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-cosine0512-v1
|
| 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-cosine0512-v1
|
| 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-cosine0512-v1", 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/o0zilidi)
|
| 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": 0.011041237461038782,
|
| 4 |
+
"train_runtime": 79107.9261,
|
| 5 |
+
"train_samples": 8888,
|
| 6 |
+
"train_samples_per_second": 0.562,
|
| 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": 0.011041237461038782,
|
| 4 |
+
"train_runtime": 79107.9261,
|
| 5 |
+
"train_samples": 8888,
|
| 6 |
+
"train_samples_per_second": 0.562,
|
| 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": 0.0279017857142857,
|
| 19 |
+
"completions/max_length": 2048.0,
|
| 20 |
+
"completions/max_terminated_length": 1978.0,
|
| 21 |
+
"completions/mean_length": 538.3058471679688,
|
| 22 |
+
"completions/mean_terminated_length": 494.97357177734375,
|
| 23 |
+
"completions/min_length": 3.0,
|
| 24 |
+
"completions/min_terminated_length": 3.0,
|
| 25 |
+
"epoch": 0.012601260126012601,
|
| 26 |
+
"grad_norm": 0.602237343788147,
|
| 27 |
+
"learning_rate": 0.0,
|
| 28 |
+
"loss": 0.0039,
|
| 29 |
+
"num_tokens": 609906.0,
|
| 30 |
+
"reward": 0.2795497477054596,
|
| 31 |
+
"reward_std": 0.41905471682548523,
|
| 32 |
+
"rewards/accuracy_reward/mean": 0.1685267835855484,
|
| 33 |
+
"rewards/accuracy_reward/std": 0.3745425045490265,
|
| 34 |
+
"rewards/cosine_scaled_reward/mean": 0.1684744656085968,
|
| 35 |
+
"rewards/cosine_scaled_reward/std": 0.37442949414253235,
|
| 36 |
+
"rewards/format_reward/mean": 0.0535714291036129,
|
| 37 |
+
"rewards/format_reward/std": 0.2252957671880722,
|
| 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": 0.021484375,
|
| 47 |
+
"completions/max_length": 2048.0,
|
| 48 |
+
"completions/max_terminated_length": 1946.5,
|
| 49 |
+
"completions/mean_length": 523.0030899047852,
|
| 50 |
+
"completions/mean_terminated_length": 489.49188232421875,
|
| 51 |
+
"completions/min_length": 2.5,
|
| 52 |
+
"completions/min_terminated_length": 2.5,
|
| 53 |
+
"epoch": 0.063006300630063,
|
| 54 |
+
"grad_norm": 0.32099631428718567,
|
| 55 |
+
"learning_rate": 1e-07,
|
| 56 |
+
"loss": 0.0035,
|
| 57 |
+
"num_tokens": 3031821.0,
|
| 58 |
+
"reward": 0.2492716759443283,
|
| 59 |
+
"reward_std": 0.3989013135433197,
|
| 60 |
+
"rewards/accuracy_reward/mean": 0.1512276753783226,
|
| 61 |
+
"rewards/accuracy_reward/std": 0.3584018871188164,
|
| 62 |
+
"rewards/cosine_scaled_reward/mean": 0.15116609632968903,
|
| 63 |
+
"rewards/cosine_scaled_reward/std": 0.35826219618320465,
|
| 64 |
+
"rewards/format_reward/mean": 0.044921875931322575,
|
| 65 |
+
"rewards/format_reward/std": 0.20668350905179977,
|
| 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": 0.024107142857142883,
|
| 75 |
+
"completions/max_length": 2048.0,
|
| 76 |
+
"completions/max_terminated_length": 1979.6,
|
| 77 |
+
"completions/mean_length": 547.3105102539063,
|
| 78 |
+
"completions/mean_terminated_length": 510.2965393066406,
|
| 79 |
+
"completions/min_length": 4.8,
|
| 80 |
+
"completions/min_terminated_length": 4.8,
|
| 81 |
+
"epoch": 0.126012601260126,
|
| 82 |
+
"grad_norm": 0.4403175711631775,
|
| 83 |
+
"learning_rate": 2.25e-07,
|
| 84 |
+
"loss": 0.0027,
|
| 85 |
+
"num_tokens": 6168628.0,
|
| 86 |
+
"reward": 0.2597510188817978,
|
| 87 |
+
"reward_std": 0.39840335249900816,
|
| 88 |
+
"rewards/accuracy_reward/mean": 0.15892857015132905,
|
| 89 |
+
"rewards/accuracy_reward/std": 0.365299665927887,
|
| 90 |
+
"rewards/cosine_scaled_reward/mean": 0.15878771841526032,
|
| 91 |
+
"rewards/cosine_scaled_reward/std": 0.3649946331977844,
|
| 92 |
+
"rewards/format_reward/mean": 0.04285714328289032,
|
| 93 |
+
"rewards/format_reward/std": 0.2020670711994171,
|
| 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": 0.02589285714285714,
|
| 103 |
+
"completions/max_length": 2048.0,
|
| 104 |
+
"completions/max_terminated_length": 1994.4,
|
| 105 |
+
"completions/mean_length": 555.5024780273437,
|
| 106 |
+
"completions/mean_terminated_length": 515.7921813964844,
|
| 107 |
+
"completions/min_length": 6.2,
|
| 108 |
+
"completions/min_terminated_length": 6.2,
|
| 109 |
+
"epoch": 0.18901890189018902,
|
| 110 |
+
"grad_norm": 1.8242275714874268,
|
| 111 |
+
"learning_rate": 3.5e-07,
|
| 112 |
+
"loss": 0.0031,
|
| 113 |
+
"num_tokens": 9301535.0,
|
| 114 |
+
"reward": 0.27981547117233274,
|
| 115 |
+
"reward_std": 0.40177412033081056,
|
| 116 |
+
"rewards/accuracy_reward/mean": 0.16138392984867095,
|
| 117 |
+
"rewards/accuracy_reward/std": 0.36725109815597534,
|
| 118 |
+
"rewards/cosine_scaled_reward/mean": 0.16119343042373657,
|
| 119 |
+
"rewards/cosine_scaled_reward/std": 0.36687323451042175,
|
| 120 |
+
"rewards/format_reward/mean": 0.07566964402794837,
|
| 121 |
+
"rewards/format_reward/std": 0.2617058604955673,
|
| 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": 0.02209821428571428,
|
| 131 |
+
"completions/max_length": 2048.0,
|
| 132 |
+
"completions/max_terminated_length": 2017.0,
|
| 133 |
+
"completions/mean_length": 548.893994140625,
|
| 134 |
+
"completions/mean_terminated_length": 515.1586364746094,
|
| 135 |
+
"completions/min_length": 3.0,
|
| 136 |
+
"completions/min_terminated_length": 3.0,
|
| 137 |
+
"epoch": 0.252025202520252,
|
| 138 |
+
"grad_norm": 0.6227929592132568,
|
| 139 |
+
"learning_rate": 4.7499999999999995e-07,
|
| 140 |
+
"loss": -0.0047,
|
| 141 |
+
"num_tokens": 12415812.0,
|
| 142 |
+
"reward": 0.32004910707473755,
|
| 143 |
+
"reward_std": 0.4481643557548523,
|
| 144 |
+
"rewards/accuracy_reward/mean": 0.17611607313156127,
|
| 145 |
+
"rewards/accuracy_reward/std": 0.3801506757736206,
|
| 146 |
+
"rewards/cosine_scaled_reward/mean": 0.1758124738931656,
|
| 147 |
+
"rewards/cosine_scaled_reward/std": 0.37958953976631166,
|
| 148 |
+
"rewards/format_reward/mean": 0.11205357164144517,
|
| 149 |
+
"rewards/format_reward/std": 0.3129072666168213,
|
| 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": 0.02075892857142858,
|
| 159 |
+
"completions/max_length": 2048.0,
|
| 160 |
+
"completions/max_terminated_length": 1966.2,
|
| 161 |
+
"completions/mean_length": 565.0823852539063,
|
| 162 |
+
"completions/mean_terminated_length": 533.6526733398438,
|
| 163 |
+
"completions/min_length": 2.8,
|
| 164 |
+
"completions/min_terminated_length": 2.8,
|
| 165 |
+
"epoch": 0.31503150315031503,
|
| 166 |
+
"grad_norm": 0.38101691007614136,
|
| 167 |
+
"learning_rate": 6e-07,
|
| 168 |
+
"loss": -0.0097,
|
| 169 |
+
"num_tokens": 15816741.0,
|
| 170 |
+
"reward": 0.5020053923130036,
|
| 171 |
+
"reward_std": 0.6075942158699036,
|
| 172 |
+
"rewards/accuracy_reward/mean": 0.2622948855161667,
|
| 173 |
+
"rewards/accuracy_reward/std": 0.4395259916782379,
|
| 174 |
+
"rewards/cosine_scaled_reward/mean": 0.26026553511619566,
|
| 175 |
+
"rewards/cosine_scaled_reward/std": 0.43781105875968934,
|
| 176 |
+
"rewards/format_reward/mean": 0.22053571343421935,
|
| 177 |
+
"rewards/format_reward/std": 0.40703149437904357,
|
| 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": 0.017187499999999977,
|
| 187 |
+
"completions/max_length": 2048.0,
|
| 188 |
+
"completions/max_terminated_length": 1967.0,
|
| 189 |
+
"completions/mean_length": 578.9469116210937,
|
| 190 |
+
"completions/mean_terminated_length": 553.24716796875,
|
| 191 |
+
"completions/min_length": 6.8,
|
| 192 |
+
"completions/min_terminated_length": 6.8,
|
| 193 |
+
"epoch": 0.37803780378037805,
|
| 194 |
+
"grad_norm": 0.33144351840019226,
|
| 195 |
+
"learning_rate": 7.249999999999999e-07,
|
| 196 |
+
"loss": -0.0055,
|
| 197 |
+
"num_tokens": 19248647.0,
|
| 198 |
+
"reward": 0.7754053592681884,
|
| 199 |
+
"reward_std": 0.677242922782898,
|
| 200 |
+
"rewards/accuracy_reward/mean": 0.35334821343421935,
|
| 201 |
+
"rewards/accuracy_reward/std": 0.4751245081424713,
|
| 202 |
+
"rewards/cosine_scaled_reward/mean": 0.3530427932739258,
|
| 203 |
+
"rewards/cosine_scaled_reward/std": 0.4747708261013031,
|
| 204 |
+
"rewards/format_reward/mean": 0.4910714328289032,
|
| 205 |
+
"rewards/format_reward/std": 0.48938958048820497,
|
| 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": 0.014285714285714258,
|
| 215 |
+
"completions/max_length": 2048.0,
|
| 216 |
+
"completions/max_terminated_length": 1959.6,
|
| 217 |
+
"completions/mean_length": 608.9741455078125,
|
| 218 |
+
"completions/mean_terminated_length": 588.129736328125,
|
| 219 |
+
"completions/min_length": 12.0,
|
| 220 |
+
"completions/min_terminated_length": 12.0,
|
| 221 |
+
"epoch": 0.44104410441044106,
|
| 222 |
+
"grad_norm": 0.2479850798845291,
|
| 223 |
+
"learning_rate": 8.499999999999999e-07,
|
| 224 |
+
"loss": 0.0144,
|
| 225 |
+
"num_tokens": 22822291.0,
|
| 226 |
+
"reward": 1.1359302043914794,
|
| 227 |
+
"reward_std": 0.633991277217865,
|
| 228 |
+
"rewards/accuracy_reward/mean": 0.5194196462631225,
|
| 229 |
+
"rewards/accuracy_reward/std": 0.49612280130386355,
|
| 230 |
+
"rewards/cosine_scaled_reward/mean": 0.518735259771347,
|
| 231 |
+
"rewards/cosine_scaled_reward/std": 0.495624178647995,
|
| 232 |
+
"rewards/format_reward/mean": 0.7142857193946839,
|
| 233 |
+
"rewards/format_reward/std": 0.4489860713481903,
|
| 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": 0.0125,
|
| 243 |
+
"completions/max_length": 2048.0,
|
| 244 |
+
"completions/max_terminated_length": 1967.4,
|
| 245 |
+
"completions/mean_length": 672.230615234375,
|
| 246 |
+
"completions/mean_terminated_length": 654.8196044921875,
|
| 247 |
+
"completions/min_length": 70.6,
|
| 248 |
+
"completions/min_terminated_length": 70.6,
|
| 249 |
+
"epoch": 0.504050405040504,
|
| 250 |
+
"grad_norm": 0.19197942316532135,
|
| 251 |
+
"learning_rate": 9.75e-07,
|
| 252 |
+
"loss": 0.0261,
|
| 253 |
+
"num_tokens": 26671876.0,
|
| 254 |
+
"reward": 1.3178162574768066,
|
| 255 |
+
"reward_std": 0.5191608130931854,
|
| 256 |
+
"rewards/accuracy_reward/mean": 0.5977678537368775,
|
| 257 |
+
"rewards/accuracy_reward/std": 0.48948062062263487,
|
| 258 |
+
"rewards/cosine_scaled_reward/mean": 0.5967931866645813,
|
| 259 |
+
"rewards/cosine_scaled_reward/std": 0.4888714671134949,
|
| 260 |
+
"rewards/format_reward/mean": 0.8433035612106323,
|
| 261 |
+
"rewards/format_reward/std": 0.35585846900939944,
|
| 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": 0.01383928571428572,
|
| 271 |
+
"completions/max_length": 2048.0,
|
| 272 |
+
"completions/max_terminated_length": 2024.0,
|
| 273 |
+
"completions/mean_length": 691.8661010742187,
|
| 274 |
+
"completions/mean_terminated_length": 672.869873046875,
|
| 275 |
+
"completions/min_length": 127.6,
|
| 276 |
+
"completions/min_terminated_length": 127.6,
|
| 277 |
+
"epoch": 0.5670567056705671,
|
| 278 |
+
"grad_norm": 0.12823984026908875,
|
| 279 |
+
"learning_rate": 9.99686774088805e-07,
|
| 280 |
+
"loss": 0.0293,
|
| 281 |
+
"num_tokens": 30644708.0,
|
| 282 |
+
"reward": 1.4662254571914672,
|
| 283 |
+
"reward_std": 0.4329299807548523,
|
| 284 |
+
"rewards/accuracy_reward/mean": 0.6647321343421936,
|
| 285 |
+
"rewards/accuracy_reward/std": 0.47048051953315734,
|
| 286 |
+
"rewards/cosine_scaled_reward/mean": 0.6637008666992188,
|
| 287 |
+
"rewards/cosine_scaled_reward/std": 0.46995849609375,
|
| 288 |
+
"rewards/format_reward/mean": 0.9392857074737548,
|
| 289 |
+
"rewards/format_reward/std": 0.23625308275222778,
|
| 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": 0.01517857142857142,
|
| 299 |
+
"completions/max_length": 2048.0,
|
| 300 |
+
"completions/max_terminated_length": 2006.2,
|
| 301 |
+
"completions/mean_length": 724.4636352539062,
|
| 302 |
+
"completions/mean_terminated_length": 704.2110107421875,
|
| 303 |
+
"completions/min_length": 183.8,
|
| 304 |
+
"completions/min_terminated_length": 183.8,
|
| 305 |
+
"epoch": 0.6300630063006301,
|
| 306 |
+
"grad_norm": 0.11830825358629227,
|
| 307 |
+
"learning_rate": 9.984149663879992e-07,
|
| 308 |
+
"loss": 0.0277,
|
| 309 |
+
"num_tokens": 34704233.0,
|
| 310 |
+
"reward": 1.4932293176651001,
|
| 311 |
+
"reward_std": 0.39716021418571473,
|
| 312 |
+
"rewards/accuracy_reward/mean": 0.6720982074737549,
|
| 313 |
+
"rewards/accuracy_reward/std": 0.46862396597862244,
|
| 314 |
+
"rewards/cosine_scaled_reward/mean": 0.6715031266212463,
|
| 315 |
+
"rewards/cosine_scaled_reward/std": 0.4683354377746582,
|
| 316 |
+
"rewards/format_reward/mean": 0.9707589268684387,
|
| 317 |
+
"rewards/format_reward/std": 0.1669136554002762,
|
| 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": 0.009821428571428559,
|
| 327 |
+
"completions/max_length": 2048.0,
|
| 328 |
+
"completions/max_terminated_length": 1976.2,
|
| 329 |
+
"completions/mean_length": 713.6201171875,
|
| 330 |
+
"completions/mean_terminated_length": 700.3927490234375,
|
| 331 |
+
"completions/min_length": 167.4,
|
| 332 |
+
"completions/min_terminated_length": 167.4,
|
| 333 |
+
"epoch": 0.693069306930693,
|
| 334 |
+
"grad_norm": 0.10666404664516449,
|
| 335 |
+
"learning_rate": 9.961674879848956e-07,
|
| 336 |
+
"loss": 0.0233,
|
| 337 |
+
"num_tokens": 38681571.0,
|
| 338 |
+
"reward": 1.5286698818206788,
|
| 339 |
+
"reward_std": 0.36640719771385194,
|
| 340 |
+
"rewards/accuracy_reward/mean": 0.6915178418159484,
|
| 341 |
+
"rewards/accuracy_reward/std": 0.4611486494541168,
|
| 342 |
+
"rewards/cosine_scaled_reward/mean": 0.6908218264579773,
|
| 343 |
+
"rewards/cosine_scaled_reward/std": 0.4608599662780762,
|
| 344 |
+
"rewards/format_reward/mean": 0.9834821462631226,
|
| 345 |
+
"rewards/format_reward/std": 0.12500424087047576,
|
| 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": 0.012053571428571419,
|
| 355 |
+
"completions/max_length": 2048.0,
|
| 356 |
+
"completions/max_terminated_length": 1924.6,
|
| 357 |
+
"completions/mean_length": 711.6618530273438,
|
| 358 |
+
"completions/mean_terminated_length": 695.3810913085938,
|
| 359 |
+
"completions/min_length": 164.0,
|
| 360 |
+
"completions/min_terminated_length": 164.0,
|
| 361 |
+
"epoch": 0.7560756075607561,
|
| 362 |
+
"grad_norm": 0.1039816215634346,
|
| 363 |
+
"learning_rate": 9.929487384240101e-07,
|
| 364 |
+
"loss": 0.0233,
|
| 365 |
+
"num_tokens": 42628176.0,
|
| 366 |
+
"reward": 1.5549091339111327,
|
| 367 |
+
"reward_std": 0.3488038718700409,
|
| 368 |
+
"rewards/accuracy_reward/mean": 0.7089285612106323,
|
| 369 |
+
"rewards/accuracy_reward/std": 0.45444890260696413,
|
| 370 |
+
"rewards/cosine_scaled_reward/mean": 0.7075859427452087,
|
| 371 |
+
"rewards/cosine_scaled_reward/std": 0.4539263606071472,
|
| 372 |
+
"rewards/format_reward/mean": 0.984375,
|
| 373 |
+
"rewards/format_reward/std": 0.1233037456870079,
|
| 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": 0.01339285714285714,
|
| 383 |
+
"completions/max_length": 2048.0,
|
| 384 |
+
"completions/max_terminated_length": 1963.2,
|
| 385 |
+
"completions/mean_length": 713.0527099609375,
|
| 386 |
+
"completions/mean_terminated_length": 694.9294799804687,
|
| 387 |
+
"completions/min_length": 205.6,
|
| 388 |
+
"completions/min_terminated_length": 205.6,
|
| 389 |
+
"epoch": 0.819081908190819,
|
| 390 |
+
"grad_norm": 0.11131048947572708,
|
| 391 |
+
"learning_rate": 9.88765018558761e-07,
|
| 392 |
+
"loss": 0.0214,
|
| 393 |
+
"num_tokens": 46613412.0,
|
| 394 |
+
"reward": 1.553797197341919,
|
| 395 |
+
"reward_std": 0.34091995358467103,
|
| 396 |
+
"rewards/accuracy_reward/mean": 0.7078125,
|
| 397 |
+
"rewards/accuracy_reward/std": 0.45394955277442933,
|
| 398 |
+
"rewards/cosine_scaled_reward/mean": 0.7067014455795289,
|
| 399 |
+
"rewards/cosine_scaled_reward/std": 0.4534592032432556,
|
| 400 |
+
"rewards/format_reward/mean": 0.9852678537368774,
|
| 401 |
+
"rewards/format_reward/std": 0.11978632807731629,
|
| 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": 0.010267857142857162,
|
| 411 |
+
"completions/max_length": 2048.0,
|
| 412 |
+
"completions/max_terminated_length": 1950.2,
|
| 413 |
+
"completions/mean_length": 687.0343994140625,
|
| 414 |
+
"completions/mean_terminated_length": 672.891162109375,
|
| 415 |
+
"completions/min_length": 155.0,
|
| 416 |
+
"completions/min_terminated_length": 155.0,
|
| 417 |
+
"epoch": 0.8820882088208821,
|
| 418 |
+
"grad_norm": 0.11210864782333374,
|
| 419 |
+
"learning_rate": 9.836245182172519e-07,
|
| 420 |
+
"loss": 0.0192,
|
| 421 |
+
"num_tokens": 50425302.0,
|
| 422 |
+
"reward": 1.568692970275879,
|
| 423 |
+
"reward_std": 0.3520396828651428,
|
| 424 |
+
"rewards/accuracy_reward/mean": 0.7180039405822753,
|
| 425 |
+
"rewards/accuracy_reward/std": 0.44985153079032897,
|
| 426 |
+
"rewards/cosine_scaled_reward/mean": 0.7167718410491943,
|
| 427 |
+
"rewards/cosine_scaled_reward/std": 0.44936810731887816,
|
| 428 |
+
"rewards/format_reward/mean": 0.9883928537368775,
|
| 429 |
+
"rewards/format_reward/std": 0.10585158914327622,
|
| 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": 0.007366071428571397,
|
| 439 |
+
"completions/max_length": 2048.0,
|
| 440 |
+
"completions/max_terminated_length": 1996.0,
|
| 441 |
+
"completions/mean_length": 702.3578369140625,
|
| 442 |
+
"completions/mean_terminated_length": 692.3953125,
|
| 443 |
+
"completions/min_length": 167.2,
|
| 444 |
+
"completions/min_terminated_length": 167.2,
|
| 445 |
+
"epoch": 0.9450945094509451,
|
| 446 |
+
"grad_norm": 0.11102982610464096,
|
| 447 |
+
"learning_rate": 9.775373001702997e-07,
|
| 448 |
+
"loss": 0.0159,
|
| 449 |
+
"num_tokens": 54367609.0,
|
| 450 |
+
"reward": 1.5509341955184937,
|
| 451 |
+
"reward_std": 0.3260269582271576,
|
| 452 |
+
"rewards/accuracy_reward/mean": 0.7037946343421936,
|
| 453 |
+
"rewards/accuracy_reward/std": 0.45627689361572266,
|
| 454 |
+
"rewards/cosine_scaled_reward/mean": 0.7023146390914917,
|
| 455 |
+
"rewards/cosine_scaled_reward/std": 0.45560853481292723,
|
| 456 |
+
"rewards/format_reward/mean": 0.9919642925262451,
|
| 457 |
+
"rewards/format_reward/std": 0.08732087463140488,
|
| 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": 0.008453970180305114,
|
| 467 |
+
"completions/max_length": 2048.0,
|
| 468 |
+
"completions/max_terminated_length": 1951.4,
|
| 469 |
+
"completions/mean_length": 707.547412109375,
|
| 470 |
+
"completions/mean_terminated_length": 696.137548828125,
|
| 471 |
+
"completions/min_length": 175.0,
|
| 472 |
+
"completions/min_terminated_length": 175.0,
|
| 473 |
+
"epoch": 1.0126012601260126,
|
| 474 |
+
"grad_norm": 0.10652182251214981,
|
| 475 |
+
"learning_rate": 9.705152804330871e-07,
|
| 476 |
+
"loss": 0.0192,
|
| 477 |
+
"num_tokens": 58375847.0,
|
| 478 |
+
"reward": 1.5461820125579835,
|
| 479 |
+
"reward_std": 0.3652195274829865,
|
| 480 |
+
"rewards/accuracy_reward/mean": 0.7006696343421936,
|
| 481 |
+
"rewards/accuracy_reward/std": 0.4574929714202881,
|
| 482 |
+
"rewards/cosine_scaled_reward/mean": 0.6995066642761231,
|
| 483 |
+
"rewards/cosine_scaled_reward/std": 0.45710301399230957,
|
| 484 |
+
"rewards/format_reward/mean": 0.9915178656578064,
|
| 485 |
+
"rewards/format_reward/std": 0.0895774394273758,
|
| 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": 0.005580357142857162,
|
| 495 |
+
"completions/max_length": 2048.0,
|
| 496 |
+
"completions/max_terminated_length": 1953.8,
|
| 497 |
+
"completions/mean_length": 682.2417724609375,
|
| 498 |
+
"completions/mean_terminated_length": 674.6181396484375,
|
| 499 |
+
"completions/min_length": 172.0,
|
| 500 |
+
"completions/min_terminated_length": 172.0,
|
| 501 |
+
"epoch": 1.0756075607560756,
|
| 502 |
+
"grad_norm": 0.11581874638795853,
|
| 503 |
+
"learning_rate": 9.625722049390045e-07,
|
| 504 |
+
"loss": 0.0092,
|
| 505 |
+
"num_tokens": 62230690.0,
|
| 506 |
+
"reward": 1.5838431358337401,
|
| 507 |
+
"reward_std": 0.3170156180858612,
|
| 508 |
+
"rewards/accuracy_reward/mean": 0.7247767925262452,
|
| 509 |
+
"rewards/accuracy_reward/std": 0.4461586236953735,
|
| 510 |
+
"rewards/cosine_scaled_reward/mean": 0.723936116695404,
|
| 511 |
+
"rewards/cosine_scaled_reward/std": 0.4458396315574646,
|
| 512 |
+
"rewards/format_reward/mean": 0.9941964268684387,
|
| 513 |
+
"rewards/format_reward/std": 0.07385859042406082,
|
| 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": 0.010714285714285721,
|
| 523 |
+
"completions/max_length": 2048.0,
|
| 524 |
+
"completions/max_terminated_length": 2015.2,
|
| 525 |
+
"completions/mean_length": 713.357861328125,
|
| 526 |
+
"completions/mean_terminated_length": 698.86904296875,
|
| 527 |
+
"completions/min_length": 212.2,
|
| 528 |
+
"completions/min_terminated_length": 212.2,
|
| 529 |
+
"epoch": 1.1386138613861387,
|
| 530 |
+
"grad_norm": 0.10194957256317139,
|
| 531 |
+
"learning_rate": 9.537236226313388e-07,
|
| 532 |
+
"loss": 0.0201,
|
| 533 |
+
"num_tokens": 66250349.0,
|
| 534 |
+
"reward": 1.5542301416397095,
|
| 535 |
+
"reward_std": 0.33158175349235536,
|
| 536 |
+
"rewards/accuracy_reward/mean": 0.7066964268684387,
|
| 537 |
+
"rewards/accuracy_reward/std": 0.453377491235733,
|
| 538 |
+
"rewards/cosine_scaled_reward/mean": 0.7060048222541809,
|
| 539 |
+
"rewards/cosine_scaled_reward/std": 0.45305579900741577,
|
| 540 |
+
"rewards/format_reward/mean": 0.989062488079071,
|
| 541 |
+
"rewards/format_reward/std": 0.1036216139793396,
|
| 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": 0.00959821428571428,
|
| 551 |
+
"completions/max_length": 2048.0,
|
| 552 |
+
"completions/max_terminated_length": 1982.4,
|
| 553 |
+
"completions/mean_length": 696.29912109375,
|
| 554 |
+
"completions/mean_terminated_length": 683.2262084960937,
|
| 555 |
+
"completions/min_length": 205.0,
|
| 556 |
+
"completions/min_terminated_length": 205.0,
|
| 557 |
+
"epoch": 1.2016201620162017,
|
| 558 |
+
"grad_norm": 0.11849887669086456,
|
| 559 |
+
"learning_rate": 9.439868550254868e-07,
|
| 560 |
+
"loss": 0.0179,
|
| 561 |
+
"num_tokens": 70144897.0,
|
| 562 |
+
"reward": 1.5924715518951416,
|
| 563 |
+
"reward_std": 0.3465422332286835,
|
| 564 |
+
"rewards/accuracy_reward/mean": 0.7321428537368775,
|
| 565 |
+
"rewards/accuracy_reward/std": 0.4421205699443817,
|
| 566 |
+
"rewards/cosine_scaled_reward/mean": 0.7307018756866455,
|
| 567 |
+
"rewards/cosine_scaled_reward/std": 0.44154341220855714,
|
| 568 |
+
"rewards/format_reward/mean": 0.9899553656578064,
|
| 569 |
+
"rewards/format_reward/std": 0.09479753226041794,
|
| 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": 0.0046875,
|
| 579 |
+
"completions/max_length": 2023.8,
|
| 580 |
+
"completions/max_terminated_length": 1962.0,
|
| 581 |
+
"completions/mean_length": 685.8980224609375,
|
| 582 |
+
"completions/mean_terminated_length": 679.4762451171875,
|
| 583 |
+
"completions/min_length": 209.0,
|
| 584 |
+
"completions/min_terminated_length": 209.0,
|
| 585 |
+
"epoch": 1.2646264626462647,
|
| 586 |
+
"grad_norm": 0.11047360301017761,
|
| 587 |
+
"learning_rate": 9.333809623012763e-07,
|
| 588 |
+
"loss": 0.0137,
|
| 589 |
+
"num_tokens": 73963520.0,
|
| 590 |
+
"reward": 1.6014429330825806,
|
| 591 |
+
"reward_std": 0.3224481403827667,
|
| 592 |
+
"rewards/accuracy_reward/mean": 0.7363839387893677,
|
| 593 |
+
"rewards/accuracy_reward/std": 0.4394433259963989,
|
| 594 |
+
"rewards/cosine_scaled_reward/mean": 0.7348053455352783,
|
| 595 |
+
"rewards/cosine_scaled_reward/std": 0.4390797853469849,
|
| 596 |
+
"rewards/format_reward/mean": 0.995312511920929,
|
| 597 |
+
"rewards/format_reward/std": 0.060620172321796416,
|
| 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": 0.006026785714285698,
|
| 607 |
+
"completions/max_length": 2048.0,
|
| 608 |
+
"completions/max_terminated_length": 1955.6,
|
| 609 |
+
"completions/mean_length": 685.2373168945312,
|
| 610 |
+
"completions/mean_terminated_length": 677.0635864257813,
|
| 611 |
+
"completions/min_length": 165.2,
|
| 612 |
+
"completions/min_terminated_length": 165.2,
|
| 613 |
+
"epoch": 1.3276327632763276,
|
| 614 |
+
"grad_norm": 0.11911053210496902,
|
| 615 |
+
"learning_rate": 9.219267059917681e-07,
|
| 616 |
+
"loss": 0.0179,
|
| 617 |
+
"num_tokens": 77867639.0,
|
| 618 |
+
"reward": 1.5839551210403442,
|
| 619 |
+
"reward_std": 0.3121181547641754,
|
| 620 |
+
"rewards/accuracy_reward/mean": 0.725,
|
| 621 |
+
"rewards/accuracy_reward/std": 0.44613986015319823,
|
| 622 |
+
"rewards/cosine_scaled_reward/mean": 0.7243833780288697,
|
| 623 |
+
"rewards/cosine_scaled_reward/std": 0.4459415376186371,
|
| 624 |
+
"rewards/format_reward/mean": 0.9935267806053162,
|
| 625 |
+
"rewards/format_reward/std": 0.07562160044908524,
|
| 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": 0.0049107142857142795,
|
| 635 |
+
"completions/max_length": 1989.2,
|
| 636 |
+
"completions/max_terminated_length": 1903.0,
|
| 637 |
+
"completions/mean_length": 677.8884155273438,
|
| 638 |
+
"completions/mean_terminated_length": 671.1434326171875,
|
| 639 |
+
"completions/min_length": 192.4,
|
| 640 |
+
"completions/min_terminated_length": 192.4,
|
| 641 |
+
"epoch": 1.3906390639063906,
|
| 642 |
+
"grad_norm": 0.12140108644962311,
|
| 643 |
+
"learning_rate": 9.096465083415807e-07,
|
| 644 |
+
"loss": 0.0172,
|
| 645 |
+
"num_tokens": 81711467.0,
|
| 646 |
+
"reward": 1.6193986415863038,
|
| 647 |
+
"reward_std": 0.3027470111846924,
|
| 648 |
+
"rewards/accuracy_reward/mean": 0.7484375,
|
| 649 |
+
"rewards/accuracy_reward/std": 0.4324600279331207,
|
| 650 |
+
"rewards/cosine_scaled_reward/mean": 0.7475025296211243,
|
| 651 |
+
"rewards/cosine_scaled_reward/std": 0.43219619393348696,
|
| 652 |
+
"rewards/format_reward/mean": 0.9944196462631225,
|
| 653 |
+
"rewards/format_reward/std": 0.07256761714816093,
|
| 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": 0.00825892857142858,
|
| 663 |
+
"completions/max_length": 2032.2,
|
| 664 |
+
"completions/max_terminated_length": 1932.0,
|
| 665 |
+
"completions/mean_length": 692.1948974609375,
|
| 666 |
+
"completions/mean_terminated_length": 680.9073486328125,
|
| 667 |
+
"completions/min_length": 187.8,
|
| 668 |
+
"completions/min_terminated_length": 187.8,
|
| 669 |
+
"epoch": 1.4536453645364538,
|
| 670 |
+
"grad_norm": 0.13913048803806305,
|
| 671 |
+
"learning_rate": 8.96564408414293e-07,
|
| 672 |
+
"loss": 0.014,
|
| 673 |
+
"num_tokens": 85604932.0,
|
| 674 |
+
"reward": 1.5589273691177368,
|
| 675 |
+
"reward_std": 0.3327997088432312,
|
| 676 |
+
"rewards/accuracy_reward/mean": 0.7130610585212708,
|
| 677 |
+
"rewards/accuracy_reward/std": 0.4512468934059143,
|
| 678 |
+
"rewards/cosine_scaled_reward/mean": 0.70667005777359,
|
| 679 |
+
"rewards/cosine_scaled_reward/std": 0.45351189374923706,
|
| 680 |
+
"rewards/format_reward/mean": 0.9915178656578064,
|
| 681 |
+
"rewards/format_reward/std": 0.07997382432222366,
|
| 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": 0.005357142857142861,
|
| 691 |
+
"completions/max_length": 2048.0,
|
| 692 |
+
"completions/max_terminated_length": 1993.8,
|
| 693 |
+
"completions/mean_length": 657.3136474609375,
|
| 694 |
+
"completions/mean_terminated_length": 649.76474609375,
|
| 695 |
+
"completions/min_length": 174.0,
|
| 696 |
+
"completions/min_terminated_length": 174.0,
|
| 697 |
+
"epoch": 1.5166516651665165,
|
| 698 |
+
"grad_norm": 0.11739910393953323,
|
| 699 |
+
"learning_rate": 8.827060150348489e-07,
|
| 700 |
+
"loss": 0.0121,
|
| 701 |
+
"num_tokens": 89320697.0,
|
| 702 |
+
"reward": 1.6037935495376587,
|
| 703 |
+
"reward_std": 0.2636448621749878,
|
| 704 |
+
"rewards/accuracy_reward/mean": 0.7381696462631225,
|
| 705 |
+
"rewards/accuracy_reward/std": 0.4392316222190857,
|
| 706 |
+
"rewards/cosine_scaled_reward/mean": 0.7377209424972534,
|
| 707 |
+
"rewards/cosine_scaled_reward/std": 0.4390567898750305,
|
| 708 |
+
"rewards/format_reward/mean": 0.9935267925262451,
|
| 709 |
+
"rewards/format_reward/std": 0.07422313913702965,
|
| 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": 0.009821428571428559,
|
| 719 |
+
"completions/max_length": 2048.0,
|
| 720 |
+
"completions/max_terminated_length": 1931.4,
|
| 721 |
+
"completions/mean_length": 687.1069580078125,
|
| 722 |
+
"completions/mean_terminated_length": 673.6233276367187,
|
| 723 |
+
"completions/min_length": 177.8,
|
| 724 |
+
"completions/min_terminated_length": 177.8,
|
| 725 |
+
"epoch": 1.5796579657965797,
|
| 726 |
+
"grad_norm": 0.11269731819629669,
|
| 727 |
+
"learning_rate": 8.680984566590792e-07,
|
| 728 |
+
"loss": 0.0104,
|
| 729 |
+
"num_tokens": 93191536.0,
|
| 730 |
+
"reward": 1.6061813592910767,
|
| 731 |
+
"reward_std": 0.3020260453224182,
|
| 732 |
+
"rewards/accuracy_reward/mean": 0.7410714387893677,
|
| 733 |
+
"rewards/accuracy_reward/std": 0.4377245783805847,
|
| 734 |
+
"rewards/cosine_scaled_reward/mean": 0.7404875159263611,
|
| 735 |
+
"rewards/cosine_scaled_reward/std": 0.43749932646751405,
|
| 736 |
+
"rewards/format_reward/mean": 0.9897321462631226,
|
| 737 |
+
"rewards/format_reward/std": 0.10028065890073776,
|
| 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": 0.0078124999999999774,
|
| 747 |
+
"completions/max_length": 2048.0,
|
| 748 |
+
"completions/max_terminated_length": 1875.6,
|
| 749 |
+
"completions/mean_length": 655.5540405273438,
|
| 750 |
+
"completions/mean_terminated_length": 644.6221923828125,
|
| 751 |
+
"completions/min_length": 189.6,
|
| 752 |
+
"completions/min_terminated_length": 189.6,
|
| 753 |
+
"epoch": 1.6426642664266426,
|
| 754 |
+
"grad_norm": 0.13505014777183533,
|
| 755 |
+
"learning_rate": 8.527703282684765e-07,
|
| 756 |
+
"loss": 0.014,
|
| 757 |
+
"num_tokens": 96899722.0,
|
| 758 |
+
"reward": 1.6154436826705934,
|
| 759 |
+
"reward_std": 0.30025253891944886,
|
| 760 |
+
"rewards/accuracy_reward/mean": 0.7470982193946838,
|
| 761 |
+
"rewards/accuracy_reward/std": 0.43418333530426023,
|
| 762 |
+
"rewards/cosine_scaled_reward/mean": 0.7465121984481812,
|
| 763 |
+
"rewards/cosine_scaled_reward/std": 0.43400861620903014,
|
| 764 |
+
"rewards/format_reward/mean": 0.9901785850524902,
|
| 765 |
+
"rewards/format_reward/std": 0.09706691503524781,
|
| 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": 0.00401785714285714,
|
| 775 |
+
"completions/max_length": 2048.0,
|
| 776 |
+
"completions/max_terminated_length": 1906.4,
|
| 777 |
+
"completions/mean_length": 651.9241333007812,
|
| 778 |
+
"completions/mean_terminated_length": 646.2785034179688,
|
| 779 |
+
"completions/min_length": 181.4,
|
| 780 |
+
"completions/min_terminated_length": 181.4,
|
| 781 |
+
"epoch": 1.7056705670567056,
|
| 782 |
+
"grad_norm": 0.1649669110774994,
|
| 783 |
+
"learning_rate": 8.36751635394175e-07,
|
| 784 |
+
"loss": 0.0071,
|
| 785 |
+
"num_tokens": 100597958.0,
|
| 786 |
+
"reward": 1.621180534362793,
|
| 787 |
+
"reward_std": 0.27085027396678923,
|
| 788 |
+
"rewards/accuracy_reward/mean": 0.7497767925262451,
|
| 789 |
+
"rewards/accuracy_reward/std": 0.4303715705871582,
|
| 790 |
+
"rewards/cosine_scaled_reward/mean": 0.7490574598312378,
|
| 791 |
+
"rewards/cosine_scaled_reward/std": 0.43027289509773253,
|
| 792 |
+
"rewards/format_reward/mean": 0.993749988079071,
|
| 793 |
+
"rewards/format_reward/std": 0.07439046055078506,
|
| 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": 0.00625,
|
| 803 |
+
"completions/max_length": 1999.2,
|
| 804 |
+
"completions/max_terminated_length": 1840.8,
|
| 805 |
+
"completions/mean_length": 629.5085205078125,
|
| 806 |
+
"completions/mean_terminated_length": 620.7006591796875,
|
| 807 |
+
"completions/min_length": 158.0,
|
| 808 |
+
"completions/min_terminated_length": 158.0,
|
| 809 |
+
"epoch": 1.7686768676867688,
|
| 810 |
+
"grad_norm": 0.15633437037467957,
|
| 811 |
+
"learning_rate": 8.200737353797148e-07,
|
| 812 |
+
"loss": 0.0149,
|
| 813 |
+
"num_tokens": 104226908.0,
|
| 814 |
+
"reward": 1.6223199844360352,
|
| 815 |
+
"reward_std": 0.28274320662021635,
|
| 816 |
+
"rewards/accuracy_reward/mean": 0.7506696343421936,
|
| 817 |
+
"rewards/accuracy_reward/std": 0.43105919361114503,
|
| 818 |
+
"rewards/cosine_scaled_reward/mean": 0.7499970197677612,
|
| 819 |
+
"rewards/cosine_scaled_reward/std": 0.43102300763130186,
|
| 820 |
+
"rewards/format_reward/mean": 0.9933035612106323,
|
| 821 |
+
"rewards/format_reward/std": 0.06884920671582222,
|
| 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": 0.004241071428571441,
|
| 831 |
+
"completions/max_length": 2048.0,
|
| 832 |
+
"completions/max_terminated_length": 1935.0,
|
| 833 |
+
"completions/mean_length": 616.7654296875,
|
| 834 |
+
"completions/mean_terminated_length": 610.688623046875,
|
| 835 |
+
"completions/min_length": 171.4,
|
| 836 |
+
"completions/min_terminated_length": 171.4,
|
| 837 |
+
"epoch": 1.8316831683168315,
|
| 838 |
+
"grad_norm": 0.16371622681617737,
|
| 839 |
+
"learning_rate": 8.027692759975693e-07,
|
| 840 |
+
"loss": 0.0091,
|
| 841 |
+
"num_tokens": 107827689.0,
|
| 842 |
+
"reward": 1.5669083833694457,
|
| 843 |
+
"reward_std": 0.308967524766922,
|
| 844 |
+
"rewards/accuracy_reward/mean": 0.7129464268684387,
|
| 845 |
+
"rewards/accuracy_reward/std": 0.44984045028686526,
|
| 846 |
+
"rewards/cosine_scaled_reward/mean": 0.7126113176345825,
|
| 847 |
+
"rewards/cosine_scaled_reward/std": 0.44967737793922424,
|
| 848 |
+
"rewards/format_reward/mean": 0.995312488079071,
|
| 849 |
+
"rewards/format_reward/std": 0.06629456728696823,
|
| 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": 0.00379464285714286,
|
| 859 |
+
"completions/max_length": 2048.0,
|
| 860 |
+
"completions/max_terminated_length": 1910.4,
|
| 861 |
+
"completions/mean_length": 605.9803833007812,
|
| 862 |
+
"completions/mean_terminated_length": 600.4914794921875,
|
| 863 |
+
"completions/min_length": 137.0,
|
| 864 |
+
"completions/min_terminated_length": 137.0,
|
| 865 |
+
"epoch": 1.8946894689468947,
|
| 866 |
+
"grad_norm": 0.16951541602611542,
|
| 867 |
+
"learning_rate": 7.848721315395974e-07,
|
| 868 |
+
"loss": 0.0083,
|
| 869 |
+
"num_tokens": 111359681.0,
|
| 870 |
+
"reward": 1.6116993427276611,
|
| 871 |
+
"reward_std": 0.3007046103477478,
|
| 872 |
+
"rewards/accuracy_reward/mean": 0.7428571343421936,
|
| 873 |
+
"rewards/accuracy_reward/std": 0.43679211139678953,
|
| 874 |
+
"rewards/cosine_scaled_reward/mean": 0.7423717975616455,
|
| 875 |
+
"rewards/cosine_scaled_reward/std": 0.4366111934185028,
|
| 876 |
+
"rewards/format_reward/mean": 0.995312488079071,
|
| 877 |
+
"rewards/format_reward/std": 0.06531310081481934,
|
| 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": 0.00401785714285714,
|
| 887 |
+
"completions/max_length": 2048.0,
|
| 888 |
+
"completions/max_terminated_length": 1928.0,
|
| 889 |
+
"completions/mean_length": 610.7569458007813,
|
| 890 |
+
"completions/mean_terminated_length": 604.954296875,
|
| 891 |
+
"completions/min_length": 176.0,
|
| 892 |
+
"completions/min_terminated_length": 176.0,
|
| 893 |
+
"epoch": 1.9576957695769577,
|
| 894 |
+
"grad_norm": 0.17909295856952667,
|
| 895 |
+
"learning_rate": 7.664173365065265e-07,
|
| 896 |
+
"loss": 0.0103,
|
| 897 |
+
"num_tokens": 114894120.0,
|
| 898 |
+
"reward": 1.6185973644256593,
|
| 899 |
+
"reward_std": 0.29384649991989137,
|
| 900 |
+
"rewards/accuracy_reward/mean": 0.7475446581840515,
|
| 901 |
+
"rewards/accuracy_reward/std": 0.4323570549488068,
|
| 902 |
+
"rewards/cosine_scaled_reward/mean": 0.7470160365104676,
|
| 903 |
+
"rewards/cosine_scaled_reward/std": 0.4321545004844666,
|
| 904 |
+
"rewards/format_reward/mean": 0.9950892806053162,
|
| 905 |
+
"rewards/format_reward/std": 0.06607652679085732,
|
| 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": 0.0036299410540915344,
|
| 915 |
+
"completions/max_length": 2048.0,
|
| 916 |
+
"completions/max_terminated_length": 1968.8,
|
| 917 |
+
"completions/mean_length": 619.8013061523437,
|
| 918 |
+
"completions/mean_terminated_length": 614.59990234375,
|
| 919 |
+
"completions/min_length": 153.6,
|
| 920 |
+
"completions/min_terminated_length": 153.6,
|
| 921 |
+
"epoch": 2.025202520252025,
|
| 922 |
+
"grad_norm": 0.14987502992153168,
|
| 923 |
+
"learning_rate": 7.474410170262718e-07,
|
| 924 |
+
"loss": 0.0109,
|
| 925 |
+
"num_tokens": 118441067.0,
|
| 926 |
+
"reward": 1.608110237121582,
|
| 927 |
+
"reward_std": 0.2891895234584808,
|
| 928 |
+
"rewards/accuracy_reward/mean": 0.7404017925262452,
|
| 929 |
+
"rewards/accuracy_reward/std": 0.437438040971756,
|
| 930 |
+
"rewards/cosine_scaled_reward/mean": 0.7396578192710876,
|
| 931 |
+
"rewards/cosine_scaled_reward/std": 0.4371993124485016,
|
| 932 |
+
"rewards/format_reward/mean": 0.9957589268684387,
|
| 933 |
+
"rewards/format_reward/std": 0.0631766937673092,
|
| 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": 0.006026785714285721,
|
| 943 |
+
"completions/max_length": 2048.0,
|
| 944 |
+
"completions/max_terminated_length": 1911.4,
|
| 945 |
+
"completions/mean_length": 626.1058227539063,
|
| 946 |
+
"completions/mean_terminated_length": 617.5382690429688,
|
| 947 |
+
"completions/min_length": 183.0,
|
| 948 |
+
"completions/min_terminated_length": 183.0,
|
| 949 |
+
"epoch": 2.0882088208820884,
|
| 950 |
+
"grad_norm": 0.17082159221172333,
|
| 951 |
+
"learning_rate": 7.279803201353449e-07,
|
| 952 |
+
"loss": 0.0123,
|
| 953 |
+
"num_tokens": 122031925.0,
|
| 954 |
+
"reward": 1.6282689809799193,
|
| 955 |
+
"reward_std": 0.2621865808963776,
|
| 956 |
+
"rewards/accuracy_reward/mean": 0.754687511920929,
|
| 957 |
+
"rewards/accuracy_reward/std": 0.42926387786865233,
|
| 958 |
+
"rewards/cosine_scaled_reward/mean": 0.7534128069877625,
|
| 959 |
+
"rewards/cosine_scaled_reward/std": 0.4291880249977112,
|
| 960 |
+
"rewards/format_reward/mean": 0.99375,
|
| 961 |
+
"rewards/format_reward/std": 0.07523396834731103,
|
| 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": 0.002678571428571419,
|
| 971 |
+
"completions/max_length": 2017.2,
|
| 972 |
+
"completions/max_terminated_length": 1854.8,
|
| 973 |
+
"completions/mean_length": 626.8839477539062,
|
| 974 |
+
"completions/mean_terminated_length": 623.02763671875,
|
| 975 |
+
"completions/min_length": 191.6,
|
| 976 |
+
"completions/min_terminated_length": 191.6,
|
| 977 |
+
"epoch": 2.151215121512151,
|
| 978 |
+
"grad_norm": 0.14197254180908203,
|
| 979 |
+
"learning_rate": 7.080733410617851e-07,
|
| 980 |
+
"loss": 0.0094,
|
| 981 |
+
"num_tokens": 125637397.0,
|
| 982 |
+
"reward": 1.6435120105743408,
|
| 983 |
+
"reward_std": 0.2684548318386078,
|
| 984 |
+
"rewards/accuracy_reward/mean": 0.7638392806053161,
|
| 985 |
+
"rewards/accuracy_reward/std": 0.42319408655166624,
|
| 986 |
+
"rewards/cosine_scaled_reward/mean": 0.7626935601234436,
|
| 987 |
+
"rewards/cosine_scaled_reward/std": 0.4231561243534088,
|
| 988 |
+
"rewards/format_reward/mean": 0.9966517925262451,
|
| 989 |
+
"rewards/format_reward/std": 0.04984662607312203,
|
| 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": 0.0033482142857143016,
|
| 999 |
+
"completions/max_length": 2048.0,
|
| 1000 |
+
"completions/max_terminated_length": 1970.8,
|
| 1001 |
+
"completions/mean_length": 644.8076293945312,
|
| 1002 |
+
"completions/mean_terminated_length": 640.0797729492188,
|
| 1003 |
+
"completions/min_length": 151.6,
|
| 1004 |
+
"completions/min_terminated_length": 151.6,
|
| 1005 |
+
"epoch": 2.2142214221422143,
|
| 1006 |
+
"grad_norm": 0.16446265578269958,
|
| 1007 |
+
"learning_rate": 6.87759048651962e-07,
|
| 1008 |
+
"loss": 0.0102,
|
| 1009 |
+
"num_tokens": 129311063.0,
|
| 1010 |
+
"reward": 1.6173259973526002,
|
| 1011 |
+
"reward_std": 0.29779149889945983,
|
| 1012 |
+
"rewards/accuracy_reward/mean": 0.7466517806053161,
|
| 1013 |
+
"rewards/accuracy_reward/std": 0.43190850019454957,
|
| 1014 |
+
"rewards/cosine_scaled_reward/mean": 0.7460358500480652,
|
| 1015 |
+
"rewards/cosine_scaled_reward/std": 0.43167824745178224,
|
| 1016 |
+
"rewards/format_reward/mean": 0.9953125,
|
| 1017 |
+
"rewards/format_reward/std": 0.06686954125761986,
|
| 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": 2048.0,
|
| 1028 |
+
"completions/max_terminated_length": 1889.6,
|
| 1029 |
+
"completions/mean_length": 629.3937744140625,
|
| 1030 |
+
"completions/mean_terminated_length": 621.8263305664062,
|
| 1031 |
+
"completions/min_length": 170.8,
|
| 1032 |
+
"completions/min_terminated_length": 170.8,
|
| 1033 |
+
"epoch": 2.2772277227722775,
|
| 1034 |
+
"grad_norm": 0.13965024054050446,
|
| 1035 |
+
"learning_rate": 6.670772090872278e-07,
|
| 1036 |
+
"loss": 0.0114,
|
| 1037 |
+
"num_tokens": 132892795.0,
|
| 1038 |
+
"reward": 1.6336923837661743,
|
| 1039 |
+
"reward_std": 0.26696730852127076,
|
| 1040 |
+
"rewards/accuracy_reward/mean": 0.7592241644859314,
|
| 1041 |
+
"rewards/accuracy_reward/std": 0.424642938375473,
|
| 1042 |
+
"rewards/cosine_scaled_reward/mean": 0.7585270166397095,
|
| 1043 |
+
"rewards/cosine_scaled_reward/std": 0.4244441330432892,
|
| 1044 |
+
"rewards/format_reward/mean": 0.9946428418159485,
|
| 1045 |
+
"rewards/format_reward/std": 0.07060872837901115,
|
| 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": 0.00401785714285714,
|
| 1055 |
+
"completions/max_length": 1994.0,
|
| 1056 |
+
"completions/max_terminated_length": 1871.0,
|
| 1057 |
+
"completions/mean_length": 623.195556640625,
|
| 1058 |
+
"completions/mean_terminated_length": 617.4927368164062,
|
| 1059 |
+
"completions/min_length": 170.2,
|
| 1060 |
+
"completions/min_terminated_length": 170.2,
|
| 1061 |
+
"epoch": 2.34023402340234,
|
| 1062 |
+
"grad_norm": 0.13541771471500397,
|
| 1063 |
+
"learning_rate": 6.460683080397508e-07,
|
| 1064 |
+
"loss": 0.0108,
|
| 1065 |
+
"num_tokens": 136497919.0,
|
| 1066 |
+
"reward": 1.638669991493225,
|
| 1067 |
+
"reward_std": 0.2737217009067535,
|
| 1068 |
+
"rewards/accuracy_reward/mean": 0.7620797157287598,
|
| 1069 |
+
"rewards/accuracy_reward/std": 0.4253371238708496,
|
| 1070 |
+
"rewards/cosine_scaled_reward/mean": 0.7601523399353027,
|
| 1071 |
+
"rewards/cosine_scaled_reward/std": 0.42597259283065797,
|
| 1072 |
+
"rewards/format_reward/mean": 0.9957589387893677,
|
| 1073 |
+
"rewards/format_reward/std": 0.05754158571362496,
|
| 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": 0.004687500000000022,
|
| 1083 |
+
"completions/max_length": 2048.0,
|
| 1084 |
+
"completions/max_terminated_length": 1951.2,
|
| 1085 |
+
"completions/mean_length": 617.7080688476562,
|
| 1086 |
+
"completions/mean_terminated_length": 610.93955078125,
|
| 1087 |
+
"completions/min_length": 158.4,
|
| 1088 |
+
"completions/min_terminated_length": 158.4,
|
| 1089 |
+
"epoch": 2.4032403240324034,
|
| 1090 |
+
"grad_norm": 0.1562889963388443,
|
| 1091 |
+
"learning_rate": 6.247734714199105e-07,
|
| 1092 |
+
"loss": 0.018,
|
| 1093 |
+
"num_tokens": 140050915.0,
|
| 1094 |
+
"reward": 1.6390478134155273,
|
| 1095 |
+
"reward_std": 0.2830404281616211,
|
| 1096 |
+
"rewards/accuracy_reward/mean": 0.7611607193946839,
|
| 1097 |
+
"rewards/accuracy_reward/std": 0.4261560559272766,
|
| 1098 |
+
"rewards/cosine_scaled_reward/mean": 0.7606848120689392,
|
| 1099 |
+
"rewards/cosine_scaled_reward/std": 0.42603325843811035,
|
| 1100 |
+
"rewards/format_reward/mean": 0.995089304447174,
|
| 1101 |
+
"rewards/format_reward/std": 0.06784935072064399,
|
| 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": 0.0049107142857142795,
|
| 1111 |
+
"completions/max_length": 2048.0,
|
| 1112 |
+
"completions/max_terminated_length": 1952.4,
|
| 1113 |
+
"completions/mean_length": 629.9161010742188,
|
| 1114 |
+
"completions/mean_terminated_length": 622.9476318359375,
|
| 1115 |
+
"completions/min_length": 170.0,
|
| 1116 |
+
"completions/min_terminated_length": 170.0,
|
| 1117 |
+
"epoch": 2.466246624662466,
|
| 1118 |
+
"grad_norm": 0.16037462651729584,
|
| 1119 |
+
"learning_rate": 6.032343848703969e-07,
|
| 1120 |
+
"loss": 0.0111,
|
| 1121 |
+
"num_tokens": 143659179.0,
|
| 1122 |
+
"reward": 1.6348266124725341,
|
| 1123 |
+
"reward_std": 0.2890359699726105,
|
| 1124 |
+
"rewards/accuracy_reward/mean": 0.7587053537368774,
|
| 1125 |
+
"rewards/accuracy_reward/std": 0.4270632088184357,
|
| 1126 |
+
"rewards/cosine_scaled_reward/mean": 0.7578226804733277,
|
| 1127 |
+
"rewards/cosine_scaled_reward/std": 0.4268454909324646,
|
| 1128 |
+
"rewards/format_reward/mean": 0.9944196343421936,
|
| 1129 |
+
"rewards/format_reward/std": 0.0683063194155693,
|
| 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": 0.0046875,
|
| 1139 |
+
"completions/max_length": 2011.6,
|
| 1140 |
+
"completions/max_terminated_length": 1926.0,
|
| 1141 |
+
"completions/mean_length": 632.8710083007812,
|
| 1142 |
+
"completions/mean_terminated_length": 626.33193359375,
|
| 1143 |
+
"completions/min_length": 161.6,
|
| 1144 |
+
"completions/min_terminated_length": 161.6,
|
| 1145 |
+
"epoch": 2.5292529252925293,
|
| 1146 |
+
"grad_norm": 0.14154884219169617,
|
| 1147 |
+
"learning_rate": 5.814932121646073e-07,
|
| 1148 |
+
"loss": 0.0079,
|
| 1149 |
+
"num_tokens": 147280177.0,
|
| 1150 |
+
"reward": 1.6299890995025634,
|
| 1151 |
+
"reward_std": 0.27802737057209015,
|
| 1152 |
+
"rewards/accuracy_reward/mean": 0.7558035850524902,
|
| 1153 |
+
"rewards/accuracy_reward/std": 0.42795242071151735,
|
| 1154 |
+
"rewards/cosine_scaled_reward/mean": 0.7550673365592957,
|
| 1155 |
+
"rewards/cosine_scaled_reward/std": 0.4278149545192719,
|
| 1156 |
+
"rewards/format_reward/mean": 0.9933035731315613,
|
| 1157 |
+
"rewards/format_reward/std": 0.06978957653045655,
|
| 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": 0.0051339285714286035,
|
| 1167 |
+
"completions/max_length": 2048.0,
|
| 1168 |
+
"completions/max_terminated_length": 1918.8,
|
| 1169 |
+
"completions/mean_length": 615.8734741210938,
|
| 1170 |
+
"completions/mean_terminated_length": 608.469384765625,
|
| 1171 |
+
"completions/min_length": 168.8,
|
| 1172 |
+
"completions/min_terminated_length": 168.8,
|
| 1173 |
+
"epoch": 2.592259225922592,
|
| 1174 |
+
"grad_norm": 0.12673905491828918,
|
| 1175 |
+
"learning_rate": 5.595925126690801e-07,
|
| 1176 |
+
"loss": 0.0128,
|
| 1177 |
+
"num_tokens": 150787250.0,
|
| 1178 |
+
"reward": 1.6726323127746583,
|
| 1179 |
+
"reward_std": 0.2540514647960663,
|
| 1180 |
+
"rewards/accuracy_reward/mean": 0.7837053537368774,
|
| 1181 |
+
"rewards/accuracy_reward/std": 0.40963547229766845,
|
| 1182 |
+
"rewards/cosine_scaled_reward/mean": 0.782987654209137,
|
| 1183 |
+
"rewards/cosine_scaled_reward/std": 0.40952636003494264,
|
| 1184 |
+
"rewards/format_reward/mean": 0.9948660850524902,
|
| 1185 |
+
"rewards/format_reward/std": 0.06850367933511733,
|
| 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": 0.003794642857142838,
|
| 1195 |
+
"completions/max_length": 2048.0,
|
| 1196 |
+
"completions/max_terminated_length": 1862.4,
|
| 1197 |
+
"completions/mean_length": 612.8245849609375,
|
| 1198 |
+
"completions/mean_terminated_length": 607.3509399414063,
|
| 1199 |
+
"completions/min_length": 166.8,
|
| 1200 |
+
"completions/min_terminated_length": 166.8,
|
| 1201 |
+
"epoch": 2.6552655265526552,
|
| 1202 |
+
"grad_norm": 0.12310470640659332,
|
| 1203 |
+
"learning_rate": 5.375751580315354e-07,
|
| 1204 |
+
"loss": 0.0103,
|
| 1205 |
+
"num_tokens": 154286136.0,
|
| 1206 |
+
"reward": 1.634846067428589,
|
| 1207 |
+
"reward_std": 0.25745274126529694,
|
| 1208 |
+
"rewards/accuracy_reward/mean": 0.7580357193946838,
|
| 1209 |
+
"rewards/accuracy_reward/std": 0.42683074474334715,
|
| 1210 |
+
"rewards/cosine_scaled_reward/mean": 0.7574151754379272,
|
| 1211 |
+
"rewards/cosine_scaled_reward/std": 0.42679900527000425,
|
| 1212 |
+
"rewards/format_reward/mean": 0.9962053537368775,
|
| 1213 |
+
"rewards/format_reward/std": 0.0598295733332634,
|
| 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": 0.0035714285714285587,
|
| 1223 |
+
"completions/max_length": 2048.0,
|
| 1224 |
+
"completions/max_terminated_length": 1913.4,
|
| 1225 |
+
"completions/mean_length": 600.7576049804687,
|
| 1226 |
+
"completions/mean_terminated_length": 595.5764038085938,
|
| 1227 |
+
"completions/min_length": 178.4,
|
| 1228 |
+
"completions/min_terminated_length": 178.4,
|
| 1229 |
+
"epoch": 2.7182718271827184,
|
| 1230 |
+
"grad_norm": 0.12419384717941284,
|
| 1231 |
+
"learning_rate": 5.154842482576126e-07,
|
| 1232 |
+
"loss": 0.0126,
|
| 1233 |
+
"num_tokens": 157779418.0,
|
| 1234 |
+
"reward": 1.658141851425171,
|
| 1235 |
+
"reward_std": 0.259933602809906,
|
| 1236 |
+
"rewards/accuracy_reward/mean": 0.7734375,
|
| 1237 |
+
"rewards/accuracy_reward/std": 0.41526424288749697,
|
| 1238 |
+
"rewards/cosine_scaled_reward/mean": 0.7729800581932068,
|
| 1239 |
+
"rewards/cosine_scaled_reward/std": 0.4152462720870972,
|
| 1240 |
+
"rewards/format_reward/mean": 0.9964285612106323,
|
| 1241 |
+
"rewards/format_reward/std": 0.05749904364347458,
|
| 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": 1845.8,
|
| 1253 |
+
"completions/mean_length": 606.7765747070313,
|
| 1254 |
+
"completions/mean_terminated_length": 600.97294921875,
|
| 1255 |
+
"completions/min_length": 168.0,
|
| 1256 |
+
"completions/min_terminated_length": 168.0,
|
| 1257 |
+
"epoch": 2.781278127812781,
|
| 1258 |
+
"grad_norm": 0.16448700428009033,
|
| 1259 |
+
"learning_rate": 4.933630273405846e-07,
|
| 1260 |
+
"loss": 0.0162,
|
| 1261 |
+
"num_tokens": 161255897.0,
|
| 1262 |
+
"reward": 1.6536934852600098,
|
| 1263 |
+
"reward_std": 0.2780122488737106,
|
| 1264 |
+
"rewards/accuracy_reward/mean": 0.7709821581840515,
|
| 1265 |
+
"rewards/accuracy_reward/std": 0.41886167526245116,
|
| 1266 |
+
"rewards/cosine_scaled_reward/mean": 0.7701100707054138,
|
| 1267 |
+
"rewards/cosine_scaled_reward/std": 0.4186771154403687,
|
| 1268 |
+
"rewards/format_reward/mean": 0.9953125,
|
| 1269 |
+
"rewards/format_reward/std": 0.06595941931009293,
|
| 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": 0.002901785714285743,
|
| 1279 |
+
"completions/max_length": 2048.0,
|
| 1280 |
+
"completions/max_terminated_length": 2013.6,
|
| 1281 |
+
"completions/mean_length": 597.0652099609375,
|
| 1282 |
+
"completions/mean_terminated_length": 592.8518188476562,
|
| 1283 |
+
"completions/min_length": 162.6,
|
| 1284 |
+
"completions/min_terminated_length": 162.6,
|
| 1285 |
+
"epoch": 2.8442844284428443,
|
| 1286 |
+
"grad_norm": 0.14639931917190552,
|
| 1287 |
+
"learning_rate": 4.712547986092099e-07,
|
| 1288 |
+
"loss": 0.0067,
|
| 1289 |
+
"num_tokens": 164732445.0,
|
| 1290 |
+
"reward": 1.6427595376968385,
|
| 1291 |
+
"reward_std": 0.27772180140018465,
|
| 1292 |
+
"rewards/accuracy_reward/mean": 0.7633928656578064,
|
| 1293 |
+
"rewards/accuracy_reward/std": 0.42306692004203794,
|
| 1294 |
+
"rewards/cosine_scaled_reward/mean": 0.7627511620521545,
|
| 1295 |
+
"rewards/cosine_scaled_reward/std": 0.4228369653224945,
|
| 1296 |
+
"rewards/format_reward/mean": 0.9959821462631225,
|
| 1297 |
+
"rewards/format_reward/std": 0.062224668264389035,
|
| 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": 0.0008928571428571397,
|
| 1307 |
+
"completions/max_length": 1944.6,
|
| 1308 |
+
"completions/max_terminated_length": 1889.6,
|
| 1309 |
+
"completions/mean_length": 599.8812866210938,
|
| 1310 |
+
"completions/mean_terminated_length": 598.5843017578125,
|
| 1311 |
+
"completions/min_length": 169.6,
|
| 1312 |
+
"completions/min_terminated_length": 169.6,
|
| 1313 |
+
"epoch": 2.9072907290729075,
|
| 1314 |
+
"grad_norm": 0.1395055204629898,
|
| 1315 |
+
"learning_rate": 4.492028399594338e-07,
|
| 1316 |
+
"loss": 0.0043,
|
| 1317 |
+
"num_tokens": 168207073.0,
|
| 1318 |
+
"reward": 1.6468029737472534,
|
| 1319 |
+
"reward_std": 0.26623707115650175,
|
| 1320 |
+
"rewards/accuracy_reward/mean": 0.7649553537368774,
|
| 1321 |
+
"rewards/accuracy_reward/std": 0.42270631790161134,
|
| 1322 |
+
"rewards/cosine_scaled_reward/mean": 0.764587914943695,
|
| 1323 |
+
"rewards/cosine_scaled_reward/std": 0.42256171107292173,
|
| 1324 |
+
"rewards/format_reward/mean": 0.9991071343421936,
|
| 1325 |
+
"rewards/format_reward/std": 0.022806894034147263,
|
| 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": 0.006026785714285721,
|
| 1335 |
+
"completions/max_length": 2048.0,
|
| 1336 |
+
"completions/max_terminated_length": 1864.6,
|
| 1337 |
+
"completions/mean_length": 641.97099609375,
|
| 1338 |
+
"completions/mean_terminated_length": 633.473095703125,
|
| 1339 |
+
"completions/min_length": 186.6,
|
| 1340 |
+
"completions/min_terminated_length": 186.6,
|
| 1341 |
+
"epoch": 2.9702970297029703,
|
| 1342 |
+
"grad_norm": 0.12765838205814362,
|
| 1343 |
+
"learning_rate": 4.2725031913587426e-07,
|
| 1344 |
+
"loss": 0.0096,
|
| 1345 |
+
"num_tokens": 171879167.0,
|
| 1346 |
+
"reward": 1.6136202335357666,
|
| 1347 |
+
"reward_std": 0.26737120449543,
|
| 1348 |
+
"rewards/accuracy_reward/mean": 0.7448660731315613,
|
| 1349 |
+
"rewards/accuracy_reward/std": 0.43231003880500796,
|
| 1350 |
+
"rewards/cosine_scaled_reward/mean": 0.7442046284675599,
|
| 1351 |
+
"rewards/cosine_scaled_reward/std": 0.43226224184036255,
|
| 1352 |
+
"rewards/format_reward/mean": 0.9933035612106323,
|
| 1353 |
+
"rewards/format_reward/std": 0.07773758172988891,
|
| 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": 0.005047243411927904,
|
| 1363 |
+
"completions/max_length": 2048.0,
|
| 1364 |
+
"completions/max_terminated_length": 1930.4,
|
| 1365 |
+
"completions/mean_length": 646.4539672851563,
|
| 1366 |
+
"completions/mean_terminated_length": 639.4364501953125,
|
| 1367 |
+
"completions/min_length": 183.2,
|
| 1368 |
+
"completions/min_terminated_length": 183.2,
|
| 1369 |
+
"epoch": 3.037803780378038,
|
| 1370 |
+
"grad_norm": 0.1862199455499649,
|
| 1371 |
+
"learning_rate": 4.054402092289354e-07,
|
| 1372 |
+
"loss": 0.0103,
|
| 1373 |
+
"num_tokens": 175632429.0,
|
| 1374 |
+
"reward": 1.5804502010345458,
|
| 1375 |
+
"reward_std": 0.29361810684204104,
|
| 1376 |
+
"rewards/accuracy_reward/mean": 0.7220982074737549,
|
| 1377 |
+
"rewards/accuracy_reward/std": 0.44716750979423525,
|
| 1378 |
+
"rewards/cosine_scaled_reward/mean": 0.7216144323348999,
|
| 1379 |
+
"rewards/cosine_scaled_reward/std": 0.44697086215019227,
|
| 1380 |
+
"rewards/format_reward/mean": 0.9950892925262451,
|
| 1381 |
+
"rewards/format_reward/std": 0.06746494099497795,
|
| 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": 0.0033482142857142795,
|
| 1391 |
+
"completions/max_length": 2020.6,
|
| 1392 |
+
"completions/max_terminated_length": 1812.2,
|
| 1393 |
+
"completions/mean_length": 615.8056030273438,
|
| 1394 |
+
"completions/mean_terminated_length": 611.0432861328125,
|
| 1395 |
+
"completions/min_length": 194.2,
|
| 1396 |
+
"completions/min_terminated_length": 194.2,
|
| 1397 |
+
"epoch": 3.1008100810081007,
|
| 1398 |
+
"grad_norm": 0.16215628385543823,
|
| 1399 |
+
"learning_rate": 3.838152045529636e-07,
|
| 1400 |
+
"loss": 0.0143,
|
| 1401 |
+
"num_tokens": 179124806.0,
|
| 1402 |
+
"reward": 1.6801328897476195,
|
| 1403 |
+
"reward_std": 0.2461833655834198,
|
| 1404 |
+
"rewards/accuracy_reward/mean": 0.7881696343421936,
|
| 1405 |
+
"rewards/accuracy_reward/std": 0.4080040454864502,
|
| 1406 |
+
"rewards/cosine_scaled_reward/mean": 0.7877209305763244,
|
| 1407 |
+
"rewards/cosine_scaled_reward/std": 0.4078561782836914,
|
| 1408 |
+
"rewards/format_reward/mean": 0.9962053537368775,
|
| 1409 |
+
"rewards/format_reward/std": 0.05407858937978745,
|
| 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": 0.002678571428571419,
|
| 1419 |
+
"completions/max_length": 2029.4,
|
| 1420 |
+
"completions/max_terminated_length": 1823.4,
|
| 1421 |
+
"completions/mean_length": 606.2763671875,
|
| 1422 |
+
"completions/mean_terminated_length": 602.3788940429688,
|
| 1423 |
+
"completions/min_length": 189.6,
|
| 1424 |
+
"completions/min_terminated_length": 189.6,
|
| 1425 |
+
"epoch": 3.163816381638164,
|
| 1426 |
+
"grad_norm": 0.14493921399116516,
|
| 1427 |
+
"learning_rate": 3.624176370701235e-07,
|
| 1428 |
+
"loss": 0.0047,
|
| 1429 |
+
"num_tokens": 182588044.0,
|
| 1430 |
+
"reward": 1.6811640739440918,
|
| 1431 |
+
"reward_std": 0.24738809168338777,
|
| 1432 |
+
"rewards/accuracy_reward/mean": 0.7888392806053162,
|
| 1433 |
+
"rewards/accuracy_reward/std": 0.4081935465335846,
|
| 1434 |
+
"rewards/cosine_scaled_reward/mean": 0.7882209300994873,
|
| 1435 |
+
"rewards/cosine_scaled_reward/std": 0.40814857482910155,
|
| 1436 |
+
"rewards/format_reward/mean": 0.9964285731315613,
|
| 1437 |
+
"rewards/format_reward/std": 0.053148042410612106,
|
| 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": 0.0022321428571428605,
|
| 1447 |
+
"completions/max_length": 2048.0,
|
| 1448 |
+
"completions/max_terminated_length": 1797.2,
|
| 1449 |
+
"completions/mean_length": 610.3625122070313,
|
| 1450 |
+
"completions/mean_terminated_length": 607.1727905273438,
|
| 1451 |
+
"completions/min_length": 166.6,
|
| 1452 |
+
"completions/min_terminated_length": 166.6,
|
| 1453 |
+
"epoch": 3.2268226822682267,
|
| 1454 |
+
"grad_norm": 0.12905967235565186,
|
| 1455 |
+
"learning_rate": 3.412893935235936e-07,
|
| 1456 |
+
"loss": 0.0075,
|
| 1457 |
+
"num_tokens": 186106108.0,
|
| 1458 |
+
"reward": 1.664694619178772,
|
| 1459 |
+
"reward_std": 0.2503175616264343,
|
| 1460 |
+
"rewards/accuracy_reward/mean": 0.7774553537368775,
|
| 1461 |
+
"rewards/accuracy_reward/std": 0.41434150338172915,
|
| 1462 |
+
"rewards/cosine_scaled_reward/mean": 0.7767105340957642,
|
| 1463 |
+
"rewards/cosine_scaled_reward/std": 0.4142517387866974,
|
| 1464 |
+
"rewards/format_reward/mean": 0.9977678418159485,
|
| 1465 |
+
"rewards/format_reward/std": 0.04592652022838593,
|
| 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": 0.003571428571428581,
|
| 1475 |
+
"completions/max_length": 2048.0,
|
| 1476 |
+
"completions/max_terminated_length": 1915.6,
|
| 1477 |
+
"completions/mean_length": 619.6020385742188,
|
| 1478 |
+
"completions/mean_terminated_length": 614.4874267578125,
|
| 1479 |
+
"completions/min_length": 146.0,
|
| 1480 |
+
"completions/min_terminated_length": 146.0,
|
| 1481 |
+
"epoch": 3.28982898289829,
|
| 1482 |
+
"grad_norm": 0.1830361932516098,
|
| 1483 |
+
"learning_rate": 3.204718334422998e-07,
|
| 1484 |
+
"loss": 0.0095,
|
| 1485 |
+
"num_tokens": 189646853.0,
|
| 1486 |
+
"reward": 1.6342453956604004,
|
| 1487 |
+
"reward_std": 0.30342464447021483,
|
| 1488 |
+
"rewards/accuracy_reward/mean": 0.7575892806053162,
|
| 1489 |
+
"rewards/accuracy_reward/std": 0.42738271951675416,
|
| 1490 |
+
"rewards/cosine_scaled_reward/mean": 0.7568834662437439,
|
| 1491 |
+
"rewards/cosine_scaled_reward/std": 0.4271435678005219,
|
| 1492 |
+
"rewards/format_reward/mean": 0.9964285731315613,
|
| 1493 |
+
"rewards/format_reward/std": 0.059244754165410994,
|
| 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": 0.0020089285714285808,
|
| 1503 |
+
"completions/max_length": 1969.0,
|
| 1504 |
+
"completions/max_terminated_length": 1903.2,
|
| 1505 |
+
"completions/mean_length": 618.0531494140625,
|
| 1506 |
+
"completions/mean_terminated_length": 615.2063110351562,
|
| 1507 |
+
"completions/min_length": 160.0,
|
| 1508 |
+
"completions/min_terminated_length": 160.0,
|
| 1509 |
+
"epoch": 3.352835283528353,
|
| 1510 |
+
"grad_norm": 0.13739930093288422,
|
| 1511 |
+
"learning_rate": 3.0000570817769464e-07,
|
| 1512 |
+
"loss": 0.0078,
|
| 1513 |
+
"num_tokens": 193209859.0,
|
| 1514 |
+
"reward": 1.6563403844833373,
|
| 1515 |
+
"reward_std": 0.2727595895528793,
|
| 1516 |
+
"rewards/accuracy_reward/mean": 0.7716517925262452,
|
| 1517 |
+
"rewards/accuracy_reward/std": 0.4175321161746979,
|
| 1518 |
+
"rewards/cosine_scaled_reward/mean": 0.7711627960205079,
|
| 1519 |
+
"rewards/cosine_scaled_reward/std": 0.4173714339733124,
|
| 1520 |
+
"rewards/format_reward/mean": 0.9982142806053161,
|
| 1521 |
+
"rewards/format_reward/std": 0.031032296270132063,
|
| 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": 0.002678571428571441,
|
| 1531 |
+
"completions/max_length": 1993.2,
|
| 1532 |
+
"completions/max_terminated_length": 1915.8,
|
| 1533 |
+
"completions/mean_length": 627.6781494140625,
|
| 1534 |
+
"completions/mean_terminated_length": 623.89052734375,
|
| 1535 |
+
"completions/min_length": 156.4,
|
| 1536 |
+
"completions/min_terminated_length": 156.4,
|
| 1537 |
+
"epoch": 3.4158415841584158,
|
| 1538 |
+
"grad_norm": 0.14828138053417206,
|
| 1539 |
+
"learning_rate": 2.7993108113107443e-07,
|
| 1540 |
+
"loss": 0.0066,
|
| 1541 |
+
"num_tokens": 196821017.0,
|
| 1542 |
+
"reward": 1.6634482383728026,
|
| 1543 |
+
"reward_std": 0.23966350257396699,
|
| 1544 |
+
"rewards/accuracy_reward/mean": 0.7767857193946839,
|
| 1545 |
+
"rewards/accuracy_reward/std": 0.41428372263908386,
|
| 1546 |
+
"rewards/cosine_scaled_reward/mean": 0.7762267231941223,
|
| 1547 |
+
"rewards/cosine_scaled_reward/std": 0.41410806179046633,
|
| 1548 |
+
"rewards/format_reward/mean": 0.9970982074737549,
|
| 1549 |
+
"rewards/format_reward/std": 0.0474703922867775,
|
| 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": 1884.6,
|
| 1561 |
+
"completions/mean_length": 609.0343994140625,
|
| 1562 |
+
"completions/mean_terminated_length": 603.2148071289063,
|
| 1563 |
+
"completions/min_length": 169.0,
|
| 1564 |
+
"completions/min_terminated_length": 169.0,
|
| 1565 |
+
"epoch": 3.478847884788479,
|
| 1566 |
+
"grad_norm": 0.1737632006406784,
|
| 1567 |
+
"learning_rate": 2.602872493275881e-07,
|
| 1568 |
+
"loss": 0.0051,
|
| 1569 |
+
"num_tokens": 200339011.0,
|
| 1570 |
+
"reward": 1.6778597354888916,
|
| 1571 |
+
"reward_std": 0.25451319515705106,
|
| 1572 |
+
"rewards/accuracy_reward/mean": 0.7868303537368775,
|
| 1573 |
+
"rewards/accuracy_reward/std": 0.4092303454875946,
|
| 1574 |
+
"rewards/cosine_scaled_reward/mean": 0.7860764384269714,
|
| 1575 |
+
"rewards/cosine_scaled_reward/std": 0.4091233015060425,
|
| 1576 |
+
"rewards/format_reward/mean": 0.9959821343421936,
|
| 1577 |
+
"rewards/format_reward/std": 0.056319963186979294,
|
| 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": 0.0024553571428571398,
|
| 1587 |
+
"completions/max_length": 2048.0,
|
| 1588 |
+
"completions/max_terminated_length": 1807.8,
|
| 1589 |
+
"completions/mean_length": 627.6569458007813,
|
| 1590 |
+
"completions/mean_terminated_length": 624.14736328125,
|
| 1591 |
+
"completions/min_length": 171.6,
|
| 1592 |
+
"completions/min_terminated_length": 171.6,
|
| 1593 |
+
"epoch": 3.5418541854185417,
|
| 1594 |
+
"grad_norm": 0.14920973777770996,
|
| 1595 |
+
"learning_rate": 2.411126664904665e-07,
|
| 1596 |
+
"loss": 0.0122,
|
| 1597 |
+
"num_tokens": 203921794.0,
|
| 1598 |
+
"reward": 1.640373420715332,
|
| 1599 |
+
"reward_std": 0.27687210142612456,
|
| 1600 |
+
"rewards/accuracy_reward/mean": 0.7618303537368775,
|
| 1601 |
+
"rewards/accuracy_reward/std": 0.424312424659729,
|
| 1602 |
+
"rewards/cosine_scaled_reward/mean": 0.7613270163536072,
|
| 1603 |
+
"rewards/cosine_scaled_reward/std": 0.42429112195968627,
|
| 1604 |
+
"rewards/format_reward/mean": 0.9957589149475098,
|
| 1605 |
+
"rewards/format_reward/std": 0.06131577789783478,
|
| 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": 0.0033482142857142795,
|
| 1615 |
+
"completions/max_length": 2048.0,
|
| 1616 |
+
"completions/max_terminated_length": 1812.4,
|
| 1617 |
+
"completions/mean_length": 599.8631958007812,
|
| 1618 |
+
"completions/mean_terminated_length": 595.0171997070313,
|
| 1619 |
+
"completions/min_length": 178.0,
|
| 1620 |
+
"completions/min_terminated_length": 178.0,
|
| 1621 |
+
"epoch": 3.604860486048605,
|
| 1622 |
+
"grad_norm": 0.16235579550266266,
|
| 1623 |
+
"learning_rate": 2.2244486776605241e-07,
|
| 1624 |
+
"loss": 0.0082,
|
| 1625 |
+
"num_tokens": 207393885.0,
|
| 1626 |
+
"reward": 1.6748189449310302,
|
| 1627 |
+
"reward_std": 0.25130158960819243,
|
| 1628 |
+
"rewards/accuracy_reward/mean": 0.784374988079071,
|
| 1629 |
+
"rewards/accuracy_reward/std": 0.4097891986370087,
|
| 1630 |
+
"rewards/cosine_scaled_reward/mean": 0.7840126991271973,
|
| 1631 |
+
"rewards/cosine_scaled_reward/std": 0.40970866084098817,
|
| 1632 |
+
"rewards/format_reward/mean": 0.996875,
|
| 1633 |
+
"rewards/format_reward/std": 0.05381677523255348,
|
| 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": 0.0020089285714285808,
|
| 1643 |
+
"completions/max_length": 1973.4,
|
| 1644 |
+
"completions/max_terminated_length": 1706.2,
|
| 1645 |
+
"completions/mean_length": 593.27236328125,
|
| 1646 |
+
"completions/mean_terminated_length": 590.3545043945312,
|
| 1647 |
+
"completions/min_length": 170.4,
|
| 1648 |
+
"completions/min_terminated_length": 170.4,
|
| 1649 |
+
"epoch": 3.667866786678668,
|
| 1650 |
+
"grad_norm": 0.14174528419971466,
|
| 1651 |
+
"learning_rate": 2.0432039624699015e-07,
|
| 1652 |
+
"loss": 0.0072,
|
| 1653 |
+
"num_tokens": 210886881.0,
|
| 1654 |
+
"reward": 1.6687217950820923,
|
| 1655 |
+
"reward_std": 0.2413029581308365,
|
| 1656 |
+
"rewards/accuracy_reward/mean": 0.7799107074737549,
|
| 1657 |
+
"rewards/accuracy_reward/std": 0.41347917914390564,
|
| 1658 |
+
"rewards/cosine_scaled_reward/mean": 0.779630982875824,
|
| 1659 |
+
"rewards/cosine_scaled_reward/std": 0.4133819997310638,
|
| 1660 |
+
"rewards/format_reward/mean": 0.9979910731315613,
|
| 1661 |
+
"rewards/format_reward/std": 0.038909842073917386,
|
| 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": 0.004241071428571419,
|
| 1671 |
+
"completions/max_length": 2048.0,
|
| 1672 |
+
"completions/max_terminated_length": 1969.2,
|
| 1673 |
+
"completions/mean_length": 635.64912109375,
|
| 1674 |
+
"completions/mean_terminated_length": 629.61572265625,
|
| 1675 |
+
"completions/min_length": 174.6,
|
| 1676 |
+
"completions/min_terminated_length": 174.6,
|
| 1677 |
+
"epoch": 3.730873087308731,
|
| 1678 |
+
"grad_norm": 0.13457916676998138,
|
| 1679 |
+
"learning_rate": 1.867747314374041e-07,
|
| 1680 |
+
"loss": 0.014,
|
| 1681 |
+
"num_tokens": 214550253.0,
|
| 1682 |
+
"reward": 1.6441724061965943,
|
| 1683 |
+
"reward_std": 0.25683614015579226,
|
| 1684 |
+
"rewards/accuracy_reward/mean": 0.7647321343421936,
|
| 1685 |
+
"rewards/accuracy_reward/std": 0.42399376034736636,
|
| 1686 |
+
"rewards/cosine_scaled_reward/mean": 0.7633446335792542,
|
| 1687 |
+
"rewards/cosine_scaled_reward/std": 0.4237023591995239,
|
| 1688 |
+
"rewards/format_reward/mean": 0.9955357193946839,
|
| 1689 |
+
"rewards/format_reward/std": 0.06432178840041161,
|
| 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": 0.006026785714285698,
|
| 1699 |
+
"completions/max_length": 2048.0,
|
| 1700 |
+
"completions/max_terminated_length": 1899.4,
|
| 1701 |
+
"completions/mean_length": 625.6288208007812,
|
| 1702 |
+
"completions/mean_terminated_length": 617.0573120117188,
|
| 1703 |
+
"completions/min_length": 183.8,
|
| 1704 |
+
"completions/min_terminated_length": 183.8,
|
| 1705 |
+
"epoch": 3.793879387938794,
|
| 1706 |
+
"grad_norm": 0.16838262975215912,
|
| 1707 |
+
"learning_rate": 1.6984221980010454e-07,
|
| 1708 |
+
"loss": 0.0107,
|
| 1709 |
+
"num_tokens": 218094638.0,
|
| 1710 |
+
"reward": 1.6642211198806762,
|
| 1711 |
+
"reward_std": 0.2543728470802307,
|
| 1712 |
+
"rewards/accuracy_reward/mean": 0.7783482193946838,
|
| 1713 |
+
"rewards/accuracy_reward/std": 0.41493446230888364,
|
| 1714 |
+
"rewards/cosine_scaled_reward/mean": 0.7779957056045532,
|
| 1715 |
+
"rewards/cosine_scaled_reward/std": 0.41481839418411254,
|
| 1716 |
+
"rewards/format_reward/mean": 0.993749988079071,
|
| 1717 |
+
"rewards/format_reward/std": 0.07161349579691886,
|
| 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": 0.004464285714285721,
|
| 1727 |
+
"completions/max_length": 2048.0,
|
| 1728 |
+
"completions/max_terminated_length": 1934.0,
|
| 1729 |
+
"completions/mean_length": 627.9725708007812,
|
| 1730 |
+
"completions/mean_terminated_length": 621.595361328125,
|
| 1731 |
+
"completions/min_length": 186.0,
|
| 1732 |
+
"completions/min_terminated_length": 186.0,
|
| 1733 |
+
"epoch": 3.856885688568857,
|
| 1734 |
+
"grad_norm": 0.14313475787639618,
|
| 1735 |
+
"learning_rate": 1.5355600752177312e-07,
|
| 1736 |
+
"loss": 0.0125,
|
| 1737 |
+
"num_tokens": 221700083.0,
|
| 1738 |
+
"reward": 1.6535644292831422,
|
| 1739 |
+
"reward_std": 0.2506650030612946,
|
| 1740 |
+
"rewards/accuracy_reward/mean": 0.7707589268684387,
|
| 1741 |
+
"rewards/accuracy_reward/std": 0.4196486592292786,
|
| 1742 |
+
"rewards/cosine_scaled_reward/mean": 0.7700751662254334,
|
| 1743 |
+
"rewards/cosine_scaled_reward/std": 0.41945862770080566,
|
| 1744 |
+
"rewards/format_reward/mean": 0.9955357074737549,
|
| 1745 |
+
"rewards/format_reward/std": 0.06590693667531014,
|
| 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": 0.003125,
|
| 1755 |
+
"completions/max_length": 2014.6,
|
| 1756 |
+
"completions/max_terminated_length": 1926.0,
|
| 1757 |
+
"completions/mean_length": 613.800927734375,
|
| 1758 |
+
"completions/mean_terminated_length": 609.321435546875,
|
| 1759 |
+
"completions/min_length": 179.8,
|
| 1760 |
+
"completions/min_terminated_length": 179.8,
|
| 1761 |
+
"epoch": 3.91989198919892,
|
| 1762 |
+
"grad_norm": 0.13859793543815613,
|
| 1763 |
+
"learning_rate": 1.3794797562774596e-07,
|
| 1764 |
+
"loss": 0.0068,
|
| 1765 |
+
"num_tokens": 225211495.0,
|
| 1766 |
+
"reward": 1.6338534832000733,
|
| 1767 |
+
"reward_std": 0.2609659105539322,
|
| 1768 |
+
"rewards/accuracy_reward/mean": 0.7595358729362488,
|
| 1769 |
+
"rewards/accuracy_reward/std": 0.4260308504104614,
|
| 1770 |
+
"rewards/cosine_scaled_reward/mean": 0.7553341269493103,
|
| 1771 |
+
"rewards/cosine_scaled_reward/std": 0.4275141417980194,
|
| 1772 |
+
"rewards/format_reward/mean": 0.9970982074737549,
|
| 1773 |
+
"rewards/format_reward/std": 0.0464889258146286,
|
| 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": 0.00625,
|
| 1783 |
+
"completions/max_length": 2048.0,
|
| 1784 |
+
"completions/max_terminated_length": 1900.4,
|
| 1785 |
+
"completions/mean_length": 611.440869140625,
|
| 1786 |
+
"completions/mean_terminated_length": 602.3813110351563,
|
| 1787 |
+
"completions/min_length": 181.0,
|
| 1788 |
+
"completions/min_terminated_length": 181.0,
|
| 1789 |
+
"epoch": 3.9828982898289826,
|
| 1790 |
+
"grad_norm": 0.1344553679227829,
|
| 1791 |
+
"learning_rate": 1.230486775734097e-07,
|
| 1792 |
+
"loss": 0.013,
|
| 1793 |
+
"num_tokens": 228791646.0,
|
| 1794 |
+
"reward": 1.6422872066497802,
|
| 1795 |
+
"reward_std": 0.23485693633556365,
|
| 1796 |
+
"rewards/accuracy_reward/mean": 0.7667692184448243,
|
| 1797 |
+
"rewards/accuracy_reward/std": 0.42179493308067323,
|
| 1798 |
+
"rewards/cosine_scaled_reward/mean": 0.761570417881012,
|
| 1799 |
+
"rewards/cosine_scaled_reward/std": 0.4240230083465576,
|
| 1800 |
+
"rewards/format_reward/mean": 0.9935267806053162,
|
| 1801 |
+
"rewards/format_reward/std": 0.07590627297759056,
|
| 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": 0.0031640083217753024,
|
| 1811 |
+
"completions/max_length": 2048.0,
|
| 1812 |
+
"completions/max_terminated_length": 1952.0,
|
| 1813 |
+
"completions/mean_length": 617.7764038085937,
|
| 1814 |
+
"completions/mean_terminated_length": 613.2669799804687,
|
| 1815 |
+
"completions/min_length": 164.4,
|
| 1816 |
+
"completions/min_terminated_length": 164.4,
|
| 1817 |
+
"epoch": 4.05040504050405,
|
| 1818 |
+
"grad_norm": 0.15917585790157318,
|
| 1819 |
+
"learning_rate": 1.0888727943437748e-07,
|
| 1820 |
+
"loss": 0.0084,
|
| 1821 |
+
"num_tokens": 232312551.0,
|
| 1822 |
+
"reward": 1.6375758409500123,
|
| 1823 |
+
"reward_std": 0.2651157915592194,
|
| 1824 |
+
"rewards/accuracy_reward/mean": 0.7629625201225281,
|
| 1825 |
+
"rewards/accuracy_reward/std": 0.42385138273239137,
|
| 1826 |
+
"rewards/cosine_scaled_reward/mean": 0.7563670516014099,
|
| 1827 |
+
"rewards/cosine_scaled_reward/std": 0.4267165422439575,
|
| 1828 |
+
"rewards/format_reward/mean": 0.996875,
|
| 1829 |
+
"rewards/format_reward/std": 0.052468103170394895,
|
| 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": 0.0022321428571428826,
|
| 1839 |
+
"completions/max_length": 2048.0,
|
| 1840 |
+
"completions/max_terminated_length": 1919.4,
|
| 1841 |
+
"completions/mean_length": 608.4109619140625,
|
| 1842 |
+
"completions/mean_terminated_length": 605.1753173828125,
|
| 1843 |
+
"completions/min_length": 191.8,
|
| 1844 |
+
"completions/min_terminated_length": 191.8,
|
| 1845 |
+
"epoch": 4.113411341134113,
|
| 1846 |
+
"grad_norm": 0.1608886420726776,
|
| 1847 |
+
"learning_rate": 9.549150281252632e-08,
|
| 1848 |
+
"loss": 0.0043,
|
| 1849 |
+
"num_tokens": 235787696.0,
|
| 1850 |
+
"reward": 1.66022846698761,
|
| 1851 |
+
"reward_std": 0.2489331156015396,
|
| 1852 |
+
"rewards/accuracy_reward/mean": 0.7745535850524903,
|
| 1853 |
+
"rewards/accuracy_reward/std": 0.41753928661346434,
|
| 1854 |
+
"rewards/cosine_scaled_reward/mean": 0.7740282416343689,
|
| 1855 |
+
"rewards/cosine_scaled_reward/std": 0.4174140512943268,
|
| 1856 |
+
"rewards/format_reward/mean": 0.9973214268684387,
|
| 1857 |
+
"rewards/format_reward/std": 0.05046965554356575,
|
| 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": 0.003794642857142838,
|
| 1867 |
+
"completions/max_length": 2048.0,
|
| 1868 |
+
"completions/max_terminated_length": 1945.6,
|
| 1869 |
+
"completions/mean_length": 622.467431640625,
|
| 1870 |
+
"completions/mean_terminated_length": 617.0385375976563,
|
| 1871 |
+
"completions/min_length": 152.6,
|
| 1872 |
+
"completions/min_terminated_length": 152.6,
|
| 1873 |
+
"epoch": 4.176417641764177,
|
| 1874 |
+
"grad_norm": 0.15062691271305084,
|
| 1875 |
+
"learning_rate": 8.288757056966017e-08,
|
| 1876 |
+
"loss": 0.0112,
|
| 1877 |
+
"num_tokens": 239378142.0,
|
| 1878 |
+
"reward": 1.6499414920806885,
|
| 1879 |
+
"reward_std": 0.2437061160802841,
|
| 1880 |
+
"rewards/accuracy_reward/mean": 0.7680803537368774,
|
| 1881 |
+
"rewards/accuracy_reward/std": 0.4208991706371307,
|
| 1882 |
+
"rewards/cosine_scaled_reward/mean": 0.7675167083740234,
|
| 1883 |
+
"rewards/cosine_scaled_reward/std": 0.4206974029541016,
|
| 1884 |
+
"rewards/format_reward/mean": 0.9962053418159484,
|
| 1885 |
+
"rewards/format_reward/std": 0.05708769783377647,
|
| 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": 0.005133928571428558,
|
| 1895 |
+
"completions/max_length": 2048.0,
|
| 1896 |
+
"completions/max_terminated_length": 1929.2,
|
| 1897 |
+
"completions/mean_length": 633.4506958007812,
|
| 1898 |
+
"completions/mean_terminated_length": 626.1871459960937,
|
| 1899 |
+
"completions/min_length": 175.4,
|
| 1900 |
+
"completions/min_terminated_length": 175.4,
|
| 1901 |
+
"epoch": 4.2394239423942395,
|
| 1902 |
+
"grad_norm": 0.14156123995780945,
|
| 1903 |
+
"learning_rate": 7.110015549502546e-08,
|
| 1904 |
+
"loss": 0.0082,
|
| 1905 |
+
"num_tokens": 243021681.0,
|
| 1906 |
+
"reward": 1.6500544548034668,
|
| 1907 |
+
"reward_std": 0.26026159822940825,
|
| 1908 |
+
"rewards/accuracy_reward/mean": 0.768749988079071,
|
| 1909 |
+
"rewards/accuracy_reward/std": 0.42072436213493347,
|
| 1910 |
+
"rewards/cosine_scaled_reward/mean": 0.7679658651351928,
|
| 1911 |
+
"rewards/cosine_scaled_reward/std": 0.4205094575881958,
|
| 1912 |
+
"rewards/format_reward/mean": 0.9946428418159485,
|
| 1913 |
+
"rewards/format_reward/std": 0.06923034116625786,
|
| 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": 0.0024553571428571177,
|
| 1923 |
+
"completions/max_length": 1975.6,
|
| 1924 |
+
"completions/max_terminated_length": 1895.2,
|
| 1925 |
+
"completions/mean_length": 627.3937744140625,
|
| 1926 |
+
"completions/mean_terminated_length": 623.917529296875,
|
| 1927 |
+
"completions/min_length": 161.8,
|
| 1928 |
+
"completions/min_terminated_length": 161.8,
|
| 1929 |
+
"epoch": 4.302430243024302,
|
| 1930 |
+
"grad_norm": 0.13976019620895386,
|
| 1931 |
+
"learning_rate": 6.015233200716874e-08,
|
| 1932 |
+
"loss": 0.0064,
|
| 1933 |
+
"num_tokens": 246632541.0,
|
| 1934 |
+
"reward": 1.6928518533706665,
|
| 1935 |
+
"reward_std": 0.2341456115245819,
|
| 1936 |
+
"rewards/accuracy_reward/mean": 0.7964285731315612,
|
| 1937 |
+
"rewards/accuracy_reward/std": 0.40205827355384827,
|
| 1938 |
+
"rewards/cosine_scaled_reward/mean": 0.7953016996383667,
|
| 1939 |
+
"rewards/cosine_scaled_reward/std": 0.40191030502319336,
|
| 1940 |
+
"rewards/format_reward/mean": 0.9975446343421936,
|
| 1941 |
+
"rewards/format_reward/std": 0.04161064326763153,
|
| 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": 0.005133928571428581,
|
| 1951 |
+
"completions/max_length": 2048.0,
|
| 1952 |
+
"completions/max_terminated_length": 1957.8,
|
| 1953 |
+
"completions/mean_length": 618.9397705078125,
|
| 1954 |
+
"completions/mean_terminated_length": 611.7075561523437,
|
| 1955 |
+
"completions/min_length": 167.0,
|
| 1956 |
+
"completions/min_terminated_length": 167.0,
|
| 1957 |
+
"epoch": 4.365436543654366,
|
| 1958 |
+
"grad_norm": 0.15124908089637756,
|
| 1959 |
+
"learning_rate": 5.006553098467764e-08,
|
| 1960 |
+
"loss": 0.0103,
|
| 1961 |
+
"num_tokens": 250178143.0,
|
| 1962 |
+
"reward": 1.6755474090576172,
|
| 1963 |
+
"reward_std": 0.2548744171857834,
|
| 1964 |
+
"rewards/accuracy_reward/mean": 0.7857142925262451,
|
| 1965 |
+
"rewards/accuracy_reward/std": 0.4092557728290558,
|
| 1966 |
+
"rewards/cosine_scaled_reward/mean": 0.7848000407218934,
|
| 1967 |
+
"rewards/cosine_scaled_reward/std": 0.4092385172843933,
|
| 1968 |
+
"rewards/format_reward/mean": 0.9948660850524902,
|
| 1969 |
+
"rewards/format_reward/std": 0.06760494336485863,
|
| 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": 0.004464285714285721,
|
| 1979 |
+
"completions/max_length": 2048.0,
|
| 1980 |
+
"completions/max_terminated_length": 1865.8,
|
| 1981 |
+
"completions/mean_length": 626.4183349609375,
|
| 1982 |
+
"completions/mean_terminated_length": 620.0564453125,
|
| 1983 |
+
"completions/min_length": 182.0,
|
| 1984 |
+
"completions/min_terminated_length": 182.0,
|
| 1985 |
+
"epoch": 4.428442844284429,
|
| 1986 |
+
"grad_norm": 0.14921754598617554,
|
| 1987 |
+
"learning_rate": 4.085949781423048e-08,
|
| 1988 |
+
"loss": 0.007,
|
| 1989 |
+
"num_tokens": 253769705.0,
|
| 1990 |
+
"reward": 1.6789589166641234,
|
| 1991 |
+
"reward_std": 0.23221377432346343,
|
| 1992 |
+
"rewards/accuracy_reward/mean": 0.7877232074737549,
|
| 1993 |
+
"rewards/accuracy_reward/std": 0.40675231218338015,
|
| 1994 |
+
"rewards/cosine_scaled_reward/mean": 0.7873820304870606,
|
| 1995 |
+
"rewards/cosine_scaled_reward/std": 0.4066896378993988,
|
| 1996 |
+
"rewards/format_reward/mean": 0.9950892806053162,
|
| 1997 |
+
"rewards/format_reward/std": 0.06925405710935592,
|
| 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": 0.002678571428571419,
|
| 2007 |
+
"completions/max_length": 2048.0,
|
| 2008 |
+
"completions/max_terminated_length": 1934.2,
|
| 2009 |
+
"completions/mean_length": 618.8587280273438,
|
| 2010 |
+
"completions/mean_terminated_length": 615.0157958984375,
|
| 2011 |
+
"completions/min_length": 160.0,
|
| 2012 |
+
"completions/min_terminated_length": 160.0,
|
| 2013 |
+
"epoch": 4.491449144914491,
|
| 2014 |
+
"grad_norm": 0.16535188257694244,
|
| 2015 |
+
"learning_rate": 3.2552253738074074e-08,
|
| 2016 |
+
"loss": 0.008,
|
| 2017 |
+
"num_tokens": 257334160.0,
|
| 2018 |
+
"reward": 1.6443732500076294,
|
| 2019 |
+
"reward_std": 0.2480471521615982,
|
| 2020 |
+
"rewards/accuracy_reward/mean": 0.7647321343421936,
|
| 2021 |
+
"rewards/accuracy_reward/std": 0.423707377910614,
|
| 2022 |
+
"rewards/cosine_scaled_reward/mean": 0.7641927599906921,
|
| 2023 |
+
"rewards/cosine_scaled_reward/std": 0.4235422372817993,
|
| 2024 |
+
"rewards/format_reward/mean": 0.9950892925262451,
|
| 2025 |
+
"rewards/format_reward/std": 0.06574137881398201,
|
| 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": 0.0024553571428571398,
|
| 2035 |
+
"completions/max_length": 1972.6,
|
| 2036 |
+
"completions/max_terminated_length": 1843.2,
|
| 2037 |
+
"completions/mean_length": 613.258740234375,
|
| 2038 |
+
"completions/mean_terminated_length": 609.7564453125,
|
| 2039 |
+
"completions/min_length": 191.4,
|
| 2040 |
+
"completions/min_terminated_length": 191.4,
|
| 2041 |
+
"epoch": 4.554455445544555,
|
| 2042 |
+
"grad_norm": 0.14730912446975708,
|
| 2043 |
+
"learning_rate": 2.5160060576596365e-08,
|
| 2044 |
+
"loss": 0.0072,
|
| 2045 |
+
"num_tokens": 260865495.0,
|
| 2046 |
+
"reward": 1.6269155740737915,
|
| 2047 |
+
"reward_std": 0.2688676297664642,
|
| 2048 |
+
"rewards/accuracy_reward/mean": 0.7524553418159485,
|
| 2049 |
+
"rewards/accuracy_reward/std": 0.42926363945007323,
|
| 2050 |
+
"rewards/cosine_scaled_reward/mean": 0.7520452857017517,
|
| 2051 |
+
"rewards/cosine_scaled_reward/std": 0.4292009472846985,
|
| 2052 |
+
"rewards/format_reward/mean": 0.996874988079071,
|
| 2053 |
+
"rewards/format_reward/std": 0.053170456737279895,
|
| 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": 0.0029017857142857205,
|
| 2063 |
+
"completions/max_length": 2048.0,
|
| 2064 |
+
"completions/max_terminated_length": 1827.2,
|
| 2065 |
+
"completions/mean_length": 618.671240234375,
|
| 2066 |
+
"completions/mean_terminated_length": 614.5585571289063,
|
| 2067 |
+
"completions/min_length": 169.0,
|
| 2068 |
+
"completions/min_terminated_length": 169.0,
|
| 2069 |
+
"epoch": 4.617461746174618,
|
| 2070 |
+
"grad_norm": 0.1736821085214615,
|
| 2071 |
+
"learning_rate": 1.8697388895048515e-08,
|
| 2072 |
+
"loss": 0.01,
|
| 2073 |
+
"num_tokens": 264474382.0,
|
| 2074 |
+
"reward": 1.6391252517700194,
|
| 2075 |
+
"reward_std": 0.25250489115715025,
|
| 2076 |
+
"rewards/accuracy_reward/mean": 0.760937488079071,
|
| 2077 |
+
"rewards/accuracy_reward/std": 0.4259650409221649,
|
| 2078 |
+
"rewards/cosine_scaled_reward/mean": 0.7601699948310852,
|
| 2079 |
+
"rewards/cosine_scaled_reward/std": 0.42582063674926757,
|
| 2080 |
+
"rewards/format_reward/mean": 0.9962053656578064,
|
| 2081 |
+
"rewards/format_reward/std": 0.05821031853556633,
|
| 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": 0.0013392857142857206,
|
| 2091 |
+
"completions/max_length": 1962.0,
|
| 2092 |
+
"completions/max_terminated_length": 1836.8,
|
| 2093 |
+
"completions/mean_length": 595.1326293945312,
|
| 2094 |
+
"completions/mean_terminated_length": 593.2115600585937,
|
| 2095 |
+
"completions/min_length": 174.4,
|
| 2096 |
+
"completions/min_terminated_length": 174.4,
|
| 2097 |
+
"epoch": 4.68046804680468,
|
| 2098 |
+
"grad_norm": 0.16435028612613678,
|
| 2099 |
+
"learning_rate": 1.3176889676734692e-08,
|
| 2100 |
+
"loss": 0.0088,
|
| 2101 |
+
"num_tokens": 267903232.0,
|
| 2102 |
+
"reward": 1.6919838905334472,
|
| 2103 |
+
"reward_std": 0.2522643059492111,
|
| 2104 |
+
"rewards/accuracy_reward/mean": 0.7953125,
|
| 2105 |
+
"rewards/accuracy_reward/std": 0.40310133099555967,
|
| 2106 |
+
"rewards/cosine_scaled_reward/mean": 0.7946819305419922,
|
| 2107 |
+
"rewards/cosine_scaled_reward/std": 0.40297635197639464,
|
| 2108 |
+
"rewards/format_reward/mean": 0.9986607074737549,
|
| 2109 |
+
"rewards/format_reward/std": 0.027685175836086272,
|
| 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": 0.00379464285714286,
|
| 2119 |
+
"completions/max_length": 2048.0,
|
| 2120 |
+
"completions/max_terminated_length": 1986.6,
|
| 2121 |
+
"completions/mean_length": 633.7323852539063,
|
| 2122 |
+
"completions/mean_terminated_length": 628.3538696289063,
|
| 2123 |
+
"completions/min_length": 181.8,
|
| 2124 |
+
"completions/min_terminated_length": 181.8,
|
| 2125 |
+
"epoch": 4.743474347434743,
|
| 2126 |
+
"grad_norm": 0.16870158910751343,
|
| 2127 |
+
"learning_rate": 8.60936955811753e-09,
|
| 2128 |
+
"loss": 0.0053,
|
| 2129 |
+
"num_tokens": 271506625.0,
|
| 2130 |
+
"reward": 1.677209448814392,
|
| 2131 |
+
"reward_std": 0.24304699897766113,
|
| 2132 |
+
"rewards/accuracy_reward/mean": 0.7866071343421936,
|
| 2133 |
+
"rewards/accuracy_reward/std": 0.4092015206813812,
|
| 2134 |
+
"rewards/cosine_scaled_reward/mean": 0.7849991202354432,
|
| 2135 |
+
"rewards/cosine_scaled_reward/std": 0.409130734205246,
|
| 2136 |
+
"rewards/format_reward/mean": 0.9962053537368775,
|
| 2137 |
+
"rewards/format_reward/std": 0.058644138276576996,
|
| 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": 0.003125,
|
| 2147 |
+
"completions/max_length": 2033.8,
|
| 2148 |
+
"completions/max_terminated_length": 1926.0,
|
| 2149 |
+
"completions/mean_length": 596.5832885742187,
|
| 2150 |
+
"completions/mean_terminated_length": 592.036572265625,
|
| 2151 |
+
"completions/min_length": 161.2,
|
| 2152 |
+
"completions/min_terminated_length": 161.2,
|
| 2153 |
+
"epoch": 4.806480648064807,
|
| 2154 |
+
"grad_norm": 0.18181964755058289,
|
| 2155 |
+
"learning_rate": 5.003769674320368e-09,
|
| 2156 |
+
"loss": 0.011,
|
| 2157 |
+
"num_tokens": 274925870.0,
|
| 2158 |
+
"reward": 1.6703053951263427,
|
| 2159 |
+
"reward_std": 0.24658992886543274,
|
| 2160 |
+
"rewards/accuracy_reward/mean": 0.7816964268684388,
|
| 2161 |
+
"rewards/accuracy_reward/std": 0.41211633682250975,
|
| 2162 |
+
"rewards/cosine_scaled_reward/mean": 0.7807891845703125,
|
| 2163 |
+
"rewards/cosine_scaled_reward/std": 0.4120659828186035,
|
| 2164 |
+
"rewards/format_reward/mean": 0.9964285731315613,
|
| 2165 |
+
"rewards/format_reward/std": 0.05088246762752533,
|
| 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": 0.00401785714285714,
|
| 2175 |
+
"completions/max_length": 2025.6,
|
| 2176 |
+
"completions/max_terminated_length": 1959.0,
|
| 2177 |
+
"completions/mean_length": 629.3553833007812,
|
| 2178 |
+
"completions/mean_terminated_length": 623.6705200195313,
|
| 2179 |
+
"completions/min_length": 182.2,
|
| 2180 |
+
"completions/min_terminated_length": 182.2,
|
| 2181 |
+
"epoch": 4.8694869486948695,
|
| 2182 |
+
"grad_norm": 0.15454868972301483,
|
| 2183 |
+
"learning_rate": 2.3671481564347396e-09,
|
| 2184 |
+
"loss": 0.0112,
|
| 2185 |
+
"num_tokens": 278501238.0,
|
| 2186 |
+
"reward": 1.6728654146194457,
|
| 2187 |
+
"reward_std": 0.24723841547966002,
|
| 2188 |
+
"rewards/accuracy_reward/mean": 0.7837053537368774,
|
| 2189 |
+
"rewards/accuracy_reward/std": 0.41015875339508057,
|
| 2190 |
+
"rewards/cosine_scaled_reward/mean": 0.7823377847671509,
|
| 2191 |
+
"rewards/cosine_scaled_reward/std": 0.40985342860221863,
|
| 2192 |
+
"rewards/format_reward/mean": 0.9959821343421936,
|
| 2193 |
+
"rewards/format_reward/std": 0.04546051248908043,
|
| 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": 0.0033482142857142795,
|
| 2203 |
+
"completions/max_length": 2048.0,
|
| 2204 |
+
"completions/max_terminated_length": 1895.4,
|
| 2205 |
+
"completions/mean_length": 639.8073974609375,
|
| 2206 |
+
"completions/mean_terminated_length": 635.0597290039062,
|
| 2207 |
+
"completions/min_length": 183.0,
|
| 2208 |
+
"completions/min_terminated_length": 183.0,
|
| 2209 |
+
"epoch": 4.932493249324932,
|
| 2210 |
+
"grad_norm": 0.15312208235263824,
|
| 2211 |
+
"learning_rate": 7.046663148979615e-10,
|
| 2212 |
+
"loss": 0.0143,
|
| 2213 |
+
"num_tokens": 282142359.0,
|
| 2214 |
+
"reward": 1.6685362339019776,
|
| 2215 |
+
"reward_std": 0.25656993985176085,
|
| 2216 |
+
"rewards/accuracy_reward/mean": 0.7805803537368774,
|
| 2217 |
+
"rewards/accuracy_reward/std": 0.4132432878017426,
|
| 2218 |
+
"rewards/cosine_scaled_reward/mean": 0.7792598366737366,
|
| 2219 |
+
"rewards/cosine_scaled_reward/std": 0.4132158994674683,
|
| 2220 |
+
"rewards/format_reward/mean": 0.9966517806053161,
|
| 2221 |
+
"rewards/format_reward/std": 0.05593275725841522,
|
| 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": 0.003833651178918163,
|
| 2231 |
+
"completions/max_length": 2048.0,
|
| 2232 |
+
"completions/max_terminated_length": 1935.2,
|
| 2233 |
+
"completions/mean_length": 615.3019409179688,
|
| 2234 |
+
"completions/mean_terminated_length": 609.7786987304687,
|
| 2235 |
+
"completions/min_length": 197.0,
|
| 2236 |
+
"completions/min_terminated_length": 197.0,
|
| 2237 |
+
"epoch": 4.995499549954996,
|
| 2238 |
+
"grad_norm": 0.15891171991825104,
|
| 2239 |
+
"learning_rate": 1.95785359849987e-11,
|
| 2240 |
+
"loss": 0.01,
|
| 2241 |
+
"num_tokens": 285681294.0,
|
| 2242 |
+
"reward": 1.6595622539520263,
|
| 2243 |
+
"reward_std": 0.2522480070590973,
|
| 2244 |
+
"rewards/accuracy_reward/mean": 0.7747767686843872,
|
| 2245 |
+
"rewards/accuracy_reward/std": 0.41717290282249453,
|
| 2246 |
+
"rewards/cosine_scaled_reward/mean": 0.7740350365638733,
|
| 2247 |
+
"rewards/cosine_scaled_reward/std": 0.4170451521873474,
|
| 2248 |
+
"rewards/format_reward/mean": 0.9955357193946839,
|
| 2249 |
+
"rewards/format_reward/std": 0.06459032222628594,
|
| 2250 |
+
"step": 395
|
| 2251 |
+
},
|
| 2252 |
+
{
|
| 2253 |
+
"epoch": 4.995499549954996,
|
| 2254 |
+
"step": 395,
|
| 2255 |
+
"total_flos": 0.0,
|
| 2256 |
+
"train_loss": 0.011041237461038782,
|
| 2257 |
+
"train_runtime": 79107.9261,
|
| 2258 |
+
"train_samples_per_second": 0.562,
|
| 2259 |
+
"train_steps_per_second": 0.005
|
| 2260 |
+
}
|
| 2261 |
+
],
|
| 2262 |
+
"logging_steps": 5,
|
| 2263 |
+
"max_steps": 395,
|
| 2264 |
+
"num_input_tokens_seen": 285681294,
|
| 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 |
+
}
|