"clip_ratio/high_max": 0.0,
|
| 1246 |
+
"clip_ratio/high_mean": 0.0,
|
| 1247 |
+
"clip_ratio/low_mean": 0.0,
|
| 1248 |
+
"clip_ratio/low_min": 0.0,
|
| 1249 |
+
"clip_ratio/region_mean": 0.0,
|
| 1250 |
+
"completions/clipped_ratio": 0.0040178571428571615,
|
| 1251 |
+
"completions/max_length": 2048.0,
|
| 1252 |
+
"completions/max_terminated_length": 1953.0,
|
| 1253 |
+
"completions/mean_length": 718.486181640625,
|
| 1254 |
+
"completions/mean_terminated_length": 713.2156860351563,
|
| 1255 |
+
"completions/min_length": 237.6,
|
| 1256 |
+
"completions/min_terminated_length": 237.6,
|
| 1257 |
+
"epoch": 2.781278127812781,
|
| 1258 |
+
"grad_norm": 0.14773806929588318,
|
| 1259 |
+
"learning_rate": 4.933630273405846e-07,
|
| 1260 |
+
"loss": 0.0133,
|
| 1261 |
+
"num_tokens": 171601929.0,
|
| 1262 |
+
"reward": 1.6592854976654052,
|
| 1263 |
+
"reward_std": 0.2805535584688187,
|
| 1264 |
+
"rewards/accuracy_reward/mean": 0.7752232193946839,
|
| 1265 |
+
"rewards/accuracy_reward/std": 0.41580758690834047,
|
| 1266 |
+
"rewards/cosine_scaled_reward/mean": 0.7723655581474305,
|
| 1267 |
+
"rewards/cosine_scaled_reward/std": 0.41581328511238097,
|
| 1268 |
+
"rewards/format_reward/mean": 0.9957589387893677,
|
| 1269 |
+
"rewards/format_reward/std": 0.06227037310600281,
|
| 1270 |
+
"step": 220
|
| 1271 |
+
},
|
| 1272 |
+
{
|
| 1273 |
+
"clip_ratio/high_max": 0.0,
|
| 1274 |
+
"clip_ratio/high_mean": 0.0,
|
| 1275 |
+
"clip_ratio/low_mean": 0.0,
|
| 1276 |
+
"clip_ratio/low_min": 0.0,
|
| 1277 |
+
"clip_ratio/region_mean": 0.0,
|
| 1278 |
+
"completions/clipped_ratio": 0.0033482142857143016,
|
| 1279 |
+
"completions/max_length": 2048.0,
|
| 1280 |
+
"completions/max_terminated_length": 1964.8,
|
| 1281 |
+
"completions/mean_length": 704.973681640625,
|
| 1282 |
+
"completions/mean_terminated_length": 700.4504150390625,
|
| 1283 |
+
"completions/min_length": 220.8,
|
| 1284 |
+
"completions/min_terminated_length": 220.8,
|
| 1285 |
+
"epoch": 2.8442844284428443,
|
| 1286 |
+
"grad_norm": 0.15314224362373352,
|
| 1287 |
+
"learning_rate": 4.712547986092099e-07,
|
| 1288 |
+
"loss": 0.0106,
|
| 1289 |
+
"num_tokens": 175538803.0,
|
| 1290 |
+
"reward": 1.6607633590698243,
|
| 1291 |
+
"reward_std": 0.2584164053201675,
|
| 1292 |
+
"rewards/accuracy_reward/mean": 0.7758928656578064,
|
| 1293 |
+
"rewards/accuracy_reward/std": 0.41488723158836366,
|
| 1294 |
+
"rewards/cosine_scaled_reward/mean": 0.7733123421669006,
|
| 1295 |
+
"rewards/cosine_scaled_reward/std": 0.41487019062042235,
|
| 1296 |
+
"rewards/format_reward/mean": 0.9964285612106323,
|
| 1297 |
+
"rewards/format_reward/std": 0.059030205756425855,
|
| 1298 |
+
"step": 225
|
| 1299 |
+
},
|
| 1300 |
+
{
|
| 1301 |
+
"clip_ratio/high_max": 0.0,
|
| 1302 |
+
"clip_ratio/high_mean": 0.0,
|
| 1303 |
+
"clip_ratio/low_mean": 0.0,
|
| 1304 |
+
"clip_ratio/low_min": 0.0,
|
| 1305 |
+
"clip_ratio/region_mean": 0.0,
|
| 1306 |
+
"completions/clipped_ratio": 0.002678571428571441,
|
| 1307 |
+
"completions/max_length": 2048.0,
|
| 1308 |
+
"completions/max_terminated_length": 1926.0,
|
| 1309 |
+
"completions/mean_length": 719.9161010742188,
|
| 1310 |
+
"completions/mean_terminated_length": 716.3463256835937,
|
| 1311 |
+
"completions/min_length": 232.4,
|
| 1312 |
+
"completions/min_terminated_length": 232.4,
|
| 1313 |
+
"epoch": 2.9072907290729075,
|
| 1314 |
+
"grad_norm": 0.13197660446166992,
|
| 1315 |
+
"learning_rate": 4.492028399594338e-07,
|
| 1316 |
+
"loss": 0.0071,
|
| 1317 |
+
"num_tokens": 179542971.0,
|
| 1318 |
+
"reward": 1.6427105903625487,
|
| 1319 |
+
"reward_std": 0.2545609176158905,
|
| 1320 |
+
"rewards/accuracy_reward/mean": 0.7636160731315613,
|
| 1321 |
+
"rewards/accuracy_reward/std": 0.422965270280838,
|
| 1322 |
+
"rewards/cosine_scaled_reward/mean": 0.7610907316207886,
|
| 1323 |
+
"rewards/cosine_scaled_reward/std": 0.42288472652435305,
|
| 1324 |
+
"rewards/format_reward/mean": 0.9970982074737549,
|
| 1325 |
+
"rewards/format_reward/std": 0.05258563756942749,
|
| 1326 |
+
"step": 230
|
| 1327 |
+
},
|
| 1328 |
+
{
|
| 1329 |
+
"clip_ratio/high_max": 0.0,
|
| 1330 |
+
"clip_ratio/high_mean": 0.0,
|
| 1331 |
+
"clip_ratio/low_mean": 0.0,
|
| 1332 |
+
"clip_ratio/low_min": 0.0,
|
| 1333 |
+
"clip_ratio/region_mean": 0.0,
|
| 1334 |
+
"completions/clipped_ratio": 0.005803571428571419,
|
| 1335 |
+
"completions/max_length": 2048.0,
|
| 1336 |
+
"completions/max_terminated_length": 1992.0,
|
| 1337 |
+
"completions/mean_length": 749.613427734375,
|
| 1338 |
+
"completions/mean_terminated_length": 742.11640625,
|
| 1339 |
+
"completions/min_length": 227.6,
|
| 1340 |
+
"completions/min_terminated_length": 227.6,
|
| 1341 |
+
"epoch": 2.9702970297029703,
|
| 1342 |
+
"grad_norm": 0.1521228849887848,
|
| 1343 |
+
"learning_rate": 4.2725031913587426e-07,
|
| 1344 |
+
"loss": 0.0102,
|
| 1345 |
+
"num_tokens": 183700159.0,
|
| 1346 |
+
"reward": 1.6156146049499511,
|
| 1347 |
+
"reward_std": 0.27087976038455963,
|
| 1348 |
+
"rewards/accuracy_reward/mean": 0.7462053656578064,
|
| 1349 |
+
"rewards/accuracy_reward/std": 0.4335868239402771,
|
| 1350 |
+
"rewards/cosine_scaled_reward/mean": 0.744845175743103,
|
| 1351 |
+
"rewards/cosine_scaled_reward/std": 0.43346109986305237,
|
| 1352 |
+
"rewards/format_reward/mean": 0.9939731955528259,
|
| 1353 |
+
"rewards/format_reward/std": 0.07259788736701012,
|
| 1354 |
+
"step": 235
|
| 1355 |
+
},
|
| 1356 |
+
{
|
| 1357 |
+
"clip_ratio/high_max": 0.0,
|
| 1358 |
+
"clip_ratio/high_mean": 0.0,
|
| 1359 |
+
"clip_ratio/low_mean": 0.0,
|
| 1360 |
+
"clip_ratio/low_min": 0.0,
|
| 1361 |
+
"clip_ratio/region_mean": 0.0,
|
| 1362 |
+
"completions/clipped_ratio": 0.005552184466019417,
|
| 1363 |
+
"completions/max_length": 2048.0,
|
| 1364 |
+
"completions/max_terminated_length": 1949.6,
|
| 1365 |
+
"completions/mean_length": 764.5406005859375,
|
| 1366 |
+
"completions/mean_terminated_length": 757.5264038085937,
|
| 1367 |
+
"completions/min_length": 240.6,
|
| 1368 |
+
"completions/min_terminated_length": 240.6,
|
| 1369 |
+
"epoch": 3.037803780378038,
|
| 1370 |
+
"grad_norm": 0.1749519556760788,
|
| 1371 |
+
"learning_rate": 4.054402092289354e-07,
|
| 1372 |
+
"loss": 0.0079,
|
| 1373 |
+
"num_tokens": 187977498.0,
|
| 1374 |
+
"reward": 1.561311173439026,
|
| 1375 |
+
"reward_std": 0.30535221099853516,
|
| 1376 |
+
"rewards/accuracy_reward/mean": 0.7107142925262451,
|
| 1377 |
+
"rewards/accuracy_reward/std": 0.45136345624923707,
|
| 1378 |
+
"rewards/cosine_scaled_reward/mean": 0.7067740797996521,
|
| 1379 |
+
"rewards/cosine_scaled_reward/std": 0.45066094398498535,
|
| 1380 |
+
"rewards/format_reward/mean": 0.9944196462631225,
|
| 1381 |
+
"rewards/format_reward/std": 0.06912030950188637,
|
| 1382 |
+
"step": 240
|
| 1383 |
+
},
|
| 1384 |
+
{
|
| 1385 |
+
"clip_ratio/high_max": 0.0,
|
| 1386 |
+
"clip_ratio/high_mean": 0.0,
|
| 1387 |
+
"clip_ratio/low_mean": 0.0,
|
| 1388 |
+
"clip_ratio/low_min": 0.0,
|
| 1389 |
+
"clip_ratio/region_mean": 0.0,
|
| 1390 |
+
"completions/clipped_ratio": 0.0049107142857142795,
|
| 1391 |
+
"completions/max_length": 2048.0,
|
| 1392 |
+
"completions/max_terminated_length": 1971.4,
|
| 1393 |
+
"completions/mean_length": 733.2201293945312,
|
| 1394 |
+
"completions/mean_terminated_length": 726.7033935546875,
|
| 1395 |
+
"completions/min_length": 226.6,
|
| 1396 |
+
"completions/min_terminated_length": 226.6,
|
| 1397 |
+
"epoch": 3.1008100810081007,
|
| 1398 |
+
"grad_norm": 0.15469872951507568,
|
| 1399 |
+
"learning_rate": 3.838152045529636e-07,
|
| 1400 |
+
"loss": 0.0114,
|
| 1401 |
+
"num_tokens": 192025564.0,
|
| 1402 |
+
"reward": 1.670283031463623,
|
| 1403 |
+
"reward_std": 0.2512336015701294,
|
| 1404 |
+
"rewards/accuracy_reward/mean": 0.7828125,
|
| 1405 |
+
"rewards/accuracy_reward/std": 0.4110615670681,
|
| 1406 |
+
"rewards/cosine_scaled_reward/mean": 0.7798516988754273,
|
| 1407 |
+
"rewards/cosine_scaled_reward/std": 0.41104503273963927,
|
| 1408 |
+
"rewards/format_reward/mean": 0.9950892806053162,
|
| 1409 |
+
"rewards/format_reward/std": 0.06898552328348159,
|
| 1410 |
+
"step": 245
|
| 1411 |
+
},
|
| 1412 |
+
{
|
| 1413 |
+
"clip_ratio/high_max": 0.0,
|
| 1414 |
+
"clip_ratio/high_mean": 0.0,
|
| 1415 |
+
"clip_ratio/low_mean": 0.0,
|
| 1416 |
+
"clip_ratio/low_min": 0.0,
|
| 1417 |
+
"clip_ratio/region_mean": 0.0,
|
| 1418 |
+
"completions/clipped_ratio": 0.004910714285714257,
|
| 1419 |
+
"completions/max_length": 2043.8,
|
| 1420 |
+
"completions/max_terminated_length": 1944.6,
|
| 1421 |
+
"completions/mean_length": 732.2582885742188,
|
| 1422 |
+
"completions/mean_terminated_length": 725.8395385742188,
|
| 1423 |
+
"completions/min_length": 203.8,
|
| 1424 |
+
"completions/min_terminated_length": 203.8,
|
| 1425 |
+
"epoch": 3.163816381638164,
|
| 1426 |
+
"grad_norm": 0.12465714663267136,
|
| 1427 |
+
"learning_rate": 3.624176370701235e-07,
|
| 1428 |
+
"loss": 0.0125,
|
| 1429 |
+
"num_tokens": 196069785.0,
|
| 1430 |
+
"reward": 1.6728124141693115,
|
| 1431 |
+
"reward_std": 0.26720473170280457,
|
| 1432 |
+
"rewards/accuracy_reward/mean": 0.7848214268684387,
|
| 1433 |
+
"rewards/accuracy_reward/std": 0.4107992649078369,
|
| 1434 |
+
"rewards/cosine_scaled_reward/mean": 0.7813390612602233,
|
| 1435 |
+
"rewards/cosine_scaled_reward/std": 0.41090287566184996,
|
| 1436 |
+
"rewards/format_reward/mean": 0.9946428537368774,
|
| 1437 |
+
"rewards/format_reward/std": 0.061242764443159105,
|
| 1438 |
+
"step": 250
|
| 1439 |
+
},
|
| 1440 |
+
{
|
| 1441 |
+
"clip_ratio/high_max": 0.0,
|
| 1442 |
+
"clip_ratio/high_mean": 0.0,
|
| 1443 |
+
"clip_ratio/low_mean": 0.0,
|
| 1444 |
+
"clip_ratio/low_min": 0.0,
|
| 1445 |
+
"clip_ratio/region_mean": 0.0,
|
| 1446 |
+
"completions/clipped_ratio": 0.004017857142857117,
|
| 1447 |
+
"completions/max_length": 2048.0,
|
| 1448 |
+
"completions/max_terminated_length": 1911.2,
|
| 1449 |
+
"completions/mean_length": 733.9761596679688,
|
| 1450 |
+
"completions/mean_terminated_length": 728.7048950195312,
|
| 1451 |
+
"completions/min_length": 227.6,
|
| 1452 |
+
"completions/min_terminated_length": 227.6,
|
| 1453 |
+
"epoch": 3.2268226822682267,
|
| 1454 |
+
"grad_norm": 0.1451456993818283,
|
| 1455 |
+
"learning_rate": 3.412893935235936e-07,
|
| 1456 |
+
"loss": 0.0097,
|
| 1457 |
+
"num_tokens": 200116558.0,
|
| 1458 |
+
"reward": 1.6486672163009644,
|
| 1459 |
+
"reward_std": 0.2642055004835129,
|
| 1460 |
+
"rewards/accuracy_reward/mean": 0.7683035731315613,
|
| 1461 |
+
"rewards/accuracy_reward/std": 0.4195379137992859,
|
| 1462 |
+
"rewards/cosine_scaled_reward/mean": 0.7640753865242005,
|
| 1463 |
+
"rewards/cosine_scaled_reward/std": 0.4194532334804535,
|
| 1464 |
+
"rewards/format_reward/mean": 0.9966517925262451,
|
| 1465 |
+
"rewards/format_reward/std": 0.05103206038475037,
|
| 1466 |
+
"step": 255
|
| 1467 |
+
},
|
| 1468 |
+
{
|
| 1469 |
+
"clip_ratio/high_max": 0.0,
|
| 1470 |
+
"clip_ratio/high_mean": 0.0,
|
| 1471 |
+
"clip_ratio/low_mean": 0.0,
|
| 1472 |
+
"clip_ratio/low_min": 0.0,
|
| 1473 |
+
"clip_ratio/region_mean": 0.0,
|
| 1474 |
+
"completions/clipped_ratio": 0.0046875,
|
| 1475 |
+
"completions/max_length": 2048.0,
|
| 1476 |
+
"completions/max_terminated_length": 1961.4,
|
| 1477 |
+
"completions/mean_length": 744.1857421875,
|
| 1478 |
+
"completions/mean_terminated_length": 738.0572265625,
|
| 1479 |
+
"completions/min_length": 229.0,
|
| 1480 |
+
"completions/min_terminated_length": 229.0,
|
| 1481 |
+
"epoch": 3.28982898289829,
|
| 1482 |
+
"grad_norm": 0.1736501306295395,
|
| 1483 |
+
"learning_rate": 3.204718334422998e-07,
|
| 1484 |
+
"loss": 0.0084,
|
| 1485 |
+
"num_tokens": 204218390.0,
|
| 1486 |
+
"reward": 1.6481477737426757,
|
| 1487 |
+
"reward_std": 0.27542482018470765,
|
| 1488 |
+
"rewards/accuracy_reward/mean": 0.7685267806053162,
|
| 1489 |
+
"rewards/accuracy_reward/std": 0.4199278116226196,
|
| 1490 |
+
"rewards/cosine_scaled_reward/mean": 0.7641526460647583,
|
| 1491 |
+
"rewards/cosine_scaled_reward/std": 0.4198519289493561,
|
| 1492 |
+
"rewards/format_reward/mean": 0.9950892925262451,
|
| 1493 |
+
"rewards/format_reward/std": 0.0682831697165966,
|
| 1494 |
+
"step": 260
|
| 1495 |
+
},
|
| 1496 |
+
{
|
| 1497 |
+
"clip_ratio/high_max": 0.0,
|
| 1498 |
+
"clip_ratio/high_mean": 0.0,
|
| 1499 |
+
"clip_ratio/low_mean": 0.0,
|
| 1500 |
+
"clip_ratio/low_min": 0.0,
|
| 1501 |
+
"clip_ratio/region_mean": 0.0,
|
| 1502 |
+
"completions/clipped_ratio": 0.0031250000000000223,
|
| 1503 |
+
"completions/max_length": 2048.0,
|
| 1504 |
+
"completions/max_terminated_length": 1957.0,
|
| 1505 |
+
"completions/mean_length": 741.0214721679688,
|
| 1506 |
+
"completions/mean_terminated_length": 736.9272338867188,
|
| 1507 |
+
"completions/min_length": 199.4,
|
| 1508 |
+
"completions/min_terminated_length": 199.4,
|
| 1509 |
+
"epoch": 3.352835283528353,
|
| 1510 |
+
"grad_norm": 0.1577436625957489,
|
| 1511 |
+
"learning_rate": 3.0000570817769464e-07,
|
| 1512 |
+
"loss": 0.0081,
|
| 1513 |
+
"num_tokens": 208315302.0,
|
| 1514 |
+
"reward": 1.6395839691162108,
|
| 1515 |
+
"reward_std": 0.29457778930664064,
|
| 1516 |
+
"rewards/accuracy_reward/mean": 0.7618303418159484,
|
| 1517 |
+
"rewards/accuracy_reward/std": 0.4226981282234192,
|
| 1518 |
+
"rewards/cosine_scaled_reward/mean": 0.7586321473121643,
|
| 1519 |
+
"rewards/cosine_scaled_reward/std": 0.42258375883102417,
|
| 1520 |
+
"rewards/format_reward/mean": 0.996874988079071,
|
| 1521 |
+
"rewards/format_reward/std": 0.05568308532238007,
|
| 1522 |
+
"step": 265
|
| 1523 |
+
},
|
| 1524 |
+
{
|
| 1525 |
+
"clip_ratio/high_max": 0.0,
|
| 1526 |
+
"clip_ratio/high_mean": 0.0,
|
| 1527 |
+
"clip_ratio/low_mean": 0.0,
|
| 1528 |
+
"clip_ratio/low_min": 0.0,
|
| 1529 |
+
"clip_ratio/region_mean": 0.0,
|
| 1530 |
+
"completions/clipped_ratio": 0.004910714285714301,
|
| 1531 |
+
"completions/max_length": 2048.0,
|
| 1532 |
+
"completions/max_terminated_length": 1973.0,
|
| 1533 |
+
"completions/mean_length": 746.5964599609375,
|
| 1534 |
+
"completions/mean_terminated_length": 740.2795288085938,
|
| 1535 |
+
"completions/min_length": 228.4,
|
| 1536 |
+
"completions/min_terminated_length": 228.4,
|
| 1537 |
+
"epoch": 3.4158415841584158,
|
| 1538 |
+
"grad_norm": 0.16179685294628143,
|
| 1539 |
+
"learning_rate": 2.7993108113107443e-07,
|
| 1540 |
+
"loss": 0.0063,
|
| 1541 |
+
"num_tokens": 212457542.0,
|
| 1542 |
+
"reward": 1.6618568658828736,
|
| 1543 |
+
"reward_std": 0.25516128838062285,
|
| 1544 |
+
"rewards/accuracy_reward/mean": 0.7776785731315613,
|
| 1545 |
+
"rewards/accuracy_reward/std": 0.4147783398628235,
|
| 1546 |
+
"rewards/cosine_scaled_reward/mean": 0.7734902620315551,
|
| 1547 |
+
"rewards/cosine_scaled_reward/std": 0.41475528478622437,
|
| 1548 |
+
"rewards/format_reward/mean": 0.9948660731315613,
|
| 1549 |
+
"rewards/format_reward/std": 0.06522349417209625,
|
| 1550 |
+
"step": 270
|
| 1551 |
+
},
|
| 1552 |
+
{
|
| 1553 |
+
"clip_ratio/high_max": 0.0,
|
| 1554 |
+
"clip_ratio/high_mean": 0.0,
|
| 1555 |
+
"clip_ratio/low_mean": 0.0,
|
| 1556 |
+
"clip_ratio/low_min": 0.0,
|
| 1557 |
+
"clip_ratio/region_mean": 0.0,
|
| 1558 |
+
"completions/clipped_ratio": 0.00401785714285714,
|
| 1559 |
+
"completions/max_length": 2048.0,
|
| 1560 |
+
"completions/max_terminated_length": 1937.4,
|
| 1561 |
+
"completions/mean_length": 733.0241333007813,
|
| 1562 |
+
"completions/mean_terminated_length": 727.7449096679687,
|
| 1563 |
+
"completions/min_length": 238.0,
|
| 1564 |
+
"completions/min_terminated_length": 238.0,
|
| 1565 |
+
"epoch": 3.478847884788479,
|
| 1566 |
+
"grad_norm": 0.13407224416732788,
|
| 1567 |
+
"learning_rate": 2.602872493275881e-07,
|
| 1568 |
+
"loss": 0.0094,
|
| 1569 |
+
"num_tokens": 216523066.0,
|
| 1570 |
+
"reward": 1.6756507635116578,
|
| 1571 |
+
"reward_std": 0.24836473166942596,
|
| 1572 |
+
"rewards/accuracy_reward/mean": 0.7861607074737549,
|
| 1573 |
+
"rewards/accuracy_reward/std": 0.40959796905517576,
|
| 1574 |
+
"rewards/cosine_scaled_reward/mean": 0.7829978227615356,
|
| 1575 |
+
"rewards/cosine_scaled_reward/std": 0.4097108542919159,
|
| 1576 |
+
"rewards/format_reward/mean": 0.9959821343421936,
|
| 1577 |
+
"rewards/format_reward/std": 0.05930030345916748,
|
| 1578 |
+
"step": 275
|
| 1579 |
+
},
|
| 1580 |
+
{
|
| 1581 |
+
"clip_ratio/high_max": 0.0,
|
| 1582 |
+
"clip_ratio/high_mean": 0.0,
|
| 1583 |
+
"clip_ratio/low_mean": 0.0,
|
| 1584 |
+
"clip_ratio/low_min": 0.0,
|
| 1585 |
+
"clip_ratio/region_mean": 0.0,
|
| 1586 |
+
"completions/clipped_ratio": 0.004687499999999978,
|
| 1587 |
+
"completions/max_length": 2048.0,
|
| 1588 |
+
"completions/max_terminated_length": 1963.0,
|
| 1589 |
+
"completions/mean_length": 758.125048828125,
|
| 1590 |
+
"completions/mean_terminated_length": 752.0442993164063,
|
| 1591 |
+
"completions/min_length": 224.2,
|
| 1592 |
+
"completions/min_terminated_length": 224.2,
|
| 1593 |
+
"epoch": 3.5418541854185417,
|
| 1594 |
+
"grad_norm": 0.24010425806045532,
|
| 1595 |
+
"learning_rate": 2.411126664904665e-07,
|
| 1596 |
+
"loss": 0.0122,
|
| 1597 |
+
"num_tokens": 220698850.0,
|
| 1598 |
+
"reward": 1.6234807968139648,
|
| 1599 |
+
"reward_std": 0.2710172593593597,
|
| 1600 |
+
"rewards/accuracy_reward/mean": 0.7522321462631225,
|
| 1601 |
+
"rewards/accuracy_reward/std": 0.4294972479343414,
|
| 1602 |
+
"rewards/cosine_scaled_reward/mean": 0.7489704251289367,
|
| 1603 |
+
"rewards/cosine_scaled_reward/std": 0.429376620054245,
|
| 1604 |
+
"rewards/format_reward/mean": 0.9935267925262451,
|
| 1605 |
+
"rewards/format_reward/std": 0.07893114537000656,
|
| 1606 |
+
"step": 280
|
| 1607 |
+
},
|
| 1608 |
+
{
|
| 1609 |
+
"clip_ratio/high_max": 0.0,
|
| 1610 |
+
"clip_ratio/high_mean": 0.0,
|
| 1611 |
+
"clip_ratio/low_mean": 0.0,
|
| 1612 |
+
"clip_ratio/low_min": 0.0,
|
| 1613 |
+
"clip_ratio/region_mean": 0.0,
|
| 1614 |
+
"completions/clipped_ratio": 0.005580357142857118,
|
| 1615 |
+
"completions/max_length": 2048.0,
|
| 1616 |
+
"completions/max_terminated_length": 1972.6,
|
| 1617 |
+
"completions/mean_length": 725.0837524414062,
|
| 1618 |
+
"completions/mean_terminated_length": 717.6320190429688,
|
| 1619 |
+
"completions/min_length": 235.8,
|
| 1620 |
+
"completions/min_terminated_length": 235.8,
|
| 1621 |
+
"epoch": 3.604860486048605,
|
| 1622 |
+
"grad_norm": 0.15291793644428253,
|
| 1623 |
+
"learning_rate": 2.2244486776605241e-07,
|
| 1624 |
+
"loss": 0.0062,
|
| 1625 |
+
"num_tokens": 224701689.0,
|
| 1626 |
+
"reward": 1.6737475872039795,
|
| 1627 |
+
"reward_std": 0.24555476903915405,
|
| 1628 |
+
"rewards/accuracy_reward/mean": 0.7852678775787354,
|
| 1629 |
+
"rewards/accuracy_reward/std": 0.4096054911613464,
|
| 1630 |
+
"rewards/cosine_scaled_reward/mean": 0.7825396418571472,
|
| 1631 |
+
"rewards/cosine_scaled_reward/std": 0.4097288966178894,
|
| 1632 |
+
"rewards/format_reward/mean": 0.9944196224212647,
|
| 1633 |
+
"rewards/format_reward/std": 0.07079662829637527,
|
| 1634 |
+
"step": 285
|
| 1635 |
+
},
|
| 1636 |
+
{
|
| 1637 |
+
"clip_ratio/high_max": 0.0,
|
| 1638 |
+
"clip_ratio/high_mean": 0.0,
|
| 1639 |
+
"clip_ratio/low_mean": 0.0,
|
| 1640 |
+
"clip_ratio/low_min": 0.0,
|
| 1641 |
+
"clip_ratio/region_mean": 0.0,
|
| 1642 |
+
"completions/clipped_ratio": 0.0029017857142856983,
|
| 1643 |
+
"completions/max_length": 2020.8,
|
| 1644 |
+
"completions/max_terminated_length": 1891.0,
|
| 1645 |
+
"completions/mean_length": 721.8882080078125,
|
| 1646 |
+
"completions/mean_terminated_length": 718.0403076171875,
|
| 1647 |
+
"completions/min_length": 218.8,
|
| 1648 |
+
"completions/min_terminated_length": 218.8,
|
| 1649 |
+
"epoch": 3.667866786678668,
|
| 1650 |
+
"grad_norm": 0.17101366817951202,
|
| 1651 |
+
"learning_rate": 2.0432039624699015e-07,
|
| 1652 |
+
"loss": 0.0122,
|
| 1653 |
+
"num_tokens": 228762668.0,
|
| 1654 |
+
"reward": 1.6519631147384644,
|
| 1655 |
+
"reward_std": 0.2486158311367035,
|
| 1656 |
+
"rewards/accuracy_reward/mean": 0.7694196343421936,
|
| 1657 |
+
"rewards/accuracy_reward/std": 0.4205454409122467,
|
| 1658 |
+
"rewards/cosine_scaled_reward/mean": 0.7682117819786072,
|
| 1659 |
+
"rewards/cosine_scaled_reward/std": 0.4205586850643158,
|
| 1660 |
+
"rewards/format_reward/mean": 0.996875,
|
| 1661 |
+
"rewards/format_reward/std": 0.048269759863615036,
|
| 1662 |
+
"step": 290
|
| 1663 |
+
},
|
| 1664 |
+
{
|
| 1665 |
+
"clip_ratio/high_max": 0.0,
|
| 1666 |
+
"clip_ratio/high_mean": 0.0,
|
| 1667 |
+
"clip_ratio/low_mean": 0.0,
|
| 1668 |
+
"clip_ratio/low_min": 0.0,
|
| 1669 |
+
"clip_ratio/region_mean": 0.0,
|
| 1670 |
+
"completions/clipped_ratio": 0.004910714285714301,
|
| 1671 |
+
"completions/max_length": 2048.0,
|
| 1672 |
+
"completions/max_terminated_length": 1985.4,
|
| 1673 |
+
"completions/mean_length": 755.2917846679687,
|
| 1674 |
+
"completions/mean_terminated_length": 748.9348266601562,
|
| 1675 |
+
"completions/min_length": 224.4,
|
| 1676 |
+
"completions/min_terminated_length": 224.4,
|
| 1677 |
+
"epoch": 3.730873087308731,
|
| 1678 |
+
"grad_norm": 0.15902669727802277,
|
| 1679 |
+
"learning_rate": 1.867747314374041e-07,
|
| 1680 |
+
"loss": 0.0129,
|
| 1681 |
+
"num_tokens": 232940423.0,
|
| 1682 |
+
"reward": 1.6354996681213378,
|
| 1683 |
+
"reward_std": 0.25759266018867494,
|
| 1684 |
+
"rewards/accuracy_reward/mean": 0.7604910731315613,
|
| 1685 |
+
"rewards/accuracy_reward/std": 0.42592642903327943,
|
| 1686 |
+
"rewards/cosine_scaled_reward/mean": 0.7549277067184448,
|
| 1687 |
+
"rewards/cosine_scaled_reward/std": 0.425705748796463,
|
| 1688 |
+
"rewards/format_reward/mean": 0.9950892806053162,
|
| 1689 |
+
"rewards/format_reward/std": 0.06402865275740624,
|
| 1690 |
+
"step": 295
|
| 1691 |
+
},
|
| 1692 |
+
{
|
| 1693 |
+
"clip_ratio/high_max": 0.0,
|
| 1694 |
+
"clip_ratio/high_mean": 0.0,
|
| 1695 |
+
"clip_ratio/low_mean": 0.0,
|
| 1696 |
+
"clip_ratio/low_min": 0.0,
|
| 1697 |
+
"clip_ratio/region_mean": 0.0,
|
| 1698 |
+
"completions/clipped_ratio": 0.004464285714285698,
|
| 1699 |
+
"completions/max_length": 2028.8,
|
| 1700 |
+
"completions/max_terminated_length": 1990.0,
|
| 1701 |
+
"completions/mean_length": 745.7913330078125,
|
| 1702 |
+
"completions/mean_terminated_length": 739.9611328125,
|
| 1703 |
+
"completions/min_length": 235.2,
|
| 1704 |
+
"completions/min_terminated_length": 235.2,
|
| 1705 |
+
"epoch": 3.793879387938794,
|
| 1706 |
+
"grad_norm": 0.16288451850414276,
|
| 1707 |
+
"learning_rate": 1.6984221980010454e-07,
|
| 1708 |
+
"loss": 0.0103,
|
| 1709 |
+
"num_tokens": 237049072.0,
|
| 1710 |
+
"reward": 1.678315019607544,
|
| 1711 |
+
"reward_std": 0.25401819348335264,
|
| 1712 |
+
"rewards/accuracy_reward/mean": 0.7883928537368774,
|
| 1713 |
+
"rewards/accuracy_reward/std": 0.4084781646728516,
|
| 1714 |
+
"rewards/cosine_scaled_reward/mean": 0.7847549080848694,
|
| 1715 |
+
"rewards/cosine_scaled_reward/std": 0.4085943877696991,
|
| 1716 |
+
"rewards/format_reward/mean": 0.9950892925262451,
|
| 1717 |
+
"rewards/format_reward/std": 0.0597273975610733,
|
| 1718 |
+
"step": 300
|
| 1719 |
+
},
|
| 1720 |
+
{
|
| 1721 |
+
"clip_ratio/high_max": 0.0,
|
| 1722 |
+
"clip_ratio/high_mean": 0.0,
|
| 1723 |
+
"clip_ratio/low_mean": 0.0,
|
| 1724 |
+
"clip_ratio/low_min": 0.0,
|
| 1725 |
+
"clip_ratio/region_mean": 0.0,
|
| 1726 |
+
"completions/clipped_ratio": 0.004464285714285698,
|
| 1727 |
+
"completions/max_length": 2048.0,
|
| 1728 |
+
"completions/max_terminated_length": 1976.0,
|
| 1729 |
+
"completions/mean_length": 747.590234375,
|
| 1730 |
+
"completions/mean_terminated_length": 741.756640625,
|
| 1731 |
+
"completions/min_length": 220.8,
|
| 1732 |
+
"completions/min_terminated_length": 220.8,
|
| 1733 |
+
"epoch": 3.856885688568857,
|
| 1734 |
+
"grad_norm": 0.16114351153373718,
|
| 1735 |
+
"learning_rate": 1.5355600752177312e-07,
|
| 1736 |
+
"loss": 0.0113,
|
| 1737 |
+
"num_tokens": 241188012.0,
|
| 1738 |
+
"reward": 1.6530835628509521,
|
| 1739 |
+
"reward_std": 0.2503146827220917,
|
| 1740 |
+
"rewards/accuracy_reward/mean": 0.7714285731315613,
|
| 1741 |
+
"rewards/accuracy_reward/std": 0.4196752727031708,
|
| 1742 |
+
"rewards/cosine_scaled_reward/mean": 0.7677741289138794,
|
| 1743 |
+
"rewards/cosine_scaled_reward/std": 0.4196808636188507,
|
| 1744 |
+
"rewards/format_reward/mean": 0.9955357074737549,
|
| 1745 |
+
"rewards/format_reward/std": 0.06474298015236854,
|
| 1746 |
+
"step": 305
|
| 1747 |
+
},
|
| 1748 |
+
{
|
| 1749 |
+
"clip_ratio/high_max": 0.0,
|
| 1750 |
+
"clip_ratio/high_mean": 0.0,
|
| 1751 |
+
"clip_ratio/low_mean": 0.0,
|
| 1752 |
+
"clip_ratio/low_min": 0.0,
|
| 1753 |
+
"clip_ratio/region_mean": 0.0,
|
| 1754 |
+
"completions/clipped_ratio": 0.005133928571428581,
|
| 1755 |
+
"completions/max_length": 2048.0,
|
| 1756 |
+
"completions/max_terminated_length": 1974.8,
|
| 1757 |
+
"completions/mean_length": 726.7525024414062,
|
| 1758 |
+
"completions/mean_terminated_length": 719.9224853515625,
|
| 1759 |
+
"completions/min_length": 192.2,
|
| 1760 |
+
"completions/min_terminated_length": 192.2,
|
| 1761 |
+
"epoch": 3.91989198919892,
|
| 1762 |
+
"grad_norm": 0.1752256602048874,
|
| 1763 |
+
"learning_rate": 1.3794797562774596e-07,
|
| 1764 |
+
"loss": 0.0065,
|
| 1765 |
+
"num_tokens": 245204647.0,
|
| 1766 |
+
"reward": 1.633379077911377,
|
| 1767 |
+
"reward_std": 0.25873309969902036,
|
| 1768 |
+
"rewards/accuracy_reward/mean": 0.7604951024055481,
|
| 1769 |
+
"rewards/accuracy_reward/std": 0.4261723875999451,
|
| 1770 |
+
"rewards/cosine_scaled_reward/mean": 0.7551171183586121,
|
| 1771 |
+
"rewards/cosine_scaled_reward/std": 0.42787304520606995,
|
| 1772 |
+
"rewards/format_reward/mean": 0.9946428537368774,
|
| 1773 |
+
"rewards/format_reward/std": 0.06888841390609741,
|
| 1774 |
+
"step": 310
|
| 1775 |
+
},
|
| 1776 |
+
{
|
| 1777 |
+
"clip_ratio/high_max": 0.0,
|
| 1778 |
+
"clip_ratio/high_mean": 0.0,
|
| 1779 |
+
"clip_ratio/low_mean": 0.0,
|
| 1780 |
+
"clip_ratio/low_min": 0.0,
|
| 1781 |
+
"clip_ratio/region_mean": 0.0,
|
| 1782 |
+
"completions/clipped_ratio": 0.006026785714285698,
|
| 1783 |
+
"completions/max_length": 2048.0,
|
| 1784 |
+
"completions/max_terminated_length": 2022.2,
|
| 1785 |
+
"completions/mean_length": 715.9703491210937,
|
| 1786 |
+
"completions/mean_terminated_length": 707.817626953125,
|
| 1787 |
+
"completions/min_length": 225.2,
|
| 1788 |
+
"completions/min_terminated_length": 225.2,
|
| 1789 |
+
"epoch": 3.9828982898289826,
|
| 1790 |
+
"grad_norm": 0.1453341394662857,
|
| 1791 |
+
"learning_rate": 1.230486775734097e-07,
|
| 1792 |
+
"loss": 0.0124,
|
| 1793 |
+
"num_tokens": 249253770.0,
|
| 1794 |
+
"reward": 1.6327741384506225,
|
| 1795 |
+
"reward_std": 0.24864841401576995,
|
| 1796 |
+
"rewards/accuracy_reward/mean": 0.7609555959701538,
|
| 1797 |
+
"rewards/accuracy_reward/std": 0.42520409226417544,
|
| 1798 |
+
"rewards/cosine_scaled_reward/mean": 0.7526016592979431,
|
| 1799 |
+
"rewards/cosine_scaled_reward/std": 0.4279878318309784,
|
| 1800 |
+
"rewards/format_reward/mean": 0.99375,
|
| 1801 |
+
"rewards/format_reward/std": 0.07615087106823921,
|
| 1802 |
+
"step": 315
|
| 1803 |
+
},
|
| 1804 |
+
{
|
| 1805 |
+
"clip_ratio/high_max": 0.0,
|
| 1806 |
+
"clip_ratio/high_mean": 0.0,
|
| 1807 |
+
"clip_ratio/low_mean": 0.0,
|
| 1808 |
+
"clip_ratio/low_min": 0.0,
|
| 1809 |
+
"clip_ratio/region_mean": 0.0,
|
| 1810 |
+
"completions/clipped_ratio": 0.003649445214979186,
|
| 1811 |
+
"completions/max_length": 2018.2,
|
| 1812 |
+
"completions/max_terminated_length": 1983.2,
|
| 1813 |
+
"completions/mean_length": 726.5947998046875,
|
| 1814 |
+
"completions/mean_terminated_length": 721.7894653320312,
|
| 1815 |
+
"completions/min_length": 218.0,
|
| 1816 |
+
"completions/min_terminated_length": 218.0,
|
| 1817 |
+
"epoch": 4.05040504050405,
|
| 1818 |
+
"grad_norm": 0.1543412208557129,
|
| 1819 |
+
"learning_rate": 1.0888727943437748e-07,
|
| 1820 |
+
"loss": 0.0102,
|
| 1821 |
+
"num_tokens": 253270490.0,
|
| 1822 |
+
"reward": 1.6437525272369384,
|
| 1823 |
+
"reward_std": 0.2744287014007568,
|
| 1824 |
+
"rewards/accuracy_reward/mean": 0.7688465356826782,
|
| 1825 |
+
"rewards/accuracy_reward/std": 0.4200101852416992,
|
| 1826 |
+
"rewards/cosine_scaled_reward/mean": 0.7564781427383422,
|
| 1827 |
+
"rewards/cosine_scaled_reward/std": 0.4258088171482086,
|
| 1828 |
+
"rewards/format_reward/mean": 0.9962053656578064,
|
| 1829 |
+
"rewards/format_reward/std": 0.054011975228786466,
|
| 1830 |
+
"step": 320
|
| 1831 |
+
},
|
| 1832 |
+
{
|
| 1833 |
+
"clip_ratio/high_max": 0.0,
|
| 1834 |
+
"clip_ratio/high_mean": 0.0,
|
| 1835 |
+
"clip_ratio/low_mean": 0.0,
|
| 1836 |
+
"clip_ratio/low_min": 0.0,
|
| 1837 |
+
"clip_ratio/region_mean": 0.0,
|
| 1838 |
+
"completions/clipped_ratio": 0.004464285714285698,
|
| 1839 |
+
"completions/max_length": 2048.0,
|
| 1840 |
+
"completions/max_terminated_length": 1982.2,
|
| 1841 |
+
"completions/mean_length": 722.0165405273438,
|
| 1842 |
+
"completions/mean_terminated_length": 716.0443603515625,
|
| 1843 |
+
"completions/min_length": 217.0,
|
| 1844 |
+
"completions/min_terminated_length": 217.0,
|
| 1845 |
+
"epoch": 4.113411341134113,
|
| 1846 |
+
"grad_norm": 0.15785960853099823,
|
| 1847 |
+
"learning_rate": 9.549150281252632e-08,
|
| 1848 |
+
"loss": 0.0099,
|
| 1849 |
+
"num_tokens": 257255212.0,
|
| 1850 |
+
"reward": 1.6709603309631347,
|
| 1851 |
+
"reward_std": 0.26430340111255646,
|
| 1852 |
+
"rewards/accuracy_reward/mean": 0.7837053656578064,
|
| 1853 |
+
"rewards/accuracy_reward/std": 0.41143102645874025,
|
| 1854 |
+
"rewards/cosine_scaled_reward/mean": 0.7803134202957154,
|
| 1855 |
+
"rewards/cosine_scaled_reward/std": 0.41151973605155945,
|
| 1856 |
+
"rewards/format_reward/mean": 0.9941964387893677,
|
| 1857 |
+
"rewards/format_reward/std": 0.07270083874464035,
|
| 1858 |
+
"step": 325
|
| 1859 |
+
},
|
| 1860 |
+
{
|
| 1861 |
+
"clip_ratio/high_max": 0.0,
|
| 1862 |
+
"clip_ratio/high_mean": 0.0,
|
| 1863 |
+
"clip_ratio/low_mean": 0.0,
|
| 1864 |
+
"clip_ratio/low_min": 0.0,
|
| 1865 |
+
"clip_ratio/region_mean": 0.0,
|
| 1866 |
+
"completions/clipped_ratio": 0.00714285714285714,
|
| 1867 |
+
"completions/max_length": 2048.0,
|
| 1868 |
+
"completions/max_terminated_length": 1994.2,
|
| 1869 |
+
"completions/mean_length": 737.9998168945312,
|
| 1870 |
+
"completions/mean_terminated_length": 728.5752319335937,
|
| 1871 |
+
"completions/min_length": 211.4,
|
| 1872 |
+
"completions/min_terminated_length": 211.4,
|
| 1873 |
+
"epoch": 4.176417641764177,
|
| 1874 |
+
"grad_norm": 0.15280981361865997,
|
| 1875 |
+
"learning_rate": 8.288757056966017e-08,
|
| 1876 |
+
"loss": 0.0079,
|
| 1877 |
+
"num_tokens": 261348219.0,
|
| 1878 |
+
"reward": 1.6478198051452637,
|
| 1879 |
+
"reward_std": 0.25256154835224154,
|
| 1880 |
+
"rewards/accuracy_reward/mean": 0.7685267806053162,
|
| 1881 |
+
"rewards/accuracy_reward/std": 0.4206309378147125,
|
| 1882 |
+
"rewards/cosine_scaled_reward/mean": 0.7661752223968505,
|
| 1883 |
+
"rewards/cosine_scaled_reward/std": 0.4206216633319855,
|
| 1884 |
+
"rewards/format_reward/mean": 0.9924107193946838,
|
| 1885 |
+
"rewards/format_reward/std": 0.08273022547364235,
|
| 1886 |
+
"step": 330
|
| 1887 |
+
},
|
| 1888 |
+
{
|
| 1889 |
+
"clip_ratio/high_max": 0.0,
|
| 1890 |
+
"clip_ratio/high_mean": 0.0,
|
| 1891 |
+
"clip_ratio/low_mean": 0.0,
|
| 1892 |
+
"clip_ratio/low_min": 0.0,
|
| 1893 |
+
"clip_ratio/region_mean": 0.0,
|
| 1894 |
+
"completions/clipped_ratio": 0.006250000000000022,
|
| 1895 |
+
"completions/max_length": 2048.0,
|
| 1896 |
+
"completions/max_terminated_length": 1988.0,
|
| 1897 |
+
"completions/mean_length": 742.7562866210938,
|
| 1898 |
+
"completions/mean_terminated_length": 734.589794921875,
|
| 1899 |
+
"completions/min_length": 242.2,
|
| 1900 |
+
"completions/min_terminated_length": 242.2,
|
| 1901 |
+
"epoch": 4.2394239423942395,
|
| 1902 |
+
"grad_norm": 0.1629822999238968,
|
| 1903 |
+
"learning_rate": 7.110015549502546e-08,
|
| 1904 |
+
"loss": 0.0115,
|
| 1905 |
+
"num_tokens": 265491519.0,
|
| 1906 |
+
"reward": 1.6362776041030884,
|
| 1907 |
+
"reward_std": 0.2605998367071152,
|
| 1908 |
+
"rewards/accuracy_reward/mean": 0.7607142925262451,
|
| 1909 |
+
"rewards/accuracy_reward/std": 0.4256579399108887,
|
| 1910 |
+
"rewards/cosine_scaled_reward/mean": 0.7573764562606812,
|
| 1911 |
+
"rewards/cosine_scaled_reward/std": 0.42556257247924806,
|
| 1912 |
+
"rewards/format_reward/mean": 0.9937500119209289,
|
| 1913 |
+
"rewards/format_reward/std": 0.07603473961353302,
|
| 1914 |
+
"step": 335
|
| 1915 |
+
},
|
| 1916 |
+
{
|
| 1917 |
+
"clip_ratio/high_max": 0.0,
|
| 1918 |
+
"clip_ratio/high_mean": 0.0,
|
| 1919 |
+
"clip_ratio/low_mean": 0.0,
|
| 1920 |
+
"clip_ratio/low_min": 0.0,
|
| 1921 |
+
"clip_ratio/region_mean": 0.0,
|
| 1922 |
+
"completions/clipped_ratio": 0.0020089285714285587,
|
| 1923 |
+
"completions/max_length": 2048.0,
|
| 1924 |
+
"completions/max_terminated_length": 1959.6,
|
| 1925 |
+
"completions/mean_length": 739.0100830078125,
|
| 1926 |
+
"completions/mean_terminated_length": 736.3754638671875,
|
| 1927 |
+
"completions/min_length": 215.8,
|
| 1928 |
+
"completions/min_terminated_length": 215.8,
|
| 1929 |
+
"epoch": 4.302430243024302,
|
| 1930 |
+
"grad_norm": 0.1339646428823471,
|
| 1931 |
+
"learning_rate": 6.015233200716874e-08,
|
| 1932 |
+
"loss": 0.0117,
|
| 1933 |
+
"num_tokens": 269585804.0,
|
| 1934 |
+
"reward": 1.6636471033096314,
|
| 1935 |
+
"reward_std": 0.25507272481918336,
|
| 1936 |
+
"rewards/accuracy_reward/mean": 0.7779017925262451,
|
| 1937 |
+
"rewards/accuracy_reward/std": 0.4153802514076233,
|
| 1938 |
+
"rewards/cosine_scaled_reward/mean": 0.7737227082252502,
|
| 1939 |
+
"rewards/cosine_scaled_reward/std": 0.4154580056667328,
|
| 1940 |
+
"rewards/format_reward/mean": 0.9977678418159485,
|
| 1941 |
+
"rewards/format_reward/std": 0.044395358115434644,
|
| 1942 |
+
"step": 340
|
| 1943 |
+
},
|
| 1944 |
+
{
|
| 1945 |
+
"clip_ratio/high_max": 0.0,
|
| 1946 |
+
"clip_ratio/high_mean": 0.0,
|
| 1947 |
+
"clip_ratio/low_mean": 0.0,
|
| 1948 |
+
"clip_ratio/low_min": 0.0,
|
| 1949 |
+
"clip_ratio/region_mean": 0.0,
|
| 1950 |
+
"completions/clipped_ratio": 0.004464285714285721,
|
| 1951 |
+
"completions/max_length": 2048.0,
|
| 1952 |
+
"completions/max_terminated_length": 1967.8,
|
| 1953 |
+
"completions/mean_length": 720.30205078125,
|
| 1954 |
+
"completions/mean_terminated_length": 714.4131469726562,
|
| 1955 |
+
"completions/min_length": 203.8,
|
| 1956 |
+
"completions/min_terminated_length": 203.8,
|
| 1957 |
+
"epoch": 4.365436543654366,
|
| 1958 |
+
"grad_norm": 0.18192198872566223,
|
| 1959 |
+
"learning_rate": 5.006553098467764e-08,
|
| 1960 |
+
"loss": 0.0152,
|
| 1961 |
+
"num_tokens": 273613733.0,
|
| 1962 |
+
"reward": 1.6808908224105834,
|
| 1963 |
+
"reward_std": 0.24185148477554322,
|
| 1964 |
+
"rewards/accuracy_reward/mean": 0.7897321343421936,
|
| 1965 |
+
"rewards/accuracy_reward/std": 0.4066182732582092,
|
| 1966 |
+
"rewards/cosine_scaled_reward/mean": 0.7867815136909485,
|
| 1967 |
+
"rewards/cosine_scaled_reward/std": 0.4068154990673065,
|
| 1968 |
+
"rewards/format_reward/mean": 0.9955357074737549,
|
| 1969 |
+
"rewards/format_reward/std": 0.06201168373227119,
|
| 1970 |
+
"step": 345
|
| 1971 |
+
},
|
| 1972 |
+
{
|
| 1973 |
+
"clip_ratio/high_max": 0.0,
|
| 1974 |
+
"clip_ratio/high_mean": 0.0,
|
| 1975 |
+
"clip_ratio/low_mean": 0.0,
|
| 1976 |
+
"clip_ratio/low_min": 0.0,
|
| 1977 |
+
"clip_ratio/region_mean": 0.0,
|
| 1978 |
+
"completions/clipped_ratio": 0.003571428571428581,
|
| 1979 |
+
"completions/max_length": 2048.0,
|
| 1980 |
+
"completions/max_terminated_length": 1949.0,
|
| 1981 |
+
"completions/mean_length": 732.9803833007812,
|
| 1982 |
+
"completions/mean_terminated_length": 728.2846801757812,
|
| 1983 |
+
"completions/min_length": 239.2,
|
| 1984 |
+
"completions/min_terminated_length": 239.2,
|
| 1985 |
+
"epoch": 4.428442844284429,
|
| 1986 |
+
"grad_norm": 0.16086089611053467,
|
| 1987 |
+
"learning_rate": 4.085949781423048e-08,
|
| 1988 |
+
"loss": 0.0097,
|
| 1989 |
+
"num_tokens": 277698637.0,
|
| 1990 |
+
"reward": 1.680655264854431,
|
| 1991 |
+
"reward_std": 0.23921382129192353,
|
| 1992 |
+
"rewards/accuracy_reward/mean": 0.7897321462631226,
|
| 1993 |
+
"rewards/accuracy_reward/std": 0.4064399242401123,
|
| 1994 |
+
"rewards/cosine_scaled_reward/mean": 0.7863104462623596,
|
| 1995 |
+
"rewards/cosine_scaled_reward/std": 0.4065011143684387,
|
| 1996 |
+
"rewards/format_reward/mean": 0.9955357193946839,
|
| 1997 |
+
"rewards/format_reward/std": 0.06422311514616012,
|
| 1998 |
+
"step": 350
|
| 1999 |
+
},
|
| 2000 |
+
{
|
| 2001 |
+
"clip_ratio/high_max": 0.0,
|
| 2002 |
+
"clip_ratio/high_mean": 0.0,
|
| 2003 |
+
"clip_ratio/low_mean": 0.0,
|
| 2004 |
+
"clip_ratio/low_min": 0.0,
|
| 2005 |
+
"clip_ratio/region_mean": 0.0,
|
| 2006 |
+
"completions/clipped_ratio": 0.004464285714285721,
|
| 2007 |
+
"completions/max_length": 2033.4,
|
| 2008 |
+
"completions/max_terminated_length": 1924.2,
|
| 2009 |
+
"completions/mean_length": 729.4888793945313,
|
| 2010 |
+
"completions/mean_terminated_length": 723.6002075195313,
|
| 2011 |
+
"completions/min_length": 190.2,
|
| 2012 |
+
"completions/min_terminated_length": 190.2,
|
| 2013 |
+
"epoch": 4.491449144914491,
|
| 2014 |
+
"grad_norm": 0.1604759842157364,
|
| 2015 |
+
"learning_rate": 3.2552253738074074e-08,
|
| 2016 |
+
"loss": 0.011,
|
| 2017 |
+
"num_tokens": 281751955.0,
|
| 2018 |
+
"reward": 1.64648494720459,
|
| 2019 |
+
"reward_std": 0.2558329701423645,
|
| 2020 |
+
"rewards/accuracy_reward/mean": 0.7671875,
|
| 2021 |
+
"rewards/accuracy_reward/std": 0.42172677516937257,
|
| 2022 |
+
"rewards/cosine_scaled_reward/mean": 0.7648447513580322,
|
| 2023 |
+
"rewards/cosine_scaled_reward/std": 0.4217764616012573,
|
| 2024 |
+
"rewards/format_reward/mean": 0.99375,
|
| 2025 |
+
"rewards/format_reward/std": 0.0674091026186943,
|
| 2026 |
+
"step": 355
|
| 2027 |
+
},
|
| 2028 |
+
{
|
| 2029 |
+
"clip_ratio/high_max": 0.0,
|
| 2030 |
+
"clip_ratio/high_mean": 0.0,
|
| 2031 |
+
"clip_ratio/low_mean": 0.0,
|
| 2032 |
+
"clip_ratio/low_min": 0.0,
|
| 2033 |
+
"clip_ratio/region_mean": 0.0,
|
| 2034 |
+
"completions/clipped_ratio": 0.004017857142857117,
|
| 2035 |
+
"completions/max_length": 2048.0,
|
| 2036 |
+
"completions/max_terminated_length": 1914.6,
|
| 2037 |
+
"completions/mean_length": 726.2551635742187,
|
| 2038 |
+
"completions/mean_terminated_length": 720.9600830078125,
|
| 2039 |
+
"completions/min_length": 226.0,
|
| 2040 |
+
"completions/min_terminated_length": 226.0,
|
| 2041 |
+
"epoch": 4.554455445544555,
|
| 2042 |
+
"grad_norm": 0.16528210043907166,
|
| 2043 |
+
"learning_rate": 2.5160060576596365e-08,
|
| 2044 |
+
"loss": 0.0093,
|
| 2045 |
+
"num_tokens": 285784210.0,
|
| 2046 |
+
"reward": 1.644165849685669,
|
| 2047 |
+
"reward_std": 0.2550088167190552,
|
| 2048 |
+
"rewards/accuracy_reward/mean": 0.7649553537368774,
|
| 2049 |
+
"rewards/accuracy_reward/std": 0.4207766890525818,
|
| 2050 |
+
"rewards/cosine_scaled_reward/mean": 0.7624387383460999,
|
| 2051 |
+
"rewards/cosine_scaled_reward/std": 0.42074838280677795,
|
| 2052 |
+
"rewards/format_reward/mean": 0.9959821343421936,
|
| 2053 |
+
"rewards/format_reward/std": 0.06139585971832275,
|
| 2054 |
+
"step": 360
|
| 2055 |
+
},
|
| 2056 |
+
{
|
| 2057 |
+
"clip_ratio/high_max": 0.0,
|
| 2058 |
+
"clip_ratio/high_mean": 0.0,
|
| 2059 |
+
"clip_ratio/low_mean": 0.0,
|
| 2060 |
+
"clip_ratio/low_min": 0.0,
|
| 2061 |
+
"clip_ratio/region_mean": 0.0,
|
| 2062 |
+
"completions/clipped_ratio": 0.00758928571428572,
|
| 2063 |
+
"completions/max_length": 2048.0,
|
| 2064 |
+
"completions/max_terminated_length": 2005.2,
|
| 2065 |
+
"completions/mean_length": 734.5837524414062,
|
| 2066 |
+
"completions/mean_terminated_length": 724.657373046875,
|
| 2067 |
+
"completions/min_length": 197.8,
|
| 2068 |
+
"completions/min_terminated_length": 197.8,
|
| 2069 |
+
"epoch": 4.617461746174618,
|
| 2070 |
+
"grad_norm": 0.15532474219799042,
|
| 2071 |
+
"learning_rate": 1.8697388895048515e-08,
|
| 2072 |
+
"loss": 0.008,
|
| 2073 |
+
"num_tokens": 289916185.0,
|
| 2074 |
+
"reward": 1.6304625272750854,
|
| 2075 |
+
"reward_std": 0.2597217381000519,
|
| 2076 |
+
"rewards/accuracy_reward/mean": 0.7575892806053162,
|
| 2077 |
+
"rewards/accuracy_reward/std": 0.42815070748329165,
|
| 2078 |
+
"rewards/cosine_scaled_reward/mean": 0.7535588622093201,
|
| 2079 |
+
"rewards/cosine_scaled_reward/std": 0.4280454754829407,
|
| 2080 |
+
"rewards/format_reward/mean": 0.9921875,
|
| 2081 |
+
"rewards/format_reward/std": 0.0802665263414383,
|
| 2082 |
+
"step": 365
|
| 2083 |
+
},
|
| 2084 |
+
{
|
| 2085 |
+
"clip_ratio/high_max": 0.0,
|
| 2086 |
+
"clip_ratio/high_mean": 0.0,
|
| 2087 |
+
"clip_ratio/low_mean": 0.0,
|
| 2088 |
+
"clip_ratio/low_min": 0.0,
|
| 2089 |
+
"clip_ratio/region_mean": 0.0,
|
| 2090 |
+
"completions/clipped_ratio": 0.0020089285714285808,
|
| 2091 |
+
"completions/max_length": 2038.2,
|
| 2092 |
+
"completions/max_terminated_length": 1860.0,
|
| 2093 |
+
"completions/mean_length": 702.7852905273437,
|
| 2094 |
+
"completions/mean_terminated_length": 700.0497802734375,
|
| 2095 |
+
"completions/min_length": 239.2,
|
| 2096 |
+
"completions/min_terminated_length": 239.2,
|
| 2097 |
+
"epoch": 4.68046804680468,
|
| 2098 |
+
"grad_norm": 0.1871415078639984,
|
| 2099 |
+
"learning_rate": 1.3176889676734692e-08,
|
| 2100 |
+
"loss": 0.0091,
|
| 2101 |
+
"num_tokens": 293819303.0,
|
| 2102 |
+
"reward": 1.6794666051864624,
|
| 2103 |
+
"reward_std": 0.26247662901878355,
|
| 2104 |
+
"rewards/accuracy_reward/mean": 0.7879464268684387,
|
| 2105 |
+
"rewards/accuracy_reward/std": 0.408376944065094,
|
| 2106 |
+
"rewards/cosine_scaled_reward/mean": 0.7854955077171326,
|
| 2107 |
+
"rewards/cosine_scaled_reward/std": 0.4085055232048035,
|
| 2108 |
+
"rewards/format_reward/mean": 0.9975446343421936,
|
| 2109 |
+
"rewards/format_reward/std": 0.0441232718527317,
|
| 2110 |
+
"step": 370
|
| 2111 |
+
},
|
| 2112 |
+
{
|
| 2113 |
+
"clip_ratio/high_max": 0.0,
|
| 2114 |
+
"clip_ratio/high_mean": 0.0,
|
| 2115 |
+
"clip_ratio/low_mean": 0.0,
|
| 2116 |
+
"clip_ratio/low_min": 0.0,
|
| 2117 |
+
"clip_ratio/region_mean": 0.0,
|
| 2118 |
+
"completions/clipped_ratio": 0.005580357142857162,
|
| 2119 |
+
"completions/max_length": 2048.0,
|
| 2120 |
+
"completions/max_terminated_length": 1976.2,
|
| 2121 |
+
"completions/mean_length": 738.7837280273437,
|
| 2122 |
+
"completions/mean_terminated_length": 731.4327758789062,
|
| 2123 |
+
"completions/min_length": 235.6,
|
| 2124 |
+
"completions/min_terminated_length": 235.6,
|
| 2125 |
+
"epoch": 4.743474347434743,
|
| 2126 |
+
"grad_norm": 0.17294685542583466,
|
| 2127 |
+
"learning_rate": 8.60936955811753e-09,
|
| 2128 |
+
"loss": 0.0117,
|
| 2129 |
+
"num_tokens": 297898502.0,
|
| 2130 |
+
"reward": 1.675231122970581,
|
| 2131 |
+
"reward_std": 0.25796766579151154,
|
| 2132 |
+
"rewards/accuracy_reward/mean": 0.7866071343421936,
|
| 2133 |
+
"rewards/accuracy_reward/std": 0.408166778087616,
|
| 2134 |
+
"rewards/cosine_scaled_reward/mean": 0.7832746028900146,
|
| 2135 |
+
"rewards/cosine_scaled_reward/std": 0.40820558071136476,
|
| 2136 |
+
"rewards/format_reward/mean": 0.9939732074737548,
|
| 2137 |
+
"rewards/format_reward/std": 0.07510094344615936,
|
| 2138 |
+
"step": 375
|
| 2139 |
+
},
|
| 2140 |
+
{
|
| 2141 |
+
"clip_ratio/high_max": 0.0,
|
| 2142 |
+
"clip_ratio/high_mean": 0.0,
|
| 2143 |
+
"clip_ratio/low_mean": 0.0,
|
| 2144 |
+
"clip_ratio/low_min": 0.0,
|
| 2145 |
+
"clip_ratio/region_mean": 0.0,
|
| 2146 |
+
"completions/clipped_ratio": 0.004464285714285721,
|
| 2147 |
+
"completions/max_length": 2048.0,
|
| 2148 |
+
"completions/max_terminated_length": 1943.2,
|
| 2149 |
+
"completions/mean_length": 707.327490234375,
|
| 2150 |
+
"completions/mean_terminated_length": 701.3208129882812,
|
| 2151 |
+
"completions/min_length": 241.4,
|
| 2152 |
+
"completions/min_terminated_length": 241.4,
|
| 2153 |
+
"epoch": 4.806480648064807,
|
| 2154 |
+
"grad_norm": 0.1627655327320099,
|
| 2155 |
+
"learning_rate": 5.003769674320368e-09,
|
| 2156 |
+
"loss": 0.0098,
|
| 2157 |
+
"num_tokens": 301826889.0,
|
| 2158 |
+
"reward": 1.6730069637298584,
|
| 2159 |
+
"reward_std": 0.25484800338745117,
|
| 2160 |
+
"rewards/accuracy_reward/mean": 0.7845982193946839,
|
| 2161 |
+
"rewards/accuracy_reward/std": 0.40905932784080506,
|
| 2162 |
+
"rewards/cosine_scaled_reward/mean": 0.7815048575401307,
|
| 2163 |
+
"rewards/cosine_scaled_reward/std": 0.40920759439468385,
|
| 2164 |
+
"rewards/format_reward/mean": 0.995312488079071,
|
| 2165 |
+
"rewards/format_reward/std": 0.06412766650319099,
|
| 2166 |
+
"step": 380
|
| 2167 |
+
},
|
| 2168 |
+
{
|
| 2169 |
+
"clip_ratio/high_max": 0.0,
|
| 2170 |
+
"clip_ratio/high_mean": 0.0,
|
| 2171 |
+
"clip_ratio/low_mean": 0.0,
|
| 2172 |
+
"clip_ratio/low_min": 0.0,
|
| 2173 |
+
"clip_ratio/region_mean": 0.0,
|
| 2174 |
+
"completions/clipped_ratio": 0.006026785714285721,
|
| 2175 |
+
"completions/max_length": 2048.0,
|
| 2176 |
+
"completions/max_terminated_length": 1974.2,
|
| 2177 |
+
"completions/mean_length": 738.5777099609375,
|
| 2178 |
+
"completions/mean_terminated_length": 730.669482421875,
|
| 2179 |
+
"completions/min_length": 202.4,
|
| 2180 |
+
"completions/min_terminated_length": 202.4,
|
| 2181 |
+
"epoch": 4.8694869486948695,
|
| 2182 |
+
"grad_norm": 0.15702560544013977,
|
| 2183 |
+
"learning_rate": 2.3671481564347396e-09,
|
| 2184 |
+
"loss": 0.0108,
|
| 2185 |
+
"num_tokens": 305901773.0,
|
| 2186 |
+
"reward": 1.666380262374878,
|
| 2187 |
+
"reward_std": 0.25681224167346955,
|
| 2188 |
+
"rewards/accuracy_reward/mean": 0.7810267686843873,
|
| 2189 |
+
"rewards/accuracy_reward/std": 0.41189714074134826,
|
| 2190 |
+
"rewards/cosine_scaled_reward/mean": 0.7774031162261963,
|
| 2191 |
+
"rewards/cosine_scaled_reward/std": 0.4118649184703827,
|
| 2192 |
+
"rewards/format_reward/mean": 0.9933035731315613,
|
| 2193 |
+
"rewards/format_reward/std": 0.07954220175743103,
|
| 2194 |
+
"step": 385
|
| 2195 |
+
},
|
| 2196 |
+
{
|
| 2197 |
+
"clip_ratio/high_max": 0.0,
|
| 2198 |
+
"clip_ratio/high_mean": 0.0,
|
| 2199 |
+
"clip_ratio/low_mean": 0.0,
|
| 2200 |
+
"clip_ratio/low_min": 0.0,
|
| 2201 |
+
"clip_ratio/region_mean": 0.0,
|
| 2202 |
+
"completions/clipped_ratio": 0.004241071428571441,
|
| 2203 |
+
"completions/max_length": 2048.0,
|
| 2204 |
+
"completions/max_terminated_length": 1968.4,
|
| 2205 |
+
"completions/mean_length": 745.6741455078125,
|
| 2206 |
+
"completions/mean_terminated_length": 740.12138671875,
|
| 2207 |
+
"completions/min_length": 237.4,
|
| 2208 |
+
"completions/min_terminated_length": 237.4,
|
| 2209 |
+
"epoch": 4.932493249324932,
|
| 2210 |
+
"grad_norm": 0.17807839810848236,
|
| 2211 |
+
"learning_rate": 7.046663148979615e-10,
|
| 2212 |
+
"loss": 0.0081,
|
| 2213 |
+
"num_tokens": 310014129.0,
|
| 2214 |
+
"reward": 1.6689030408859253,
|
| 2215 |
+
"reward_std": 0.2475011855363846,
|
| 2216 |
+
"rewards/accuracy_reward/mean": 0.7819196343421936,
|
| 2217 |
+
"rewards/accuracy_reward/std": 0.41213822960853574,
|
| 2218 |
+
"rewards/cosine_scaled_reward/mean": 0.7782077431678772,
|
| 2219 |
+
"rewards/cosine_scaled_reward/std": 0.4122960567474365,
|
| 2220 |
+
"rewards/format_reward/mean": 0.9957589387893677,
|
| 2221 |
+
"rewards/format_reward/std": 0.06400550231337547,
|
| 2222 |
+
"step": 390
|
| 2223 |
+
},
|
| 2224 |
+
{
|
| 2225 |
+
"clip_ratio/high_max": 0.0,
|
| 2226 |
+
"clip_ratio/high_mean": 0.0,
|
| 2227 |
+
"clip_ratio/low_mean": 0.0,
|
| 2228 |
+
"clip_ratio/low_min": 0.0,
|
| 2229 |
+
"clip_ratio/region_mean": 0.0,
|
| 2230 |
+
"completions/clipped_ratio": 0.0043580963938973484,
|
| 2231 |
+
"completions/max_length": 2048.0,
|
| 2232 |
+
"completions/max_terminated_length": 1954.6,
|
| 2233 |
+
"completions/mean_length": 715.9204833984375,
|
| 2234 |
+
"completions/mean_terminated_length": 710.0726928710938,
|
| 2235 |
+
"completions/min_length": 241.8,
|
| 2236 |
+
"completions/min_terminated_length": 241.8,
|
| 2237 |
+
"epoch": 4.995499549954996,
|
| 2238 |
+
"grad_norm": 0.17002251744270325,
|
| 2239 |
+
"learning_rate": 1.95785359849987e-11,
|
| 2240 |
+
"loss": 0.0128,
|
| 2241 |
+
"num_tokens": 314023514.0,
|
| 2242 |
+
"reward": 1.6619688987731933,
|
| 2243 |
+
"reward_std": 0.24724579453468323,
|
| 2244 |
+
"rewards/accuracy_reward/mean": 0.7774553656578064,
|
| 2245 |
+
"rewards/accuracy_reward/std": 0.4148886203765869,
|
| 2246 |
+
"rewards/cosine_scaled_reward/mean": 0.7746072888374329,
|
| 2247 |
+
"rewards/cosine_scaled_reward/std": 0.4147594332695007,
|
| 2248 |
+
"rewards/format_reward/mean": 0.9944196462631225,
|
| 2249 |
+
"rewards/format_reward/std": 0.07256761640310287,
|
| 2250 |
+
"step": 395
|
| 2251 |
+
},
|
| 2252 |
+
{
|
| 2253 |
+
"epoch": 4.995499549954996,
|
| 2254 |
+
"step": 395,
|
| 2255 |
+
"total_flos": 0.0,
|
| 2256 |
+
"train_loss": 0.012358659607228598,
|
| 2257 |
+
"train_runtime": 80233.6203,
|
| 2258 |
+
"train_samples_per_second": 0.554,
|
| 2259 |
+
"train_steps_per_second": 0.005
|
| 2260 |
+
}
|
| 2261 |
+
],
|
| 2262 |
+
"logging_steps": 5,
|
| 2263 |
+
"max_steps": 395,
|
| 2264 |
+
"num_input_tokens_seen": 314023514,
|
| 2265 |
+
"num_train_epochs": 5,
|
| 2266 |
+
"save_steps": 15,
|
| 2267 |
+
"stateful_callbacks": {
|
| 2268 |
+
"TrainerControl": {
|
| 2269 |
+
"args": {
|
| 2270 |
+
"should_epoch_stop": false,
|
| 2271 |
+
"should_evaluate": false,
|
| 2272 |
+
"should_log": false,
|
| 2273 |
+
"should_save": true,
|
| 2274 |
+
"should_training_stop": true
|
| 2275 |
+
},
|
| 2276 |
+
"attributes": {}
|
| 2277 |
+
}
|
| 2278 |
+
},
|
| 2279 |
+
"total_flos": 0.0,
|
| 2280 |
+
"train_batch_size": 8,
|
| 2281 |
+
"trial_name": null,
|
| 2282 |
+
"trial_params": null
|
| 2283 |
+
}
|