Model save
Browse files- README.md +67 -0
- all_results.json +8 -0
- generation_config.json +6 -0
- train_results.json +8 -0
- trainer_state.json +2292 -0
README.md
ADDED
|
@@ -0,0 +1,67 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
library_name: transformers
|
| 3 |
+
model_name: Qwen2.5-7B-Open-R1-GRPO-Connected-50000-easy-3
|
| 4 |
+
tags:
|
| 5 |
+
- generated_from_trainer
|
| 6 |
+
- trl
|
| 7 |
+
- grpo
|
| 8 |
+
licence: license
|
| 9 |
+
---
|
| 10 |
+
|
| 11 |
+
# Model Card for Qwen2.5-7B-Open-R1-GRPO-Connected-50000-easy-3
|
| 12 |
+
|
| 13 |
+
This model is a fine-tuned version of [None](https://huggingface.co/None).
|
| 14 |
+
It has been trained using [TRL](https://github.com/huggingface/trl).
|
| 15 |
+
|
| 16 |
+
## Quick start
|
| 17 |
+
|
| 18 |
+
```python
|
| 19 |
+
from transformers import pipeline
|
| 20 |
+
|
| 21 |
+
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?"
|
| 22 |
+
generator = pipeline("text-generation", model="AdAstraAbyssoque/Qwen2.5-7B-Open-R1-GRPO-Connected-50000-easy-3", device="cuda")
|
| 23 |
+
output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
|
| 24 |
+
print(output["generated_text"])
|
| 25 |
+
```
|
| 26 |
+
|
| 27 |
+
## Training procedure
|
| 28 |
+
|
| 29 |
+
[<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/bowen_liu-hkust/huggingface/runs/7aewc3l7)
|
| 30 |
+
|
| 31 |
+
|
| 32 |
+
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).
|
| 33 |
+
|
| 34 |
+
### Framework versions
|
| 35 |
+
|
| 36 |
+
- TRL: 0.16.0.dev0
|
| 37 |
+
- Transformers: 4.49.0
|
| 38 |
+
- Pytorch: 2.5.1
|
| 39 |
+
- Datasets: 3.3.2
|
| 40 |
+
- Tokenizers: 0.21.0
|
| 41 |
+
|
| 42 |
+
## Citations
|
| 43 |
+
|
| 44 |
+
Cite GRPO as:
|
| 45 |
+
|
| 46 |
+
```bibtex
|
| 47 |
+
@article{zhihong2024deepseekmath,
|
| 48 |
+
title = {{DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models}},
|
| 49 |
+
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},
|
| 50 |
+
year = 2024,
|
| 51 |
+
eprint = {arXiv:2402.03300},
|
| 52 |
+
}
|
| 53 |
+
|
| 54 |
+
```
|
| 55 |
+
|
| 56 |
+
Cite TRL as:
|
| 57 |
+
|
| 58 |
+
```bibtex
|
| 59 |
+
@misc{vonwerra2022trl,
|
| 60 |
+
title = {{TRL: Transformer Reinforcement Learning}},
|
| 61 |
+
author = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang and Kashif Rasul and Quentin Gallouédec},
|
| 62 |
+
year = 2020,
|
| 63 |
+
journal = {GitHub repository},
|
| 64 |
+
publisher = {GitHub},
|
| 65 |
+
howpublished = {\url{https://github.com/huggingface/trl}}
|
| 66 |
+
}
|
| 67 |
+
```
|
all_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"total_flos": 0.0,
|
| 3 |
+
"train_loss": 0.31898806129537965,
|
| 4 |
+
"train_runtime": 4133.207,
|
| 5 |
+
"train_samples": 50000,
|
| 6 |
+
"train_samples_per_second": 2.032,
|
| 7 |
+
"train_steps_per_second": 0.036
|
| 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": "4.49.0"
|
| 6 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"total_flos": 0.0,
|
| 3 |
+
"train_loss": 0.31898806129537965,
|
| 4 |
+
"train_runtime": 4133.207,
|
| 5 |
+
"train_samples": 50000,
|
| 6 |
+
"train_samples_per_second": 2.032,
|
| 7 |
+
"train_steps_per_second": 0.036
|
| 8 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,2292 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 0.024,
|
| 5 |
+
"eval_steps": 500,
|
| 6 |
+
"global_step": 150,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"clip_ratio": 0.0,
|
| 13 |
+
"completion_length": 423.1785888671875,
|
| 14 |
+
"epoch": 0.00016,
|
| 15 |
+
"grad_norm": 1.5622539520263672,
|
| 16 |
+
"kl": 0.0,
|
| 17 |
+
"learning_rate": 4.999451708687114e-06,
|
| 18 |
+
"loss": 0.084,
|
| 19 |
+
"reward": 0.3575893044471741,
|
| 20 |
+
"reward_std": 0.4995622932910919,
|
| 21 |
+
"rewards/connected_reward": 0.1919642984867096,
|
| 22 |
+
"rewards/format_reward": 0.125,
|
| 23 |
+
"rewards/tag_count_reward": 0.3437500298023224,
|
| 24 |
+
"step": 1
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"clip_ratio": 0.0,
|
| 28 |
+
"completion_length": 328.01788330078125,
|
| 29 |
+
"epoch": 0.00032,
|
| 30 |
+
"grad_norm": 1.7536635398864746,
|
| 31 |
+
"kl": 0.00445556640625,
|
| 32 |
+
"learning_rate": 4.997807075247147e-06,
|
| 33 |
+
"loss": 0.0909,
|
| 34 |
+
"reward": 0.8333333134651184,
|
| 35 |
+
"reward_std": 0.7626916766166687,
|
| 36 |
+
"rewards/connected_reward": 0.5029762387275696,
|
| 37 |
+
"rewards/format_reward": 0.2857142984867096,
|
| 38 |
+
"rewards/tag_count_reward": 0.625,
|
| 39 |
+
"step": 2
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"clip_ratio": 0.0,
|
| 43 |
+
"completion_length": 365.0714416503906,
|
| 44 |
+
"epoch": 0.00048,
|
| 45 |
+
"grad_norm": 1.7543888092041016,
|
| 46 |
+
"kl": 0.0247802734375,
|
| 47 |
+
"learning_rate": 4.9950668210706795e-06,
|
| 48 |
+
"loss": 0.1728,
|
| 49 |
+
"reward": 1.1412203311920166,
|
| 50 |
+
"reward_std": 0.7671074271202087,
|
| 51 |
+
"rewards/connected_reward": 0.6309523582458496,
|
| 52 |
+
"rewards/format_reward": 0.535714328289032,
|
| 53 |
+
"rewards/tag_count_reward": 0.8080357313156128,
|
| 54 |
+
"step": 3
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"clip_ratio": 0.0,
|
| 58 |
+
"completion_length": 359.2500305175781,
|
| 59 |
+
"epoch": 0.00064,
|
| 60 |
+
"grad_norm": 1.619277000427246,
|
| 61 |
+
"kl": 0.03369140625,
|
| 62 |
+
"learning_rate": 4.9912321481237616e-06,
|
| 63 |
+
"loss": 0.1173,
|
| 64 |
+
"reward": 1.1927084922790527,
|
| 65 |
+
"reward_std": 0.7748870253562927,
|
| 66 |
+
"rewards/connected_reward": 0.7127975821495056,
|
| 67 |
+
"rewards/format_reward": 0.5178571939468384,
|
| 68 |
+
"rewards/tag_count_reward": 0.7366071939468384,
|
| 69 |
+
"step": 4
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"clip_ratio": 0.0,
|
| 73 |
+
"completion_length": 246.6785888671875,
|
| 74 |
+
"epoch": 0.0008,
|
| 75 |
+
"grad_norm": 2.3274552822113037,
|
| 76 |
+
"kl": 0.0732421875,
|
| 77 |
+
"learning_rate": 4.986304738420684e-06,
|
| 78 |
+
"loss": 0.088,
|
| 79 |
+
"reward": 1.4244047403335571,
|
| 80 |
+
"reward_std": 0.7808605432510376,
|
| 81 |
+
"rewards/connected_reward": 0.8422619700431824,
|
| 82 |
+
"rewards/format_reward": 0.660714328289032,
|
| 83 |
+
"rewards/tag_count_reward": 0.8392857313156128,
|
| 84 |
+
"step": 5
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"clip_ratio": 0.0,
|
| 88 |
+
"completion_length": 256.3214416503906,
|
| 89 |
+
"epoch": 0.00096,
|
| 90 |
+
"grad_norm": 5.211669921875,
|
| 91 |
+
"kl": 0.1904296875,
|
| 92 |
+
"learning_rate": 4.980286753286196e-06,
|
| 93 |
+
"loss": 0.0374,
|
| 94 |
+
"reward": 1.3794643878936768,
|
| 95 |
+
"reward_std": 0.6322147846221924,
|
| 96 |
+
"rewards/connected_reward": 0.7660714983940125,
|
| 97 |
+
"rewards/format_reward": 0.6964285969734192,
|
| 98 |
+
"rewards/tag_count_reward": 0.8839285969734192,
|
| 99 |
+
"step": 6
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"clip_ratio": 0.0,
|
| 103 |
+
"completion_length": 234.5178680419922,
|
| 104 |
+
"epoch": 0.00112,
|
| 105 |
+
"grad_norm": 7.583590984344482,
|
| 106 |
+
"kl": 0.2197265625,
|
| 107 |
+
"learning_rate": 4.973180832407471e-06,
|
| 108 |
+
"loss": 0.178,
|
| 109 |
+
"reward": 1.4800595045089722,
|
| 110 |
+
"reward_std": 0.7939525246620178,
|
| 111 |
+
"rewards/connected_reward": 0.867559552192688,
|
| 112 |
+
"rewards/format_reward": 0.6785714626312256,
|
| 113 |
+
"rewards/tag_count_reward": 0.910714328289032,
|
| 114 |
+
"step": 7
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"clip_ratio": 0.0,
|
| 118 |
+
"completion_length": 250.96429443359375,
|
| 119 |
+
"epoch": 0.00128,
|
| 120 |
+
"grad_norm": 2.215085029602051,
|
| 121 |
+
"kl": 0.13671875,
|
| 122 |
+
"learning_rate": 4.964990092676263e-06,
|
| 123 |
+
"loss": 0.1382,
|
| 124 |
+
"reward": 1.7797619104385376,
|
| 125 |
+
"reward_std": 0.8585469722747803,
|
| 126 |
+
"rewards/connected_reward": 1.1279761791229248,
|
| 127 |
+
"rewards/format_reward": 0.7678571939468384,
|
| 128 |
+
"rewards/tag_count_reward": 0.8928571939468384,
|
| 129 |
+
"step": 8
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"clip_ratio": 0.0,
|
| 133 |
+
"completion_length": 282.08929443359375,
|
| 134 |
+
"epoch": 0.00144,
|
| 135 |
+
"grad_norm": 2.3200507164001465,
|
| 136 |
+
"kl": 0.1650390625,
|
| 137 |
+
"learning_rate": 4.9557181268217225e-06,
|
| 138 |
+
"loss": 0.2382,
|
| 139 |
+
"reward": 1.753422737121582,
|
| 140 |
+
"reward_std": 0.8391357660293579,
|
| 141 |
+
"rewards/connected_reward": 1.0386905670166016,
|
| 142 |
+
"rewards/format_reward": 0.8750000596046448,
|
| 143 |
+
"rewards/tag_count_reward": 0.9241071939468384,
|
| 144 |
+
"step": 9
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"clip_ratio": 0.0,
|
| 148 |
+
"completion_length": 306.64288330078125,
|
| 149 |
+
"epoch": 0.0016,
|
| 150 |
+
"grad_norm": 2.2042906284332275,
|
| 151 |
+
"kl": 0.123046875,
|
| 152 |
+
"learning_rate": 4.9453690018345144e-06,
|
| 153 |
+
"loss": 0.0315,
|
| 154 |
+
"reward": 1.4244047403335571,
|
| 155 |
+
"reward_std": 0.8331058025360107,
|
| 156 |
+
"rewards/connected_reward": 0.7226191163063049,
|
| 157 |
+
"rewards/format_reward": 0.8571429252624512,
|
| 158 |
+
"rewards/tag_count_reward": 0.910714328289032,
|
| 159 |
+
"step": 10
|
| 160 |
+
},
|
| 161 |
+
{
|
| 162 |
+
"clip_ratio": 0.0,
|
| 163 |
+
"completion_length": 361.2500305175781,
|
| 164 |
+
"epoch": 0.00176,
|
| 165 |
+
"grad_norm": 1.6821162700653076,
|
| 166 |
+
"kl": 0.138671875,
|
| 167 |
+
"learning_rate": 4.933947257182901e-06,
|
| 168 |
+
"loss": 0.0162,
|
| 169 |
+
"reward": 1.227083444595337,
|
| 170 |
+
"reward_std": 0.550583004951477,
|
| 171 |
+
"rewards/connected_reward": 0.4940476715564728,
|
| 172 |
+
"rewards/format_reward": 0.8928571939468384,
|
| 173 |
+
"rewards/tag_count_reward": 0.9553571939468384,
|
| 174 |
+
"step": 11
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"clip_ratio": 0.0,
|
| 178 |
+
"completion_length": 392.6785888671875,
|
| 179 |
+
"epoch": 0.00192,
|
| 180 |
+
"grad_norm": 1.718521237373352,
|
| 181 |
+
"kl": 0.12158203125,
|
| 182 |
+
"learning_rate": 4.921457902821578e-06,
|
| 183 |
+
"loss": -0.0295,
|
| 184 |
+
"reward": 1.6671130657196045,
|
| 185 |
+
"reward_std": 0.7108415961265564,
|
| 186 |
+
"rewards/connected_reward": 0.9630953073501587,
|
| 187 |
+
"rewards/format_reward": 0.8214285969734192,
|
| 188 |
+
"rewards/tag_count_reward": 0.9776785969734192,
|
| 189 |
+
"step": 12
|
| 190 |
+
},
|
| 191 |
+
{
|
| 192 |
+
"clip_ratio": 0.0,
|
| 193 |
+
"completion_length": 373.5000305175781,
|
| 194 |
+
"epoch": 0.00208,
|
| 195 |
+
"grad_norm": 1.7214635610580444,
|
| 196 |
+
"kl": 0.1328125,
|
| 197 |
+
"learning_rate": 4.907906416994146e-06,
|
| 198 |
+
"loss": 0.0016,
|
| 199 |
+
"reward": 1.7629464864730835,
|
| 200 |
+
"reward_std": 0.7587335705757141,
|
| 201 |
+
"rewards/connected_reward": 1.0848214626312256,
|
| 202 |
+
"rewards/format_reward": 0.785714328289032,
|
| 203 |
+
"rewards/tag_count_reward": 0.9508929252624512,
|
| 204 |
+
"step": 13
|
| 205 |
+
},
|
| 206 |
+
{
|
| 207 |
+
"clip_ratio": 0.0,
|
| 208 |
+
"completion_length": 434.39288330078125,
|
| 209 |
+
"epoch": 0.00224,
|
| 210 |
+
"grad_norm": 1.6192271709442139,
|
| 211 |
+
"kl": 0.1416015625,
|
| 212 |
+
"learning_rate": 4.893298743830168e-06,
|
| 213 |
+
"loss": 0.08,
|
| 214 |
+
"reward": 1.7056547403335571,
|
| 215 |
+
"reward_std": 0.5545923709869385,
|
| 216 |
+
"rewards/connected_reward": 1.0029762983322144,
|
| 217 |
+
"rewards/format_reward": 0.8214285969734192,
|
| 218 |
+
"rewards/tag_count_reward": 0.973214328289032,
|
| 219 |
+
"step": 14
|
| 220 |
+
},
|
| 221 |
+
{
|
| 222 |
+
"clip_ratio": 0.0,
|
| 223 |
+
"completion_length": 426.3750305175781,
|
| 224 |
+
"epoch": 0.0024,
|
| 225 |
+
"grad_norm": 1.4105689525604248,
|
| 226 |
+
"kl": 0.1298828125,
|
| 227 |
+
"learning_rate": 4.8776412907378845e-06,
|
| 228 |
+
"loss": -0.0215,
|
| 229 |
+
"reward": 1.6488096714019775,
|
| 230 |
+
"reward_std": 0.5783611536026001,
|
| 231 |
+
"rewards/connected_reward": 0.9666666984558105,
|
| 232 |
+
"rewards/format_reward": 0.785714328289032,
|
| 233 |
+
"rewards/tag_count_reward": 0.9642857313156128,
|
| 234 |
+
"step": 15
|
| 235 |
+
},
|
| 236 |
+
{
|
| 237 |
+
"clip_ratio": 0.0,
|
| 238 |
+
"completion_length": 499.39288330078125,
|
| 239 |
+
"epoch": 0.00256,
|
| 240 |
+
"grad_norm": 1.6216644048690796,
|
| 241 |
+
"kl": 0.126953125,
|
| 242 |
+
"learning_rate": 4.860940925593703e-06,
|
| 243 |
+
"loss": -0.0328,
|
| 244 |
+
"reward": 1.387202501296997,
|
| 245 |
+
"reward_std": 0.5975491404533386,
|
| 246 |
+
"rewards/connected_reward": 0.6934524178504944,
|
| 247 |
+
"rewards/format_reward": 0.8035714626312256,
|
| 248 |
+
"rewards/tag_count_reward": 0.973214328289032,
|
| 249 |
+
"step": 16
|
| 250 |
+
},
|
| 251 |
+
{
|
| 252 |
+
"clip_ratio": 0.0,
|
| 253 |
+
"completion_length": 491.76788330078125,
|
| 254 |
+
"epoch": 0.00272,
|
| 255 |
+
"grad_norm": 1.512536883354187,
|
| 256 |
+
"kl": 0.130859375,
|
| 257 |
+
"learning_rate": 4.84320497372973e-06,
|
| 258 |
+
"loss": 0.0206,
|
| 259 |
+
"reward": 1.6599701642990112,
|
| 260 |
+
"reward_std": 0.5700646638870239,
|
| 261 |
+
"rewards/connected_reward": 0.9175595045089722,
|
| 262 |
+
"rewards/format_reward": 0.8928571939468384,
|
| 263 |
+
"rewards/tag_count_reward": 0.9866071939468384,
|
| 264 |
+
"step": 17
|
| 265 |
+
},
|
| 266 |
+
{
|
| 267 |
+
"clip_ratio": 0.0,
|
| 268 |
+
"completion_length": 455.5714416503906,
|
| 269 |
+
"epoch": 0.00288,
|
| 270 |
+
"grad_norm": 233.13291931152344,
|
| 271 |
+
"kl": 6.40625,
|
| 272 |
+
"learning_rate": 4.824441214720629e-06,
|
| 273 |
+
"loss": 0.2602,
|
| 274 |
+
"reward": 1.8773809671401978,
|
| 275 |
+
"reward_std": 0.8444122076034546,
|
| 276 |
+
"rewards/connected_reward": 1.1398810148239136,
|
| 277 |
+
"rewards/format_reward": 0.8750000596046448,
|
| 278 |
+
"rewards/tag_count_reward": 1.0,
|
| 279 |
+
"step": 18
|
| 280 |
+
},
|
| 281 |
+
{
|
| 282 |
+
"clip_ratio": 0.0,
|
| 283 |
+
"completion_length": 523.2857666015625,
|
| 284 |
+
"epoch": 0.00304,
|
| 285 |
+
"grad_norm": 1.4070404767990112,
|
| 286 |
+
"kl": 0.1328125,
|
| 287 |
+
"learning_rate": 4.804657878971252e-06,
|
| 288 |
+
"loss": -0.006,
|
| 289 |
+
"reward": 1.7507867813110352,
|
| 290 |
+
"reward_std": 0.6262676119804382,
|
| 291 |
+
"rewards/connected_reward": 1.0378402471542358,
|
| 292 |
+
"rewards/format_reward": 0.8392857313156128,
|
| 293 |
+
"rewards/tag_count_reward": 0.9776785969734192,
|
| 294 |
+
"step": 19
|
| 295 |
+
},
|
| 296 |
+
{
|
| 297 |
+
"clip_ratio": 0.0,
|
| 298 |
+
"completion_length": 505.01788330078125,
|
| 299 |
+
"epoch": 0.0032,
|
| 300 |
+
"grad_norm": 2.4147706031799316,
|
| 301 |
+
"kl": 0.1533203125,
|
| 302 |
+
"learning_rate": 4.783863644106502e-06,
|
| 303 |
+
"loss": 0.0332,
|
| 304 |
+
"reward": 1.725595235824585,
|
| 305 |
+
"reward_std": 0.7785912156105042,
|
| 306 |
+
"rewards/connected_reward": 1.0193452835083008,
|
| 307 |
+
"rewards/format_reward": 0.8392857313156128,
|
| 308 |
+
"rewards/tag_count_reward": 0.9553571939468384,
|
| 309 |
+
"step": 20
|
| 310 |
+
},
|
| 311 |
+
{
|
| 312 |
+
"clip_ratio": 0.0,
|
| 313 |
+
"completion_length": 438.2321472167969,
|
| 314 |
+
"epoch": 0.00336,
|
| 315 |
+
"grad_norm": 1.6368823051452637,
|
| 316 |
+
"kl": 0.1630859375,
|
| 317 |
+
"learning_rate": 4.762067631165049e-06,
|
| 318 |
+
"loss": 0.0295,
|
| 319 |
+
"reward": 1.9540605545043945,
|
| 320 |
+
"reward_std": 0.8223637938499451,
|
| 321 |
+
"rewards/connected_reward": 1.2402212619781494,
|
| 322 |
+
"rewards/format_reward": 0.8571429252624512,
|
| 323 |
+
"rewards/tag_count_reward": 0.9508929252624512,
|
| 324 |
+
"step": 21
|
| 325 |
+
},
|
| 326 |
+
{
|
| 327 |
+
"clip_ratio": 0.0,
|
| 328 |
+
"completion_length": 453.8750305175781,
|
| 329 |
+
"epoch": 0.00352,
|
| 330 |
+
"grad_norm": 1.7482272386550903,
|
| 331 |
+
"kl": 0.1572265625,
|
| 332 |
+
"learning_rate": 4.7392794005985324e-06,
|
| 333 |
+
"loss": 0.0902,
|
| 334 |
+
"reward": 1.8142857551574707,
|
| 335 |
+
"reward_std": 0.8429319858551025,
|
| 336 |
+
"rewards/connected_reward": 1.133928656578064,
|
| 337 |
+
"rewards/format_reward": 0.8035714626312256,
|
| 338 |
+
"rewards/tag_count_reward": 0.9285714626312256,
|
| 339 |
+
"step": 22
|
| 340 |
+
},
|
| 341 |
+
{
|
| 342 |
+
"clip_ratio": 0.0,
|
| 343 |
+
"completion_length": 466.607177734375,
|
| 344 |
+
"epoch": 0.00368,
|
| 345 |
+
"grad_norm": 1.5820465087890625,
|
| 346 |
+
"kl": 0.1435546875,
|
| 347 |
+
"learning_rate": 4.715508948078037e-06,
|
| 348 |
+
"loss": 0.0366,
|
| 349 |
+
"reward": 1.6919643878936768,
|
| 350 |
+
"reward_std": 0.7337676882743835,
|
| 351 |
+
"rewards/connected_reward": 0.9741071462631226,
|
| 352 |
+
"rewards/format_reward": 0.8571429252624512,
|
| 353 |
+
"rewards/tag_count_reward": 0.9642857313156128,
|
| 354 |
+
"step": 23
|
| 355 |
+
},
|
| 356 |
+
{
|
| 357 |
+
"clip_ratio": 0.0,
|
| 358 |
+
"completion_length": 436.9821472167969,
|
| 359 |
+
"epoch": 0.00384,
|
| 360 |
+
"grad_norm": 1.3989168405532837,
|
| 361 |
+
"kl": 0.1650390625,
|
| 362 |
+
"learning_rate": 4.690766700109659e-06,
|
| 363 |
+
"loss": -0.0232,
|
| 364 |
+
"reward": 2.0504465103149414,
|
| 365 |
+
"reward_std": 0.7145797610282898,
|
| 366 |
+
"rewards/connected_reward": 1.3071428537368774,
|
| 367 |
+
"rewards/format_reward": 0.910714328289032,
|
| 368 |
+
"rewards/tag_count_reward": 0.9598214626312256,
|
| 369 |
+
"step": 24
|
| 370 |
+
},
|
| 371 |
+
{
|
| 372 |
+
"clip_ratio": 0.0,
|
| 373 |
+
"completion_length": 507.6964416503906,
|
| 374 |
+
"epoch": 0.004,
|
| 375 |
+
"grad_norm": 1.4435826539993286,
|
| 376 |
+
"kl": 0.1484375,
|
| 377 |
+
"learning_rate": 4.665063509461098e-06,
|
| 378 |
+
"loss": 0.0336,
|
| 379 |
+
"reward": 1.3611607551574707,
|
| 380 |
+
"reward_std": 0.5452148914337158,
|
| 381 |
+
"rewards/connected_reward": 0.5625,
|
| 382 |
+
"rewards/format_reward": 1.0,
|
| 383 |
+
"rewards/tag_count_reward": 0.9955357313156128,
|
| 384 |
+
"step": 25
|
| 385 |
+
},
|
| 386 |
+
{
|
| 387 |
+
"clip_ratio": 0.0,
|
| 388 |
+
"completion_length": 416.6607360839844,
|
| 389 |
+
"epoch": 0.00416,
|
| 390 |
+
"grad_norm": 1.6018980741500854,
|
| 391 |
+
"kl": 0.1513671875,
|
| 392 |
+
"learning_rate": 4.638410650401267e-06,
|
| 393 |
+
"loss": -0.0475,
|
| 394 |
+
"reward": 1.834970235824585,
|
| 395 |
+
"reward_std": 0.934838056564331,
|
| 396 |
+
"rewards/connected_reward": 1.086309552192688,
|
| 397 |
+
"rewards/format_reward": 0.910714328289032,
|
| 398 |
+
"rewards/tag_count_reward": 0.9776785969734192,
|
| 399 |
+
"step": 26
|
| 400 |
+
},
|
| 401 |
+
{
|
| 402 |
+
"clip_ratio": 0.0,
|
| 403 |
+
"completion_length": 433.6250305175781,
|
| 404 |
+
"epoch": 0.00432,
|
| 405 |
+
"grad_norm": 1.635309100151062,
|
| 406 |
+
"kl": 0.15625,
|
| 407 |
+
"learning_rate": 4.610819813755038e-06,
|
| 408 |
+
"loss": 0.0671,
|
| 409 |
+
"reward": 1.7008929252624512,
|
| 410 |
+
"reward_std": 0.8552356362342834,
|
| 411 |
+
"rewards/connected_reward": 0.9508929252624512,
|
| 412 |
+
"rewards/format_reward": 0.910714328289032,
|
| 413 |
+
"rewards/tag_count_reward": 0.9821429252624512,
|
| 414 |
+
"step": 27
|
| 415 |
+
},
|
| 416 |
+
{
|
| 417 |
+
"clip_ratio": 0.0,
|
| 418 |
+
"completion_length": 446.3214416503906,
|
| 419 |
+
"epoch": 0.00448,
|
| 420 |
+
"grad_norm": 1.4390822649002075,
|
| 421 |
+
"kl": 0.15234375,
|
| 422 |
+
"learning_rate": 4.582303101775249e-06,
|
| 423 |
+
"loss": 0.0515,
|
| 424 |
+
"reward": 1.7136904001235962,
|
| 425 |
+
"reward_std": 0.6487642526626587,
|
| 426 |
+
"rewards/connected_reward": 0.9315477609634399,
|
| 427 |
+
"rewards/format_reward": 0.9642857313156128,
|
| 428 |
+
"rewards/tag_count_reward": 1.0,
|
| 429 |
+
"step": 28
|
| 430 |
+
},
|
| 431 |
+
{
|
| 432 |
+
"clip_ratio": 0.0,
|
| 433 |
+
"completion_length": 428.0000305175781,
|
| 434 |
+
"epoch": 0.00464,
|
| 435 |
+
"grad_norm": 1.5933189392089844,
|
| 436 |
+
"kl": 0.1513671875,
|
| 437 |
+
"learning_rate": 4.55287302283426e-06,
|
| 438 |
+
"loss": 0.0457,
|
| 439 |
+
"reward": 1.5874149799346924,
|
| 440 |
+
"reward_std": 0.7879763245582581,
|
| 441 |
+
"rewards/connected_reward": 0.7963436245918274,
|
| 442 |
+
"rewards/format_reward": 0.9821429252624512,
|
| 443 |
+
"rewards/tag_count_reward": 1.0,
|
| 444 |
+
"step": 29
|
| 445 |
+
},
|
| 446 |
+
{
|
| 447 |
+
"clip_ratio": 0.0,
|
| 448 |
+
"completion_length": 478.607177734375,
|
| 449 |
+
"epoch": 0.0048,
|
| 450 |
+
"grad_norm": 1.511976718902588,
|
| 451 |
+
"kl": 0.1435546875,
|
| 452 |
+
"learning_rate": 4.522542485937369e-06,
|
| 453 |
+
"loss": 0.0491,
|
| 454 |
+
"reward": 1.3787201642990112,
|
| 455 |
+
"reward_std": 0.7991003394126892,
|
| 456 |
+
"rewards/connected_reward": 0.644345223903656,
|
| 457 |
+
"rewards/format_reward": 0.8928571939468384,
|
| 458 |
+
"rewards/tag_count_reward": 0.9598214626312256,
|
| 459 |
+
"step": 30
|
| 460 |
+
},
|
| 461 |
+
{
|
| 462 |
+
"clip_ratio": 0.0,
|
| 463 |
+
"completion_length": 430.71429443359375,
|
| 464 |
+
"epoch": 0.00496,
|
| 465 |
+
"grad_norm": 1.6736035346984863,
|
| 466 |
+
"kl": 0.16015625,
|
| 467 |
+
"learning_rate": 4.491324795060491e-06,
|
| 468 |
+
"loss": -0.0175,
|
| 469 |
+
"reward": 1.6544642448425293,
|
| 470 |
+
"reward_std": 0.6817477941513062,
|
| 471 |
+
"rewards/connected_reward": 0.9151785969734192,
|
| 472 |
+
"rewards/format_reward": 0.910714328289032,
|
| 473 |
+
"rewards/tag_count_reward": 0.9464285969734192,
|
| 474 |
+
"step": 31
|
| 475 |
+
},
|
| 476 |
+
{
|
| 477 |
+
"clip_ratio": 0.0,
|
| 478 |
+
"completion_length": 396.0000305175781,
|
| 479 |
+
"epoch": 0.00512,
|
| 480 |
+
"grad_norm": 1.8634084463119507,
|
| 481 |
+
"kl": 0.1513671875,
|
| 482 |
+
"learning_rate": 4.4592336433146e-06,
|
| 483 |
+
"loss": 0.1726,
|
| 484 |
+
"reward": 1.756250023841858,
|
| 485 |
+
"reward_std": 0.9827747941017151,
|
| 486 |
+
"rewards/connected_reward": 1.0178571939468384,
|
| 487 |
+
"rewards/format_reward": 0.8928571939468384,
|
| 488 |
+
"rewards/tag_count_reward": 0.973214328289032,
|
| 489 |
+
"step": 32
|
| 490 |
+
},
|
| 491 |
+
{
|
| 492 |
+
"clip_ratio": 0.0,
|
| 493 |
+
"completion_length": 371.26788330078125,
|
| 494 |
+
"epoch": 0.00528,
|
| 495 |
+
"grad_norm": 1.7837437391281128,
|
| 496 |
+
"kl": 0.1484375,
|
| 497 |
+
"learning_rate": 4.426283106939474e-06,
|
| 498 |
+
"loss": -0.023,
|
| 499 |
+
"reward": 1.8660714626312256,
|
| 500 |
+
"reward_std": 0.8486083149909973,
|
| 501 |
+
"rewards/connected_reward": 1.0982143878936768,
|
| 502 |
+
"rewards/format_reward": 0.9464285969734192,
|
| 503 |
+
"rewards/tag_count_reward": 0.9821429252624512,
|
| 504 |
+
"step": 33
|
| 505 |
+
},
|
| 506 |
+
{
|
| 507 |
+
"clip_ratio": 0.0,
|
| 508 |
+
"completion_length": 427.0535888671875,
|
| 509 |
+
"epoch": 0.00544,
|
| 510 |
+
"grad_norm": 5.339620590209961,
|
| 511 |
+
"kl": 0.1552734375,
|
| 512 |
+
"learning_rate": 4.3924876391293915e-06,
|
| 513 |
+
"loss": 0.0789,
|
| 514 |
+
"reward": 1.8375002145767212,
|
| 515 |
+
"reward_std": 0.7801174521446228,
|
| 516 |
+
"rewards/connected_reward": 1.1482142210006714,
|
| 517 |
+
"rewards/format_reward": 0.8214285969734192,
|
| 518 |
+
"rewards/tag_count_reward": 0.9285714626312256,
|
| 519 |
+
"step": 34
|
| 520 |
+
},
|
| 521 |
+
{
|
| 522 |
+
"clip_ratio": 0.0,
|
| 523 |
+
"completion_length": 405.0000305175781,
|
| 524 |
+
"epoch": 0.0056,
|
| 525 |
+
"grad_norm": 3.4218180179595947,
|
| 526 |
+
"kl": 0.1923828125,
|
| 527 |
+
"learning_rate": 4.357862063693486e-06,
|
| 528 |
+
"loss": 0.026,
|
| 529 |
+
"reward": 1.5308035612106323,
|
| 530 |
+
"reward_std": 0.8105348944664001,
|
| 531 |
+
"rewards/connected_reward": 0.8080357313156128,
|
| 532 |
+
"rewards/format_reward": 0.8750000596046448,
|
| 533 |
+
"rewards/tag_count_reward": 0.9508929252624512,
|
| 534 |
+
"step": 35
|
| 535 |
+
},
|
| 536 |
+
{
|
| 537 |
+
"clip_ratio": 0.0,
|
| 538 |
+
"completion_length": 439.5000305175781,
|
| 539 |
+
"epoch": 0.00576,
|
| 540 |
+
"grad_norm": 2.0338456630706787,
|
| 541 |
+
"kl": 0.173828125,
|
| 542 |
+
"learning_rate": 4.322421568553529e-06,
|
| 543 |
+
"loss": 0.1697,
|
| 544 |
+
"reward": 1.2683035135269165,
|
| 545 |
+
"reward_std": 0.7643690705299377,
|
| 546 |
+
"rewards/connected_reward": 0.6214286684989929,
|
| 547 |
+
"rewards/format_reward": 0.7500000596046448,
|
| 548 |
+
"rewards/tag_count_reward": 0.9062500596046448,
|
| 549 |
+
"step": 36
|
| 550 |
+
},
|
| 551 |
+
{
|
| 552 |
+
"clip_ratio": 0.0,
|
| 553 |
+
"completion_length": 401.2321472167969,
|
| 554 |
+
"epoch": 0.00592,
|
| 555 |
+
"grad_norm": 1.8888318538665771,
|
| 556 |
+
"kl": 0.1787109375,
|
| 557 |
+
"learning_rate": 4.286181699082008e-06,
|
| 558 |
+
"loss": 0.1236,
|
| 559 |
+
"reward": 1.7342263460159302,
|
| 560 |
+
"reward_std": 0.9654321670532227,
|
| 561 |
+
"rewards/connected_reward": 1.0208333730697632,
|
| 562 |
+
"rewards/format_reward": 0.8750000596046448,
|
| 563 |
+
"rewards/tag_count_reward": 0.9196429252624512,
|
| 564 |
+
"step": 37
|
| 565 |
+
},
|
| 566 |
+
{
|
| 567 |
+
"clip_ratio": 0.0,
|
| 568 |
+
"completion_length": 473.96429443359375,
|
| 569 |
+
"epoch": 0.00608,
|
| 570 |
+
"grad_norm": 1.792199969291687,
|
| 571 |
+
"kl": 0.1982421875,
|
| 572 |
+
"learning_rate": 4.249158351283414e-06,
|
| 573 |
+
"loss": 0.154,
|
| 574 |
+
"reward": 1.37901771068573,
|
| 575 |
+
"reward_std": 0.8872238397598267,
|
| 576 |
+
"rewards/connected_reward": 0.7366071939468384,
|
| 577 |
+
"rewards/format_reward": 0.7678571939468384,
|
| 578 |
+
"rewards/tag_count_reward": 0.8616071939468384,
|
| 579 |
+
"step": 38
|
| 580 |
+
},
|
| 581 |
+
{
|
| 582 |
+
"clip_ratio": 0.0,
|
| 583 |
+
"completion_length": 512.232177734375,
|
| 584 |
+
"epoch": 0.00624,
|
| 585 |
+
"grad_norm": 2.252385377883911,
|
| 586 |
+
"kl": 0.1826171875,
|
| 587 |
+
"learning_rate": 4.211367764821722e-06,
|
| 588 |
+
"loss": 0.3264,
|
| 589 |
+
"reward": 1.412500023841858,
|
| 590 |
+
"reward_std": 0.7886937856674194,
|
| 591 |
+
"rewards/connected_reward": 0.754464328289032,
|
| 592 |
+
"rewards/format_reward": 0.785714328289032,
|
| 593 |
+
"rewards/tag_count_reward": 0.8839285969734192,
|
| 594 |
+
"step": 39
|
| 595 |
+
},
|
| 596 |
+
{
|
| 597 |
+
"clip_ratio": 0.0,
|
| 598 |
+
"completion_length": 517.7678833007812,
|
| 599 |
+
"epoch": 0.0064,
|
| 600 |
+
"grad_norm": 2.793456554412842,
|
| 601 |
+
"kl": 0.203125,
|
| 602 |
+
"learning_rate": 4.172826515897146e-06,
|
| 603 |
+
"loss": 0.264,
|
| 604 |
+
"reward": 1.2334821224212646,
|
| 605 |
+
"reward_std": 0.624849796295166,
|
| 606 |
+
"rewards/connected_reward": 0.5687500238418579,
|
| 607 |
+
"rewards/format_reward": 0.785714328289032,
|
| 608 |
+
"rewards/tag_count_reward": 0.9062500596046448,
|
| 609 |
+
"step": 40
|
| 610 |
+
},
|
| 611 |
+
{
|
| 612 |
+
"clip_ratio": 0.0,
|
| 613 |
+
"completion_length": 460.5535888671875,
|
| 614 |
+
"epoch": 0.00656,
|
| 615 |
+
"grad_norm": 2.782038688659668,
|
| 616 |
+
"kl": 0.1826171875,
|
| 617 |
+
"learning_rate": 4.133551509975264e-06,
|
| 618 |
+
"loss": 0.3976,
|
| 619 |
+
"reward": 1.6736607551574707,
|
| 620 |
+
"reward_std": 0.9631600379943848,
|
| 621 |
+
"rewards/connected_reward": 1.0625,
|
| 622 |
+
"rewards/format_reward": 0.6785714626312256,
|
| 623 |
+
"rewards/tag_count_reward": 0.9062500596046448,
|
| 624 |
+
"step": 41
|
| 625 |
+
},
|
| 626 |
+
{
|
| 627 |
+
"clip_ratio": 0.0,
|
| 628 |
+
"completion_length": 591.5714721679688,
|
| 629 |
+
"epoch": 0.00672,
|
| 630 |
+
"grad_norm": 1.9981597661972046,
|
| 631 |
+
"kl": 0.2255859375,
|
| 632 |
+
"learning_rate": 4.093559974371725e-06,
|
| 633 |
+
"loss": 0.359,
|
| 634 |
+
"reward": 1.3278273344039917,
|
| 635 |
+
"reward_std": 0.9624171257019043,
|
| 636 |
+
"rewards/connected_reward": 0.7380953431129456,
|
| 637 |
+
"rewards/format_reward": 0.6785714626312256,
|
| 638 |
+
"rewards/tag_count_reward": 0.8348214626312256,
|
| 639 |
+
"step": 42
|
| 640 |
+
},
|
| 641 |
+
{
|
| 642 |
+
"clip_ratio": 0.0,
|
| 643 |
+
"completion_length": 728.2678833007812,
|
| 644 |
+
"epoch": 0.00688,
|
| 645 |
+
"grad_norm": 2.215883731842041,
|
| 646 |
+
"kl": 0.2138671875,
|
| 647 |
+
"learning_rate": 4.052869450695776e-06,
|
| 648 |
+
"loss": 0.4698,
|
| 649 |
+
"reward": 1.158928632736206,
|
| 650 |
+
"reward_std": 0.7686693072319031,
|
| 651 |
+
"rewards/connected_reward": 0.5973214507102966,
|
| 652 |
+
"rewards/format_reward": 0.625,
|
| 653 |
+
"rewards/tag_count_reward": 0.8303571939468384,
|
| 654 |
+
"step": 43
|
| 655 |
+
},
|
| 656 |
+
{
|
| 657 |
+
"clip_ratio": 0.0,
|
| 658 |
+
"completion_length": 780.5357666015625,
|
| 659 |
+
"epoch": 0.00704,
|
| 660 |
+
"grad_norm": 2.267688035964966,
|
| 661 |
+
"kl": 0.2734375,
|
| 662 |
+
"learning_rate": 4.011497787155938e-06,
|
| 663 |
+
"loss": 0.578,
|
| 664 |
+
"reward": 1.3976190090179443,
|
| 665 |
+
"reward_std": 1.026655673980713,
|
| 666 |
+
"rewards/connected_reward": 0.8404762148857117,
|
| 667 |
+
"rewards/format_reward": 0.6428571939468384,
|
| 668 |
+
"rewards/tag_count_reward": 0.785714328289032,
|
| 669 |
+
"step": 44
|
| 670 |
+
},
|
| 671 |
+
{
|
| 672 |
+
"clip_ratio": 0.0,
|
| 673 |
+
"completion_length": 870.9464721679688,
|
| 674 |
+
"epoch": 0.0072,
|
| 675 |
+
"grad_norm": 2.2561347484588623,
|
| 676 |
+
"kl": 0.32421875,
|
| 677 |
+
"learning_rate": 3.969463130731183e-06,
|
| 678 |
+
"loss": 0.5883,
|
| 679 |
+
"reward": 1.2087798118591309,
|
| 680 |
+
"reward_std": 0.9154091477394104,
|
| 681 |
+
"rewards/connected_reward": 0.6610119938850403,
|
| 682 |
+
"rewards/format_reward": 0.6428571939468384,
|
| 683 |
+
"rewards/tag_count_reward": 0.754464328289032,
|
| 684 |
+
"step": 45
|
| 685 |
+
},
|
| 686 |
+
{
|
| 687 |
+
"clip_ratio": 0.0,
|
| 688 |
+
"completion_length": 786.7678833007812,
|
| 689 |
+
"epoch": 0.00736,
|
| 690 |
+
"grad_norm": 1.954925298690796,
|
| 691 |
+
"kl": 0.38671875,
|
| 692 |
+
"learning_rate": 3.92678391921108e-06,
|
| 693 |
+
"loss": 0.671,
|
| 694 |
+
"reward": 1.5040180683135986,
|
| 695 |
+
"reward_std": 1.171386957168579,
|
| 696 |
+
"rewards/connected_reward": 0.9571430087089539,
|
| 697 |
+
"rewards/format_reward": 0.625,
|
| 698 |
+
"rewards/tag_count_reward": 0.7812500596046448,
|
| 699 |
+
"step": 46
|
| 700 |
+
},
|
| 701 |
+
{
|
| 702 |
+
"clip_ratio": 0.0,
|
| 703 |
+
"completion_length": 843.0892944335938,
|
| 704 |
+
"epoch": 0.00752,
|
| 705 |
+
"grad_norm": 1.6565457582473755,
|
| 706 |
+
"kl": 0.375,
|
| 707 |
+
"learning_rate": 3.88347887310836e-06,
|
| 708 |
+
"loss": 0.7635,
|
| 709 |
+
"reward": 1.400744080543518,
|
| 710 |
+
"reward_std": 1.0551815032958984,
|
| 711 |
+
"rewards/connected_reward": 0.8717262148857117,
|
| 712 |
+
"rewards/format_reward": 0.5892857313156128,
|
| 713 |
+
"rewards/tag_count_reward": 0.7812500596046448,
|
| 714 |
+
"step": 47
|
| 715 |
+
},
|
| 716 |
+
{
|
| 717 |
+
"clip_ratio": 0.0,
|
| 718 |
+
"completion_length": 819.2142944335938,
|
| 719 |
+
"epoch": 0.00768,
|
| 720 |
+
"grad_norm": 1.5234383344650269,
|
| 721 |
+
"kl": 0.35546875,
|
| 722 |
+
"learning_rate": 3.839566987447492e-06,
|
| 723 |
+
"loss": 0.5719,
|
| 724 |
+
"reward": 1.422767996788025,
|
| 725 |
+
"reward_std": 1.039860725402832,
|
| 726 |
+
"rewards/connected_reward": 0.9017857313156128,
|
| 727 |
+
"rewards/format_reward": 0.5892857313156128,
|
| 728 |
+
"rewards/tag_count_reward": 0.754464328289032,
|
| 729 |
+
"step": 48
|
| 730 |
+
},
|
| 731 |
+
{
|
| 732 |
+
"clip_ratio": 0.0,
|
| 733 |
+
"completion_length": 1017.5178833007812,
|
| 734 |
+
"epoch": 0.00784,
|
| 735 |
+
"grad_norm": 1.8639246225357056,
|
| 736 |
+
"kl": 0.470703125,
|
| 737 |
+
"learning_rate": 3.795067523432826e-06,
|
| 738 |
+
"loss": 0.4938,
|
| 739 |
+
"reward": 1.012946605682373,
|
| 740 |
+
"reward_std": 1.0206222534179688,
|
| 741 |
+
"rewards/connected_reward": 0.5464286208152771,
|
| 742 |
+
"rewards/format_reward": 0.5178571939468384,
|
| 743 |
+
"rewards/tag_count_reward": 0.691964328289032,
|
| 744 |
+
"step": 49
|
| 745 |
+
},
|
| 746 |
+
{
|
| 747 |
+
"clip_ratio": 0.0,
|
| 748 |
+
"completion_length": 789.6428833007812,
|
| 749 |
+
"epoch": 0.008,
|
| 750 |
+
"grad_norm": 7.7003021240234375,
|
| 751 |
+
"kl": 0.431640625,
|
| 752 |
+
"learning_rate": 3.7500000000000005e-06,
|
| 753 |
+
"loss": 0.6166,
|
| 754 |
+
"reward": 1.5044643878936768,
|
| 755 |
+
"reward_std": 1.1621365547180176,
|
| 756 |
+
"rewards/connected_reward": 0.9642857313156128,
|
| 757 |
+
"rewards/format_reward": 0.625,
|
| 758 |
+
"rewards/tag_count_reward": 0.7589285969734192,
|
| 759 |
+
"step": 50
|
| 760 |
+
},
|
| 761 |
+
{
|
| 762 |
+
"clip_ratio": 0.0,
|
| 763 |
+
"completion_length": 767.1428833007812,
|
| 764 |
+
"epoch": 0.00816,
|
| 765 |
+
"grad_norm": 17.987171173095703,
|
| 766 |
+
"kl": 0.482421875,
|
| 767 |
+
"learning_rate": 3.7043841852542884e-06,
|
| 768 |
+
"loss": 0.6032,
|
| 769 |
+
"reward": 1.5376489162445068,
|
| 770 |
+
"reward_std": 0.8768205642700195,
|
| 771 |
+
"rewards/connected_reward": 0.9755952954292297,
|
| 772 |
+
"rewards/format_reward": 0.660714328289032,
|
| 773 |
+
"rewards/tag_count_reward": 0.7723214626312256,
|
| 774 |
+
"step": 51
|
| 775 |
+
},
|
| 776 |
+
{
|
| 777 |
+
"clip_ratio": 0.0,
|
| 778 |
+
"completion_length": 1012.2500610351562,
|
| 779 |
+
"epoch": 0.00832,
|
| 780 |
+
"grad_norm": 62.031070709228516,
|
| 781 |
+
"kl": 0.90625,
|
| 782 |
+
"learning_rate": 3.658240087799655e-06,
|
| 783 |
+
"loss": 0.7438,
|
| 784 |
+
"reward": 1.2361607551574707,
|
| 785 |
+
"reward_std": 1.0895289182662964,
|
| 786 |
+
"rewards/connected_reward": 0.7866071462631226,
|
| 787 |
+
"rewards/format_reward": 0.5,
|
| 788 |
+
"rewards/tag_count_reward": 0.6651785969734192,
|
| 789 |
+
"step": 52
|
| 790 |
+
},
|
| 791 |
+
{
|
| 792 |
+
"clip_ratio": 0.0,
|
| 793 |
+
"completion_length": 790.0714721679688,
|
| 794 |
+
"epoch": 0.00848,
|
| 795 |
+
"grad_norm": 54.658687591552734,
|
| 796 |
+
"kl": 1.1015625,
|
| 797 |
+
"learning_rate": 3.611587947962319e-06,
|
| 798 |
+
"loss": 0.6992,
|
| 799 |
+
"reward": 1.5057541131973267,
|
| 800 |
+
"reward_std": 1.0789897441864014,
|
| 801 |
+
"rewards/connected_reward": 0.9771825671195984,
|
| 802 |
+
"rewards/format_reward": 0.6071428656578064,
|
| 803 |
+
"rewards/tag_count_reward": 0.7500000596046448,
|
| 804 |
+
"step": 53
|
| 805 |
+
},
|
| 806 |
+
{
|
| 807 |
+
"clip_ratio": 0.0,
|
| 808 |
+
"completion_length": 955.4285888671875,
|
| 809 |
+
"epoch": 0.00864,
|
| 810 |
+
"grad_norm": 5.784551620483398,
|
| 811 |
+
"kl": 0.322265625,
|
| 812 |
+
"learning_rate": 3.564448228912682e-06,
|
| 813 |
+
"loss": 0.5783,
|
| 814 |
+
"reward": 1.15814208984375,
|
| 815 |
+
"reward_std": 0.956221878528595,
|
| 816 |
+
"rewards/connected_reward": 0.6451956033706665,
|
| 817 |
+
"rewards/format_reward": 0.5892857313156128,
|
| 818 |
+
"rewards/tag_count_reward": 0.7276785969734192,
|
| 819 |
+
"step": 54
|
| 820 |
+
},
|
| 821 |
+
{
|
| 822 |
+
"clip_ratio": 0.0,
|
| 823 |
+
"completion_length": 829.9285888671875,
|
| 824 |
+
"epoch": 0.0088,
|
| 825 |
+
"grad_norm": 3.2477169036865234,
|
| 826 |
+
"kl": 0.146484375,
|
| 827 |
+
"learning_rate": 3.516841607689501e-06,
|
| 828 |
+
"loss": 0.5911,
|
| 829 |
+
"reward": 1.6458333730697632,
|
| 830 |
+
"reward_std": 1.0081928968429565,
|
| 831 |
+
"rewards/connected_reward": 1.0967262983322144,
|
| 832 |
+
"rewards/format_reward": 0.6428571939468384,
|
| 833 |
+
"rewards/tag_count_reward": 0.7589285969734192,
|
| 834 |
+
"step": 55
|
| 835 |
+
},
|
| 836 |
+
{
|
| 837 |
+
"clip_ratio": 0.0,
|
| 838 |
+
"completion_length": 627.4642944335938,
|
| 839 |
+
"epoch": 0.00896,
|
| 840 |
+
"grad_norm": 14.7655611038208,
|
| 841 |
+
"kl": 0.1455078125,
|
| 842 |
+
"learning_rate": 3.4687889661302577e-06,
|
| 843 |
+
"loss": 0.831,
|
| 844 |
+
"reward": 2.1366071701049805,
|
| 845 |
+
"reward_std": 0.9424195289611816,
|
| 846 |
+
"rewards/connected_reward": 1.4866071939468384,
|
| 847 |
+
"rewards/format_reward": 0.785714328289032,
|
| 848 |
+
"rewards/tag_count_reward": 0.8571429252624512,
|
| 849 |
+
"step": 56
|
| 850 |
+
},
|
| 851 |
+
{
|
| 852 |
+
"clip_ratio": 0.0,
|
| 853 |
+
"completion_length": 1031.732177734375,
|
| 854 |
+
"epoch": 0.00912,
|
| 855 |
+
"grad_norm": 123.79815673828125,
|
| 856 |
+
"kl": 0.216796875,
|
| 857 |
+
"learning_rate": 3.4203113817116955e-06,
|
| 858 |
+
"loss": 0.7199,
|
| 859 |
+
"reward": 1.3994048833847046,
|
| 860 |
+
"reward_std": 1.1809272766113281,
|
| 861 |
+
"rewards/connected_reward": 0.9047619700431824,
|
| 862 |
+
"rewards/format_reward": 0.5714285969734192,
|
| 863 |
+
"rewards/tag_count_reward": 0.6964285969734192,
|
| 864 |
+
"step": 57
|
| 865 |
+
},
|
| 866 |
+
{
|
| 867 |
+
"clip_ratio": 0.0,
|
| 868 |
+
"completion_length": 1176.5357666015625,
|
| 869 |
+
"epoch": 0.00928,
|
| 870 |
+
"grad_norm": 49.686466217041016,
|
| 871 |
+
"kl": 0.4140625,
|
| 872 |
+
"learning_rate": 3.3714301183045382e-06,
|
| 873 |
+
"loss": 0.573,
|
| 874 |
+
"reward": 1.180803656578064,
|
| 875 |
+
"reward_std": 1.069697380065918,
|
| 876 |
+
"rewards/connected_reward": 0.7767857313156128,
|
| 877 |
+
"rewards/format_reward": 0.4464285969734192,
|
| 878 |
+
"rewards/tag_count_reward": 0.6026785969734192,
|
| 879 |
+
"step": 58
|
| 880 |
+
},
|
| 881 |
+
{
|
| 882 |
+
"clip_ratio": 0.0,
|
| 883 |
+
"completion_length": 1143.0179443359375,
|
| 884 |
+
"epoch": 0.00944,
|
| 885 |
+
"grad_norm": 12.297547340393066,
|
| 886 |
+
"kl": 0.1728515625,
|
| 887 |
+
"learning_rate": 3.3221666168464584e-06,
|
| 888 |
+
"loss": 0.43,
|
| 889 |
+
"reward": 1.1979166269302368,
|
| 890 |
+
"reward_std": 0.9061146378517151,
|
| 891 |
+
"rewards/connected_reward": 0.7523809671401978,
|
| 892 |
+
"rewards/format_reward": 0.5,
|
| 893 |
+
"rewards/tag_count_reward": 0.6517857313156128,
|
| 894 |
+
"step": 59
|
| 895 |
+
},
|
| 896 |
+
{
|
| 897 |
+
"clip_ratio": 0.0,
|
| 898 |
+
"completion_length": 744.482177734375,
|
| 899 |
+
"epoch": 0.0096,
|
| 900 |
+
"grad_norm": 1.3184255361557007,
|
| 901 |
+
"kl": 0.13671875,
|
| 902 |
+
"learning_rate": 3.272542485937369e-06,
|
| 903 |
+
"loss": 0.7022,
|
| 904 |
+
"reward": 1.6336311101913452,
|
| 905 |
+
"reward_std": 1.0127315521240234,
|
| 906 |
+
"rewards/connected_reward": 1.0148810148239136,
|
| 907 |
+
"rewards/format_reward": 0.7500000596046448,
|
| 908 |
+
"rewards/tag_count_reward": 0.8125000596046448,
|
| 909 |
+
"step": 60
|
| 910 |
+
},
|
| 911 |
+
{
|
| 912 |
+
"clip_ratio": 0.0,
|
| 913 |
+
"completion_length": 775.5892944335938,
|
| 914 |
+
"epoch": 0.00976,
|
| 915 |
+
"grad_norm": 0.7490354776382446,
|
| 916 |
+
"kl": 0.12109375,
|
| 917 |
+
"learning_rate": 3.222579492361179e-06,
|
| 918 |
+
"loss": 0.5858,
|
| 919 |
+
"reward": 1.6793580055236816,
|
| 920 |
+
"reward_std": 1.2150492668151855,
|
| 921 |
+
"rewards/connected_reward": 1.073554515838623,
|
| 922 |
+
"rewards/format_reward": 0.7321428656578064,
|
| 923 |
+
"rewards/tag_count_reward": 0.7991071939468384,
|
| 924 |
+
"step": 61
|
| 925 |
+
},
|
| 926 |
+
{
|
| 927 |
+
"clip_ratio": 0.0,
|
| 928 |
+
"completion_length": 409.3571472167969,
|
| 929 |
+
"epoch": 0.00992,
|
| 930 |
+
"grad_norm": 1.7477023601531982,
|
| 931 |
+
"kl": 0.1572265625,
|
| 932 |
+
"learning_rate": 3.1722995515381644e-06,
|
| 933 |
+
"loss": 0.3222,
|
| 934 |
+
"reward": 1.969642996788025,
|
| 935 |
+
"reward_std": 0.9241483807563782,
|
| 936 |
+
"rewards/connected_reward": 1.21875,
|
| 937 |
+
"rewards/format_reward": 0.9285714626312256,
|
| 938 |
+
"rewards/tag_count_reward": 0.9553571939468384,
|
| 939 |
+
"step": 62
|
| 940 |
+
},
|
| 941 |
+
{
|
| 942 |
+
"clip_ratio": 0.0,
|
| 943 |
+
"completion_length": 625.5357666015625,
|
| 944 |
+
"epoch": 0.01008,
|
| 945 |
+
"grad_norm": 11.000500679016113,
|
| 946 |
+
"kl": 0.1787109375,
|
| 947 |
+
"learning_rate": 3.121724717912138e-06,
|
| 948 |
+
"loss": 0.5757,
|
| 949 |
+
"reward": 1.6025298833847046,
|
| 950 |
+
"reward_std": 0.8033274412155151,
|
| 951 |
+
"rewards/connected_reward": 0.944940447807312,
|
| 952 |
+
"rewards/format_reward": 0.8035714626312256,
|
| 953 |
+
"rewards/tag_count_reward": 0.8526785969734192,
|
| 954 |
+
"step": 63
|
| 955 |
+
},
|
| 956 |
+
{
|
| 957 |
+
"clip_ratio": 0.0,
|
| 958 |
+
"completion_length": 288.9107360839844,
|
| 959 |
+
"epoch": 0.01024,
|
| 960 |
+
"grad_norm": 47.38264083862305,
|
| 961 |
+
"kl": 0.6328125,
|
| 962 |
+
"learning_rate": 3.0708771752766397e-06,
|
| 963 |
+
"loss": 0.0216,
|
| 964 |
+
"reward": 1.6721726655960083,
|
| 965 |
+
"reward_std": 0.7864125967025757,
|
| 966 |
+
"rewards/connected_reward": 0.9494048357009888,
|
| 967 |
+
"rewards/format_reward": 0.8750000596046448,
|
| 968 |
+
"rewards/tag_count_reward": 0.9508929252624512,
|
| 969 |
+
"step": 64
|
| 970 |
+
},
|
| 971 |
+
{
|
| 972 |
+
"clip_ratio": 0.0,
|
| 973 |
+
"completion_length": 254.85714721679688,
|
| 974 |
+
"epoch": 0.0104,
|
| 975 |
+
"grad_norm": 13.367755889892578,
|
| 976 |
+
"kl": 0.412109375,
|
| 977 |
+
"learning_rate": 3.019779227044398e-06,
|
| 978 |
+
"loss": -0.0184,
|
| 979 |
+
"reward": 2.1937499046325684,
|
| 980 |
+
"reward_std": 0.6596731543540955,
|
| 981 |
+
"rewards/connected_reward": 1.4285714626312256,
|
| 982 |
+
"rewards/format_reward": 0.9464285969734192,
|
| 983 |
+
"rewards/tag_count_reward": 0.973214328289032,
|
| 984 |
+
"step": 65
|
| 985 |
+
},
|
| 986 |
+
{
|
| 987 |
+
"clip_ratio": 0.0,
|
| 988 |
+
"completion_length": 268.46429443359375,
|
| 989 |
+
"epoch": 0.01056,
|
| 990 |
+
"grad_norm": 31.148229598999023,
|
| 991 |
+
"kl": 0.6484375,
|
| 992 |
+
"learning_rate": 2.9684532864643123e-06,
|
| 993 |
+
"loss": 0.0553,
|
| 994 |
+
"reward": 1.8163691759109497,
|
| 995 |
+
"reward_std": 0.6786539554595947,
|
| 996 |
+
"rewards/connected_reward": 1.0627976655960083,
|
| 997 |
+
"rewards/format_reward": 0.9285714626312256,
|
| 998 |
+
"rewards/tag_count_reward": 0.9642857313156128,
|
| 999 |
+
"step": 66
|
| 1000 |
+
},
|
| 1001 |
+
{
|
| 1002 |
+
"clip_ratio": 0.0,
|
| 1003 |
+
"completion_length": 436.2857360839844,
|
| 1004 |
+
"epoch": 0.01072,
|
| 1005 |
+
"grad_norm": 69.53382110595703,
|
| 1006 |
+
"kl": 0.2431640625,
|
| 1007 |
+
"learning_rate": 2.9169218667902562e-06,
|
| 1008 |
+
"loss": 0.4608,
|
| 1009 |
+
"reward": 2.0997023582458496,
|
| 1010 |
+
"reward_std": 0.7058864235877991,
|
| 1011 |
+
"rewards/connected_reward": 1.3720238208770752,
|
| 1012 |
+
"rewards/format_reward": 0.8928571939468384,
|
| 1013 |
+
"rewards/tag_count_reward": 0.9375000596046448,
|
| 1014 |
+
"step": 67
|
| 1015 |
+
},
|
| 1016 |
+
{
|
| 1017 |
+
"clip_ratio": 0.0,
|
| 1018 |
+
"completion_length": 387.9285888671875,
|
| 1019 |
+
"epoch": 0.01088,
|
| 1020 |
+
"grad_norm": 93.46610260009766,
|
| 1021 |
+
"kl": 0.43359375,
|
| 1022 |
+
"learning_rate": 2.8652075714060296e-06,
|
| 1023 |
+
"loss": 0.1717,
|
| 1024 |
+
"reward": 1.9021683931350708,
|
| 1025 |
+
"reward_std": 0.8603525757789612,
|
| 1026 |
+
"rewards/connected_reward": 1.1977040767669678,
|
| 1027 |
+
"rewards/format_reward": 0.8571429252624512,
|
| 1028 |
+
"rewards/tag_count_reward": 0.9196429252624512,
|
| 1029 |
+
"step": 68
|
| 1030 |
+
},
|
| 1031 |
+
{
|
| 1032 |
+
"clip_ratio": 0.0,
|
| 1033 |
+
"completion_length": 297.6607360839844,
|
| 1034 |
+
"epoch": 0.01104,
|
| 1035 |
+
"grad_norm": 102.03744506835938,
|
| 1036 |
+
"kl": 0.88671875,
|
| 1037 |
+
"learning_rate": 2.813333083910761e-06,
|
| 1038 |
+
"loss": 0.2127,
|
| 1039 |
+
"reward": 2.1446430683135986,
|
| 1040 |
+
"reward_std": 0.40179580450057983,
|
| 1041 |
+
"rewards/connected_reward": 1.3964285850524902,
|
| 1042 |
+
"rewards/format_reward": 0.9285714626312256,
|
| 1043 |
+
"rewards/tag_count_reward": 0.9464285969734192,
|
| 1044 |
+
"step": 69
|
| 1045 |
+
},
|
| 1046 |
+
{
|
| 1047 |
+
"clip_ratio": 0.0,
|
| 1048 |
+
"completion_length": 405.4285888671875,
|
| 1049 |
+
"epoch": 0.0112,
|
| 1050 |
+
"grad_norm": 617.213623046875,
|
| 1051 |
+
"kl": 3.0625,
|
| 1052 |
+
"learning_rate": 2.761321158169134e-06,
|
| 1053 |
+
"loss": 0.5334,
|
| 1054 |
+
"reward": 1.8612104654312134,
|
| 1055 |
+
"reward_std": 0.869892418384552,
|
| 1056 |
+
"rewards/connected_reward": 1.1478174924850464,
|
| 1057 |
+
"rewards/format_reward": 0.8750000596046448,
|
| 1058 |
+
"rewards/tag_count_reward": 0.9196429252624512,
|
| 1059 |
+
"step": 70
|
| 1060 |
+
},
|
| 1061 |
+
{
|
| 1062 |
+
"clip_ratio": 0.0,
|
| 1063 |
+
"completion_length": 276.8571472167969,
|
| 1064 |
+
"epoch": 0.01136,
|
| 1065 |
+
"grad_norm": 107.47020721435547,
|
| 1066 |
+
"kl": 0.97265625,
|
| 1067 |
+
"learning_rate": 2.70919460833079e-06,
|
| 1068 |
+
"loss": 0.1194,
|
| 1069 |
+
"reward": 2.099404811859131,
|
| 1070 |
+
"reward_std": 0.595194935798645,
|
| 1071 |
+
"rewards/connected_reward": 1.3690476417541504,
|
| 1072 |
+
"rewards/format_reward": 0.8928571939468384,
|
| 1073 |
+
"rewards/tag_count_reward": 0.9464285969734192,
|
| 1074 |
+
"step": 71
|
| 1075 |
+
},
|
| 1076 |
+
{
|
| 1077 |
+
"clip_ratio": 0.0,
|
| 1078 |
+
"completion_length": 269.4821472167969,
|
| 1079 |
+
"epoch": 0.01152,
|
| 1080 |
+
"grad_norm": 60.09501266479492,
|
| 1081 |
+
"kl": 0.77734375,
|
| 1082 |
+
"learning_rate": 2.6569762988232838e-06,
|
| 1083 |
+
"loss": 0.044,
|
| 1084 |
+
"reward": 1.8062500953674316,
|
| 1085 |
+
"reward_std": 0.6972793340682983,
|
| 1086 |
+
"rewards/connected_reward": 1.0642857551574707,
|
| 1087 |
+
"rewards/format_reward": 0.910714328289032,
|
| 1088 |
+
"rewards/tag_count_reward": 0.9553571939468384,
|
| 1089 |
+
"step": 72
|
| 1090 |
+
},
|
| 1091 |
+
{
|
| 1092 |
+
"clip_ratio": 0.0,
|
| 1093 |
+
"completion_length": 262.46429443359375,
|
| 1094 |
+
"epoch": 0.01168,
|
| 1095 |
+
"grad_norm": 9.094280242919922,
|
| 1096 |
+
"kl": 0.3046875,
|
| 1097 |
+
"learning_rate": 2.604689134322999e-06,
|
| 1098 |
+
"loss": -0.0869,
|
| 1099 |
+
"reward": 1.8462797403335571,
|
| 1100 |
+
"reward_std": 0.6357587575912476,
|
| 1101 |
+
"rewards/connected_reward": 1.1056549549102783,
|
| 1102 |
+
"rewards/format_reward": 0.910714328289032,
|
| 1103 |
+
"rewards/tag_count_reward": 0.9508929252624512,
|
| 1104 |
+
"step": 73
|
| 1105 |
+
},
|
| 1106 |
+
{
|
| 1107 |
+
"clip_ratio": 0.0,
|
| 1108 |
+
"completion_length": 253.87501525878906,
|
| 1109 |
+
"epoch": 0.01184,
|
| 1110 |
+
"grad_norm": 481.7274475097656,
|
| 1111 |
+
"kl": 0.953125,
|
| 1112 |
+
"learning_rate": 2.5523560497083927e-06,
|
| 1113 |
+
"loss": 0.0369,
|
| 1114 |
+
"reward": 2.012946605682373,
|
| 1115 |
+
"reward_std": 0.9039466381072998,
|
| 1116 |
+
"rewards/connected_reward": 1.2723214626312256,
|
| 1117 |
+
"rewards/format_reward": 0.910714328289032,
|
| 1118 |
+
"rewards/tag_count_reward": 0.9508929252624512,
|
| 1119 |
+
"step": 74
|
| 1120 |
+
},
|
| 1121 |
+
{
|
| 1122 |
+
"clip_ratio": 0.0,
|
| 1123 |
+
"completion_length": 237.75001525878906,
|
| 1124 |
+
"epoch": 0.012,
|
| 1125 |
+
"grad_norm": 213.04286193847656,
|
| 1126 |
+
"kl": 0.390625,
|
| 1127 |
+
"learning_rate": 2.5e-06,
|
| 1128 |
+
"loss": -0.0586,
|
| 1129 |
+
"reward": 2.0683035850524902,
|
| 1130 |
+
"reward_std": 0.8664366006851196,
|
| 1131 |
+
"rewards/connected_reward": 1.3571429252624512,
|
| 1132 |
+
"rewards/format_reward": 0.8571429252624512,
|
| 1133 |
+
"rewards/tag_count_reward": 0.941964328289032,
|
| 1134 |
+
"step": 75
|
| 1135 |
+
},
|
| 1136 |
+
{
|
| 1137 |
+
"clip_ratio": 0.0,
|
| 1138 |
+
"completion_length": 270.3571472167969,
|
| 1139 |
+
"epoch": 0.01216,
|
| 1140 |
+
"grad_norm": 52.30439376831055,
|
| 1141 |
+
"kl": 0.419921875,
|
| 1142 |
+
"learning_rate": 2.447643950291608e-06,
|
| 1143 |
+
"loss": -0.061,
|
| 1144 |
+
"reward": 1.9406251907348633,
|
| 1145 |
+
"reward_std": 0.8179541826248169,
|
| 1146 |
+
"rewards/connected_reward": 1.2116072177886963,
|
| 1147 |
+
"rewards/format_reward": 0.8928571939468384,
|
| 1148 |
+
"rewards/tag_count_reward": 0.941964328289032,
|
| 1149 |
+
"step": 76
|
| 1150 |
+
},
|
| 1151 |
+
{
|
| 1152 |
+
"clip_ratio": 0.0,
|
| 1153 |
+
"completion_length": 258.71429443359375,
|
| 1154 |
+
"epoch": 0.01232,
|
| 1155 |
+
"grad_norm": 346.89080810546875,
|
| 1156 |
+
"kl": 0.39453125,
|
| 1157 |
+
"learning_rate": 2.3953108656770018e-06,
|
| 1158 |
+
"loss": -0.0469,
|
| 1159 |
+
"reward": 2.028571367263794,
|
| 1160 |
+
"reward_std": 0.7005811333656311,
|
| 1161 |
+
"rewards/connected_reward": 1.298214316368103,
|
| 1162 |
+
"rewards/format_reward": 0.8928571939468384,
|
| 1163 |
+
"rewards/tag_count_reward": 0.9464285969734192,
|
| 1164 |
+
"step": 77
|
| 1165 |
+
},
|
| 1166 |
+
{
|
| 1167 |
+
"clip_ratio": 0.0,
|
| 1168 |
+
"completion_length": 303.1964416503906,
|
| 1169 |
+
"epoch": 0.01248,
|
| 1170 |
+
"grad_norm": 469.85699462890625,
|
| 1171 |
+
"kl": 1.015625,
|
| 1172 |
+
"learning_rate": 2.3430237011767166e-06,
|
| 1173 |
+
"loss": 0.149,
|
| 1174 |
+
"reward": 1.832887053489685,
|
| 1175 |
+
"reward_std": 0.8886847496032715,
|
| 1176 |
+
"rewards/connected_reward": 1.1413692235946655,
|
| 1177 |
+
"rewards/format_reward": 0.8392857313156128,
|
| 1178 |
+
"rewards/tag_count_reward": 0.9062500596046448,
|
| 1179 |
+
"step": 78
|
| 1180 |
+
},
|
| 1181 |
+
{
|
| 1182 |
+
"clip_ratio": 0.0,
|
| 1183 |
+
"completion_length": 303.08929443359375,
|
| 1184 |
+
"epoch": 0.01264,
|
| 1185 |
+
"grad_norm": 2587.5205078125,
|
| 1186 |
+
"kl": 2.0625,
|
| 1187 |
+
"learning_rate": 2.290805391669212e-06,
|
| 1188 |
+
"loss": 0.1151,
|
| 1189 |
+
"reward": 1.914987325668335,
|
| 1190 |
+
"reward_std": 0.6788126230239868,
|
| 1191 |
+
"rewards/connected_reward": 1.31632661819458,
|
| 1192 |
+
"rewards/format_reward": 0.6964285969734192,
|
| 1193 |
+
"rewards/tag_count_reward": 0.8348214626312256,
|
| 1194 |
+
"step": 79
|
| 1195 |
+
},
|
| 1196 |
+
{
|
| 1197 |
+
"clip_ratio": 0.0,
|
| 1198 |
+
"completion_length": 290.64288330078125,
|
| 1199 |
+
"epoch": 0.0128,
|
| 1200 |
+
"grad_norm": 576.2128295898438,
|
| 1201 |
+
"kl": 0.44921875,
|
| 1202 |
+
"learning_rate": 2.238678841830867e-06,
|
| 1203 |
+
"loss": 0.0522,
|
| 1204 |
+
"reward": 1.7459821701049805,
|
| 1205 |
+
"reward_std": 0.8939936757087708,
|
| 1206 |
+
"rewards/connected_reward": 1.2142857313156128,
|
| 1207 |
+
"rewards/format_reward": 0.5892857313156128,
|
| 1208 |
+
"rewards/tag_count_reward": 0.7901785969734192,
|
| 1209 |
+
"step": 80
|
| 1210 |
+
},
|
| 1211 |
+
{
|
| 1212 |
+
"clip_ratio": 0.0,
|
| 1213 |
+
"completion_length": 262.1785888671875,
|
| 1214 |
+
"epoch": 0.01296,
|
| 1215 |
+
"grad_norm": 227.37603759765625,
|
| 1216 |
+
"kl": 0.6640625,
|
| 1217 |
+
"learning_rate": 2.186666916089239e-06,
|
| 1218 |
+
"loss": 0.0319,
|
| 1219 |
+
"reward": 1.686011791229248,
|
| 1220 |
+
"reward_std": 0.7109549641609192,
|
| 1221 |
+
"rewards/connected_reward": 1.2485120296478271,
|
| 1222 |
+
"rewards/format_reward": 0.4464285969734192,
|
| 1223 |
+
"rewards/tag_count_reward": 0.7142857313156128,
|
| 1224 |
+
"step": 81
|
| 1225 |
+
},
|
| 1226 |
+
{
|
| 1227 |
+
"clip_ratio": 0.0,
|
| 1228 |
+
"completion_length": 321.1071472167969,
|
| 1229 |
+
"epoch": 0.01312,
|
| 1230 |
+
"grad_norm": 1797.874755859375,
|
| 1231 |
+
"kl": 1.859375,
|
| 1232 |
+
"learning_rate": 2.134792428593971e-06,
|
| 1233 |
+
"loss": 0.2395,
|
| 1234 |
+
"reward": 1.648958444595337,
|
| 1235 |
+
"reward_std": 0.9454951286315918,
|
| 1236 |
+
"rewards/connected_reward": 1.2708333730697632,
|
| 1237 |
+
"rewards/format_reward": 0.3571428656578064,
|
| 1238 |
+
"rewards/tag_count_reward": 0.6651785969734192,
|
| 1239 |
+
"step": 82
|
| 1240 |
+
},
|
| 1241 |
+
{
|
| 1242 |
+
"clip_ratio": 0.0,
|
| 1243 |
+
"completion_length": 318.3035888671875,
|
| 1244 |
+
"epoch": 0.01328,
|
| 1245 |
+
"grad_norm": 911.9273681640625,
|
| 1246 |
+
"kl": 2.53125,
|
| 1247 |
+
"learning_rate": 2.0830781332097446e-06,
|
| 1248 |
+
"loss": 0.1478,
|
| 1249 |
+
"reward": 1.5819941759109497,
|
| 1250 |
+
"reward_std": 0.9505893588066101,
|
| 1251 |
+
"rewards/connected_reward": 1.2172619104385376,
|
| 1252 |
+
"rewards/format_reward": 0.3571428656578064,
|
| 1253 |
+
"rewards/tag_count_reward": 0.6205357313156128,
|
| 1254 |
+
"step": 83
|
| 1255 |
+
},
|
| 1256 |
+
{
|
| 1257 |
+
"clip_ratio": 0.0,
|
| 1258 |
+
"completion_length": 304.875,
|
| 1259 |
+
"epoch": 0.01344,
|
| 1260 |
+
"grad_norm": 2088.1962890625,
|
| 1261 |
+
"kl": 5.3125,
|
| 1262 |
+
"learning_rate": 2.031546713535688e-06,
|
| 1263 |
+
"loss": 0.3074,
|
| 1264 |
+
"reward": 1.2762542963027954,
|
| 1265 |
+
"reward_std": 0.9391399025917053,
|
| 1266 |
+
"rewards/connected_reward": 0.9124150276184082,
|
| 1267 |
+
"rewards/format_reward": 0.3392857313156128,
|
| 1268 |
+
"rewards/tag_count_reward": 0.6473214626312256,
|
| 1269 |
+
"step": 84
|
| 1270 |
+
},
|
| 1271 |
+
{
|
| 1272 |
+
"clip_ratio": 0.0,
|
| 1273 |
+
"completion_length": 316.4464416503906,
|
| 1274 |
+
"epoch": 0.0136,
|
| 1275 |
+
"grad_norm": 549.3711547851562,
|
| 1276 |
+
"kl": 1.71875,
|
| 1277 |
+
"learning_rate": 1.9802207729556023e-06,
|
| 1278 |
+
"loss": 0.172,
|
| 1279 |
+
"reward": 1.3764880895614624,
|
| 1280 |
+
"reward_std": 0.7936123609542847,
|
| 1281 |
+
"rewards/connected_reward": 0.9791667461395264,
|
| 1282 |
+
"rewards/format_reward": 0.392857164144516,
|
| 1283 |
+
"rewards/tag_count_reward": 0.6696428656578064,
|
| 1284 |
+
"step": 85
|
| 1285 |
+
},
|
| 1286 |
+
{
|
| 1287 |
+
"clip_ratio": 0.0,
|
| 1288 |
+
"completion_length": 285.26788330078125,
|
| 1289 |
+
"epoch": 0.01376,
|
| 1290 |
+
"grad_norm": 483.4754943847656,
|
| 1291 |
+
"kl": 0.6953125,
|
| 1292 |
+
"learning_rate": 1.9291228247233607e-06,
|
| 1293 |
+
"loss": 0.0362,
|
| 1294 |
+
"reward": 1.3629463911056519,
|
| 1295 |
+
"reward_std": 0.7181650400161743,
|
| 1296 |
+
"rewards/connected_reward": 1.0598214864730835,
|
| 1297 |
+
"rewards/format_reward": 0.25,
|
| 1298 |
+
"rewards/tag_count_reward": 0.59375,
|
| 1299 |
+
"step": 86
|
| 1300 |
+
},
|
| 1301 |
+
{
|
| 1302 |
+
"clip_ratio": 0.0,
|
| 1303 |
+
"completion_length": 289.4821472167969,
|
| 1304 |
+
"epoch": 0.01392,
|
| 1305 |
+
"grad_norm": 431.3416442871094,
|
| 1306 |
+
"kl": 0.71875,
|
| 1307 |
+
"learning_rate": 1.8782752820878636e-06,
|
| 1308 |
+
"loss": 0.1296,
|
| 1309 |
+
"reward": 1.6370537281036377,
|
| 1310 |
+
"reward_std": 0.7323834896087646,
|
| 1311 |
+
"rewards/connected_reward": 1.3392857313156128,
|
| 1312 |
+
"rewards/format_reward": 0.25,
|
| 1313 |
+
"rewards/tag_count_reward": 0.5758928656578064,
|
| 1314 |
+
"step": 87
|
| 1315 |
+
},
|
| 1316 |
+
{
|
| 1317 |
+
"clip_ratio": 0.0,
|
| 1318 |
+
"completion_length": 272.3035888671875,
|
| 1319 |
+
"epoch": 0.01408,
|
| 1320 |
+
"grad_norm": 178.10662841796875,
|
| 1321 |
+
"kl": 0.88671875,
|
| 1322 |
+
"learning_rate": 1.827700448461836e-06,
|
| 1323 |
+
"loss": 0.0167,
|
| 1324 |
+
"reward": 1.5172619819641113,
|
| 1325 |
+
"reward_std": 0.9202746152877808,
|
| 1326 |
+
"rewards/connected_reward": 1.202529788017273,
|
| 1327 |
+
"rewards/format_reward": 0.267857164144516,
|
| 1328 |
+
"rewards/tag_count_reward": 0.6026785969734192,
|
| 1329 |
+
"step": 88
|
| 1330 |
+
},
|
| 1331 |
+
{
|
| 1332 |
+
"clip_ratio": 0.0,
|
| 1333 |
+
"completion_length": 253.57144165039062,
|
| 1334 |
+
"epoch": 0.01424,
|
| 1335 |
+
"grad_norm": 58.177215576171875,
|
| 1336 |
+
"kl": 2.171875,
|
| 1337 |
+
"learning_rate": 1.7774205076388207e-06,
|
| 1338 |
+
"loss": 0.0,
|
| 1339 |
+
"reward": 0.9139881730079651,
|
| 1340 |
+
"reward_std": 0.7583447098731995,
|
| 1341 |
+
"rewards/connected_reward": 0.77827388048172,
|
| 1342 |
+
"rewards/format_reward": 0.0357142873108387,
|
| 1343 |
+
"rewards/tag_count_reward": 0.392857164144516,
|
| 1344 |
+
"step": 89
|
| 1345 |
+
},
|
| 1346 |
+
{
|
| 1347 |
+
"clip_ratio": 0.0,
|
| 1348 |
+
"completion_length": 276.6071472167969,
|
| 1349 |
+
"epoch": 0.0144,
|
| 1350 |
+
"grad_norm": 205.79637145996094,
|
| 1351 |
+
"kl": 1.6484375,
|
| 1352 |
+
"learning_rate": 1.7274575140626318e-06,
|
| 1353 |
+
"loss": 0.0985,
|
| 1354 |
+
"reward": 1.4693453311920166,
|
| 1355 |
+
"reward_std": 1.1510217189788818,
|
| 1356 |
+
"rewards/connected_reward": 1.0863096714019775,
|
| 1357 |
+
"rewards/format_reward": 0.3750000298023224,
|
| 1358 |
+
"rewards/tag_count_reward": 0.6517857313156128,
|
| 1359 |
+
"step": 90
|
| 1360 |
+
},
|
| 1361 |
+
{
|
| 1362 |
+
"clip_ratio": 0.0,
|
| 1363 |
+
"completion_length": 303.625,
|
| 1364 |
+
"epoch": 0.01456,
|
| 1365 |
+
"grad_norm": 187.00259399414062,
|
| 1366 |
+
"kl": 3.203125,
|
| 1367 |
+
"learning_rate": 1.677833383153542e-06,
|
| 1368 |
+
"loss": 0.1504,
|
| 1369 |
+
"reward": 0.8815476894378662,
|
| 1370 |
+
"reward_std": 0.8259068131446838,
|
| 1371 |
+
"rewards/connected_reward": 0.7217261791229248,
|
| 1372 |
+
"rewards/format_reward": 0.0892857164144516,
|
| 1373 |
+
"rewards/tag_count_reward": 0.3839285969734192,
|
| 1374 |
+
"step": 91
|
| 1375 |
+
},
|
| 1376 |
+
{
|
| 1377 |
+
"clip_ratio": 0.0,
|
| 1378 |
+
"completion_length": 280.625,
|
| 1379 |
+
"epoch": 0.01472,
|
| 1380 |
+
"grad_norm": 47.36444854736328,
|
| 1381 |
+
"kl": 1.5078125,
|
| 1382 |
+
"learning_rate": 1.6285698816954626e-06,
|
| 1383 |
+
"loss": -0.0548,
|
| 1384 |
+
"reward": 1.3680061101913452,
|
| 1385 |
+
"reward_std": 0.9850864410400391,
|
| 1386 |
+
"rewards/connected_reward": 1.0461310148239136,
|
| 1387 |
+
"rewards/format_reward": 0.3035714328289032,
|
| 1388 |
+
"rewards/tag_count_reward": 0.566964328289032,
|
| 1389 |
+
"step": 92
|
| 1390 |
+
},
|
| 1391 |
+
{
|
| 1392 |
+
"clip_ratio": 0.0,
|
| 1393 |
+
"completion_length": 322.5357360839844,
|
| 1394 |
+
"epoch": 0.01488,
|
| 1395 |
+
"grad_norm": 84.1014633178711,
|
| 1396 |
+
"kl": 1.859375,
|
| 1397 |
+
"learning_rate": 1.5796886182883053e-06,
|
| 1398 |
+
"loss": -0.08,
|
| 1399 |
+
"reward": 1.18556547164917,
|
| 1400 |
+
"reward_std": 0.9476414918899536,
|
| 1401 |
+
"rewards/connected_reward": 0.9136905670166016,
|
| 1402 |
+
"rewards/format_reward": 0.2142857313156128,
|
| 1403 |
+
"rewards/tag_count_reward": 0.5491071939468384,
|
| 1404 |
+
"step": 93
|
| 1405 |
+
},
|
| 1406 |
+
{
|
| 1407 |
+
"clip_ratio": 0.0,
|
| 1408 |
+
"completion_length": 296.89288330078125,
|
| 1409 |
+
"epoch": 0.01504,
|
| 1410 |
+
"grad_norm": 2105.967041015625,
|
| 1411 |
+
"kl": 14.9375,
|
| 1412 |
+
"learning_rate": 1.5312110338697427e-06,
|
| 1413 |
+
"loss": 0.4129,
|
| 1414 |
+
"reward": 1.0491071939468384,
|
| 1415 |
+
"reward_std": 0.9280386567115784,
|
| 1416 |
+
"rewards/connected_reward": 0.8214285969734192,
|
| 1417 |
+
"rewards/format_reward": 0.1607142984867096,
|
| 1418 |
+
"rewards/tag_count_reward": 0.4910714626312256,
|
| 1419 |
+
"step": 94
|
| 1420 |
+
},
|
| 1421 |
+
{
|
| 1422 |
+
"clip_ratio": 0.0,
|
| 1423 |
+
"completion_length": 268.8214416503906,
|
| 1424 |
+
"epoch": 0.0152,
|
| 1425 |
+
"grad_norm": 181.77145385742188,
|
| 1426 |
+
"kl": 3.625,
|
| 1427 |
+
"learning_rate": 1.4831583923105e-06,
|
| 1428 |
+
"loss": 0.1499,
|
| 1429 |
+
"reward": 1.3781250715255737,
|
| 1430 |
+
"reward_std": 0.7101474404335022,
|
| 1431 |
+
"rewards/connected_reward": 1.1294643878936768,
|
| 1432 |
+
"rewards/format_reward": 0.1785714328289032,
|
| 1433 |
+
"rewards/tag_count_reward": 0.53125,
|
| 1434 |
+
"step": 95
|
| 1435 |
+
},
|
| 1436 |
+
{
|
| 1437 |
+
"clip_ratio": 0.0,
|
| 1438 |
+
"completion_length": 282.96429443359375,
|
| 1439 |
+
"epoch": 0.01536,
|
| 1440 |
+
"grad_norm": 159.0343017578125,
|
| 1441 |
+
"kl": 2.625,
|
| 1442 |
+
"learning_rate": 1.4355517710873184e-06,
|
| 1443 |
+
"loss": 0.0932,
|
| 1444 |
+
"reward": 1.0690478086471558,
|
| 1445 |
+
"reward_std": 0.6700408458709717,
|
| 1446 |
+
"rewards/connected_reward": 0.8779762387275696,
|
| 1447 |
+
"rewards/format_reward": 0.125,
|
| 1448 |
+
"rewards/tag_count_reward": 0.4285714626312256,
|
| 1449 |
+
"step": 96
|
| 1450 |
+
},
|
| 1451 |
+
{
|
| 1452 |
+
"clip_ratio": 0.0,
|
| 1453 |
+
"completion_length": 273.39288330078125,
|
| 1454 |
+
"epoch": 0.01552,
|
| 1455 |
+
"grad_norm": 4359.7138671875,
|
| 1456 |
+
"kl": 6.34375,
|
| 1457 |
+
"learning_rate": 1.388412052037682e-06,
|
| 1458 |
+
"loss": 0.348,
|
| 1459 |
+
"reward": 1.5040180683135986,
|
| 1460 |
+
"reward_std": 0.8597317337989807,
|
| 1461 |
+
"rewards/connected_reward": 1.3125,
|
| 1462 |
+
"rewards/format_reward": 0.1071428656578064,
|
| 1463 |
+
"rewards/tag_count_reward": 0.4598214626312256,
|
| 1464 |
+
"step": 97
|
| 1465 |
+
},
|
| 1466 |
+
{
|
| 1467 |
+
"clip_ratio": 0.0,
|
| 1468 |
+
"completion_length": 411.46429443359375,
|
| 1469 |
+
"epoch": 0.01568,
|
| 1470 |
+
"grad_norm": 464.1128845214844,
|
| 1471 |
+
"kl": 2.40625,
|
| 1472 |
+
"learning_rate": 1.3417599122003464e-06,
|
| 1473 |
+
"loss": 0.2797,
|
| 1474 |
+
"reward": 1.3879464864730835,
|
| 1475 |
+
"reward_std": 0.7324258089065552,
|
| 1476 |
+
"rewards/connected_reward": 1.1473214626312256,
|
| 1477 |
+
"rewards/format_reward": 0.1785714328289032,
|
| 1478 |
+
"rewards/tag_count_reward": 0.504464328289032,
|
| 1479 |
+
"step": 98
|
| 1480 |
+
},
|
| 1481 |
+
{
|
| 1482 |
+
"clip_ratio": 0.0,
|
| 1483 |
+
"completion_length": 336.4464416503906,
|
| 1484 |
+
"epoch": 0.01584,
|
| 1485 |
+
"grad_norm": 185.98046875,
|
| 1486 |
+
"kl": 1.4921875,
|
| 1487 |
+
"learning_rate": 1.2956158147457116e-06,
|
| 1488 |
+
"loss": 0.2559,
|
| 1489 |
+
"reward": 1.360267996788025,
|
| 1490 |
+
"reward_std": 0.7969329357147217,
|
| 1491 |
+
"rewards/connected_reward": 1.2008929252624512,
|
| 1492 |
+
"rewards/format_reward": 0.0535714328289032,
|
| 1493 |
+
"rewards/tag_count_reward": 0.4419642984867096,
|
| 1494 |
+
"step": 99
|
| 1495 |
+
},
|
| 1496 |
+
{
|
| 1497 |
+
"clip_ratio": 0.0,
|
| 1498 |
+
"completion_length": 463.232177734375,
|
| 1499 |
+
"epoch": 0.016,
|
| 1500 |
+
"grad_norm": 105.13380432128906,
|
| 1501 |
+
"kl": 2.3125,
|
| 1502 |
+
"learning_rate": 1.2500000000000007e-06,
|
| 1503 |
+
"loss": 0.3902,
|
| 1504 |
+
"reward": 1.274553656578064,
|
| 1505 |
+
"reward_std": 0.7874523997306824,
|
| 1506 |
+
"rewards/connected_reward": 1.0973215103149414,
|
| 1507 |
+
"rewards/format_reward": 0.0892857164144516,
|
| 1508 |
+
"rewards/tag_count_reward": 0.4419642984867096,
|
| 1509 |
+
"step": 100
|
| 1510 |
+
},
|
| 1511 |
+
{
|
| 1512 |
+
"clip_ratio": 0.0,
|
| 1513 |
+
"completion_length": 479.0535888671875,
|
| 1514 |
+
"epoch": 0.01616,
|
| 1515 |
+
"grad_norm": 685.9603881835938,
|
| 1516 |
+
"kl": 1.3515625,
|
| 1517 |
+
"learning_rate": 1.204932476567175e-06,
|
| 1518 |
+
"loss": 0.287,
|
| 1519 |
+
"reward": 1.4331845045089722,
|
| 1520 |
+
"reward_std": 0.7297948002815247,
|
| 1521 |
+
"rewards/connected_reward": 1.2380951642990112,
|
| 1522 |
+
"rewards/format_reward": 0.125,
|
| 1523 |
+
"rewards/tag_count_reward": 0.4419642984867096,
|
| 1524 |
+
"step": 101
|
| 1525 |
+
},
|
| 1526 |
+
{
|
| 1527 |
+
"clip_ratio": 0.0,
|
| 1528 |
+
"completion_length": 527.8928833007812,
|
| 1529 |
+
"epoch": 0.01632,
|
| 1530 |
+
"grad_norm": 436.9136657714844,
|
| 1531 |
+
"kl": 1.4609375,
|
| 1532 |
+
"learning_rate": 1.160433012552508e-06,
|
| 1533 |
+
"loss": 0.3846,
|
| 1534 |
+
"reward": 1.3385417461395264,
|
| 1535 |
+
"reward_std": 1.1538331508636475,
|
| 1536 |
+
"rewards/connected_reward": 1.0684523582458496,
|
| 1537 |
+
"rewards/format_reward": 0.2321428656578064,
|
| 1538 |
+
"rewards/tag_count_reward": 0.5133928656578064,
|
| 1539 |
+
"step": 102
|
| 1540 |
+
},
|
| 1541 |
+
{
|
| 1542 |
+
"clip_ratio": 0.0,
|
| 1543 |
+
"completion_length": 310.4821472167969,
|
| 1544 |
+
"epoch": 0.01648,
|
| 1545 |
+
"grad_norm": 5630.6962890625,
|
| 1546 |
+
"kl": 12.8125,
|
| 1547 |
+
"learning_rate": 1.11652112689164e-06,
|
| 1548 |
+
"loss": 1.1389,
|
| 1549 |
+
"reward": 1.5440478324890137,
|
| 1550 |
+
"reward_std": 0.8701499700546265,
|
| 1551 |
+
"rewards/connected_reward": 1.351190447807312,
|
| 1552 |
+
"rewards/format_reward": 0.1071428656578064,
|
| 1553 |
+
"rewards/tag_count_reward": 0.4642857313156128,
|
| 1554 |
+
"step": 103
|
| 1555 |
+
},
|
| 1556 |
+
{
|
| 1557 |
+
"clip_ratio": 0.0,
|
| 1558 |
+
"completion_length": 379.3571472167969,
|
| 1559 |
+
"epoch": 0.01664,
|
| 1560 |
+
"grad_norm": 410.12548828125,
|
| 1561 |
+
"kl": 1.9296875,
|
| 1562 |
+
"learning_rate": 1.073216080788921e-06,
|
| 1563 |
+
"loss": 0.3541,
|
| 1564 |
+
"reward": 1.1357142925262451,
|
| 1565 |
+
"reward_std": 0.719650149345398,
|
| 1566 |
+
"rewards/connected_reward": 1.0,
|
| 1567 |
+
"rewards/format_reward": 0.0357142873108387,
|
| 1568 |
+
"rewards/tag_count_reward": 0.392857164144516,
|
| 1569 |
+
"step": 104
|
| 1570 |
+
},
|
| 1571 |
+
{
|
| 1572 |
+
"clip_ratio": 0.0,
|
| 1573 |
+
"completion_length": 333.51788330078125,
|
| 1574 |
+
"epoch": 0.0168,
|
| 1575 |
+
"grad_norm": 348.0798034667969,
|
| 1576 |
+
"kl": 2.28125,
|
| 1577 |
+
"learning_rate": 1.0305368692688175e-06,
|
| 1578 |
+
"loss": 0.0754,
|
| 1579 |
+
"reward": 1.1610119342803955,
|
| 1580 |
+
"reward_std": 0.7080978751182556,
|
| 1581 |
+
"rewards/connected_reward": 0.8940476179122925,
|
| 1582 |
+
"rewards/format_reward": 0.196428582072258,
|
| 1583 |
+
"rewards/tag_count_reward": 0.5625,
|
| 1584 |
+
"step": 105
|
| 1585 |
+
},
|
| 1586 |
+
{
|
| 1587 |
+
"clip_ratio": 0.0,
|
| 1588 |
+
"completion_length": 329.4464416503906,
|
| 1589 |
+
"epoch": 0.01696,
|
| 1590 |
+
"grad_norm": 1205.54443359375,
|
| 1591 |
+
"kl": 0.62109375,
|
| 1592 |
+
"learning_rate": 9.88502212844063e-07,
|
| 1593 |
+
"loss": 0.0577,
|
| 1594 |
+
"reward": 1.4229167699813843,
|
| 1595 |
+
"reward_std": 0.8394676446914673,
|
| 1596 |
+
"rewards/connected_reward": 1.159523844718933,
|
| 1597 |
+
"rewards/format_reward": 0.1785714328289032,
|
| 1598 |
+
"rewards/tag_count_reward": 0.5803571939468384,
|
| 1599 |
+
"step": 106
|
| 1600 |
+
},
|
| 1601 |
+
{
|
| 1602 |
+
"clip_ratio": 0.0,
|
| 1603 |
+
"completion_length": 376.3750305175781,
|
| 1604 |
+
"epoch": 0.01712,
|
| 1605 |
+
"grad_norm": 774.3217163085938,
|
| 1606 |
+
"kl": 2.46875,
|
| 1607 |
+
"learning_rate": 9.471305493042243e-07,
|
| 1608 |
+
"loss": 0.3784,
|
| 1609 |
+
"reward": 1.0742347240447998,
|
| 1610 |
+
"reward_std": 0.9993730783462524,
|
| 1611 |
+
"rewards/connected_reward": 0.8813775181770325,
|
| 1612 |
+
"rewards/format_reward": 0.1071428656578064,
|
| 1613 |
+
"rewards/tag_count_reward": 0.4642857313156128,
|
| 1614 |
+
"step": 107
|
| 1615 |
+
},
|
| 1616 |
+
{
|
| 1617 |
+
"clip_ratio": 0.0,
|
| 1618 |
+
"completion_length": 333.4821472167969,
|
| 1619 |
+
"epoch": 0.01728,
|
| 1620 |
+
"grad_norm": 157.8820343017578,
|
| 1621 |
+
"kl": 1.671875,
|
| 1622 |
+
"learning_rate": 9.064400256282757e-07,
|
| 1623 |
+
"loss": 0.1336,
|
| 1624 |
+
"reward": 1.3309524059295654,
|
| 1625 |
+
"reward_std": 0.7281343340873718,
|
| 1626 |
+
"rewards/connected_reward": 1.1220239400863647,
|
| 1627 |
+
"rewards/format_reward": 0.1071428656578064,
|
| 1628 |
+
"rewards/tag_count_reward": 0.5178571939468384,
|
| 1629 |
+
"step": 108
|
| 1630 |
+
},
|
| 1631 |
+
{
|
| 1632 |
+
"clip_ratio": 0.0,
|
| 1633 |
+
"completion_length": 389.08929443359375,
|
| 1634 |
+
"epoch": 0.01744,
|
| 1635 |
+
"grad_norm": 95.9675521850586,
|
| 1636 |
+
"kl": 2.6875,
|
| 1637 |
+
"learning_rate": 8.664484900247363e-07,
|
| 1638 |
+
"loss": 0.269,
|
| 1639 |
+
"reward": 1.4202382564544678,
|
| 1640 |
+
"reward_std": 0.5553953051567078,
|
| 1641 |
+
"rewards/connected_reward": 1.1175596714019775,
|
| 1642 |
+
"rewards/format_reward": 0.267857164144516,
|
| 1643 |
+
"rewards/tag_count_reward": 0.5625,
|
| 1644 |
+
"step": 109
|
| 1645 |
+
},
|
| 1646 |
+
{
|
| 1647 |
+
"clip_ratio": 0.0,
|
| 1648 |
+
"completion_length": 341.21429443359375,
|
| 1649 |
+
"epoch": 0.0176,
|
| 1650 |
+
"grad_norm": 478.20684814453125,
|
| 1651 |
+
"kl": 3.71875,
|
| 1652 |
+
"learning_rate": 8.271734841028553e-07,
|
| 1653 |
+
"loss": 0.4583,
|
| 1654 |
+
"reward": 1.6011906862258911,
|
| 1655 |
+
"reward_std": 0.7328930497169495,
|
| 1656 |
+
"rewards/connected_reward": 1.3565478324890137,
|
| 1657 |
+
"rewards/format_reward": 0.1785714328289032,
|
| 1658 |
+
"rewards/tag_count_reward": 0.5178571939468384,
|
| 1659 |
+
"step": 110
|
| 1660 |
+
},
|
| 1661 |
+
{
|
| 1662 |
+
"clip_ratio": 0.0,
|
| 1663 |
+
"completion_length": 362.71429443359375,
|
| 1664 |
+
"epoch": 0.01776,
|
| 1665 |
+
"grad_norm": 570.6087036132812,
|
| 1666 |
+
"kl": 1.578125,
|
| 1667 |
+
"learning_rate": 7.886322351782782e-07,
|
| 1668 |
+
"loss": 0.2897,
|
| 1669 |
+
"reward": 1.381250023841858,
|
| 1670 |
+
"reward_std": 0.9994818568229675,
|
| 1671 |
+
"rewards/connected_reward": 1.1053571701049805,
|
| 1672 |
+
"rewards/format_reward": 0.2142857313156128,
|
| 1673 |
+
"rewards/tag_count_reward": 0.5625,
|
| 1674 |
+
"step": 111
|
| 1675 |
+
},
|
| 1676 |
+
{
|
| 1677 |
+
"clip_ratio": 0.0,
|
| 1678 |
+
"completion_length": 411.64288330078125,
|
| 1679 |
+
"epoch": 0.01792,
|
| 1680 |
+
"grad_norm": 227.14622497558594,
|
| 1681 |
+
"kl": 0.8984375,
|
| 1682 |
+
"learning_rate": 7.508416487165862e-07,
|
| 1683 |
+
"loss": 0.2723,
|
| 1684 |
+
"reward": 1.3443453311920166,
|
| 1685 |
+
"reward_std": 0.8701754808425903,
|
| 1686 |
+
"rewards/connected_reward": 1.108630895614624,
|
| 1687 |
+
"rewards/format_reward": 0.1607142984867096,
|
| 1688 |
+
"rewards/tag_count_reward": 0.5178571939468384,
|
| 1689 |
+
"step": 112
|
| 1690 |
+
},
|
| 1691 |
+
{
|
| 1692 |
+
"clip_ratio": 0.0,
|
| 1693 |
+
"completion_length": 404.9464416503906,
|
| 1694 |
+
"epoch": 0.01808,
|
| 1695 |
+
"grad_norm": 133.9763641357422,
|
| 1696 |
+
"kl": 0.7109375,
|
| 1697 |
+
"learning_rate": 7.138183009179922e-07,
|
| 1698 |
+
"loss": 0.1851,
|
| 1699 |
+
"reward": 1.5876487493515015,
|
| 1700 |
+
"reward_std": 0.7359150648117065,
|
| 1701 |
+
"rewards/connected_reward": 1.194345235824585,
|
| 1702 |
+
"rewards/format_reward": 0.392857164144516,
|
| 1703 |
+
"rewards/tag_count_reward": 0.65625,
|
| 1704 |
+
"step": 113
|
| 1705 |
+
},
|
| 1706 |
+
{
|
| 1707 |
+
"clip_ratio": 0.0,
|
| 1708 |
+
"completion_length": 373.89288330078125,
|
| 1709 |
+
"epoch": 0.01824,
|
| 1710 |
+
"grad_norm": 73.8996810913086,
|
| 1711 |
+
"kl": 1.4765625,
|
| 1712 |
+
"learning_rate": 6.775784314464717e-07,
|
| 1713 |
+
"loss": 0.3297,
|
| 1714 |
+
"reward": 1.5834821462631226,
|
| 1715 |
+
"reward_std": 0.7138813138008118,
|
| 1716 |
+
"rewards/connected_reward": 1.3071428537368774,
|
| 1717 |
+
"rewards/format_reward": 0.196428582072258,
|
| 1718 |
+
"rewards/tag_count_reward": 0.59375,
|
| 1719 |
+
"step": 114
|
| 1720 |
+
},
|
| 1721 |
+
{
|
| 1722 |
+
"clip_ratio": 0.0,
|
| 1723 |
+
"completion_length": 355.51788330078125,
|
| 1724 |
+
"epoch": 0.0184,
|
| 1725 |
+
"grad_norm": 283.4354553222656,
|
| 1726 |
+
"kl": 1.5078125,
|
| 1727 |
+
"learning_rate": 6.421379363065142e-07,
|
| 1728 |
+
"loss": 0.3945,
|
| 1729 |
+
"reward": 1.6325894594192505,
|
| 1730 |
+
"reward_std": 0.7952902913093567,
|
| 1731 |
+
"rewards/connected_reward": 1.2625001668930054,
|
| 1732 |
+
"rewards/format_reward": 0.3571428656578064,
|
| 1733 |
+
"rewards/tag_count_reward": 0.6383928656578064,
|
| 1734 |
+
"step": 115
|
| 1735 |
+
},
|
| 1736 |
+
{
|
| 1737 |
+
"clip_ratio": 0.0,
|
| 1738 |
+
"completion_length": 580.7857666015625,
|
| 1739 |
+
"epoch": 0.01856,
|
| 1740 |
+
"grad_norm": 158.34906005859375,
|
| 1741 |
+
"kl": 2.71875,
|
| 1742 |
+
"learning_rate": 6.075123608706093e-07,
|
| 1743 |
+
"loss": 0.432,
|
| 1744 |
+
"reward": 1.3404762744903564,
|
| 1745 |
+
"reward_std": 0.909478485584259,
|
| 1746 |
+
"rewards/connected_reward": 1.0833333730697632,
|
| 1747 |
+
"rewards/format_reward": 0.2142857313156128,
|
| 1748 |
+
"rewards/tag_count_reward": 0.5,
|
| 1749 |
+
"step": 116
|
| 1750 |
+
},
|
| 1751 |
+
{
|
| 1752 |
+
"clip_ratio": 0.0,
|
| 1753 |
+
"completion_length": 327.875,
|
| 1754 |
+
"epoch": 0.01872,
|
| 1755 |
+
"grad_norm": 87.71063995361328,
|
| 1756 |
+
"kl": 0.51953125,
|
| 1757 |
+
"learning_rate": 5.737168930605272e-07,
|
| 1758 |
+
"loss": 0.0656,
|
| 1759 |
+
"reward": 1.6141369342803955,
|
| 1760 |
+
"reward_std": 0.8700290322303772,
|
| 1761 |
+
"rewards/connected_reward": 1.1404762268066406,
|
| 1762 |
+
"rewards/format_reward": 0.5,
|
| 1763 |
+
"rewards/tag_count_reward": 0.7455357313156128,
|
| 1764 |
+
"step": 117
|
| 1765 |
+
},
|
| 1766 |
+
{
|
| 1767 |
+
"clip_ratio": 0.0,
|
| 1768 |
+
"completion_length": 376.0000305175781,
|
| 1769 |
+
"epoch": 0.01888,
|
| 1770 |
+
"grad_norm": 83.34387969970703,
|
| 1771 |
+
"kl": 2.625,
|
| 1772 |
+
"learning_rate": 5.407663566854008e-07,
|
| 1773 |
+
"loss": 0.4935,
|
| 1774 |
+
"reward": 1.892113208770752,
|
| 1775 |
+
"reward_std": 0.8559380769729614,
|
| 1776 |
+
"rewards/connected_reward": 1.3943451642990112,
|
| 1777 |
+
"rewards/format_reward": 0.5535714626312256,
|
| 1778 |
+
"rewards/tag_count_reward": 0.7366071939468384,
|
| 1779 |
+
"step": 118
|
| 1780 |
+
},
|
| 1781 |
+
{
|
| 1782 |
+
"clip_ratio": 0.0,
|
| 1783 |
+
"completion_length": 435.9285888671875,
|
| 1784 |
+
"epoch": 0.01904,
|
| 1785 |
+
"grad_norm": 2181.889892578125,
|
| 1786 |
+
"kl": 6.625,
|
| 1787 |
+
"learning_rate": 5.086752049395094e-07,
|
| 1788 |
+
"loss": 1.0453,
|
| 1789 |
+
"reward": 1.2653273344039917,
|
| 1790 |
+
"reward_std": 0.9918228387832642,
|
| 1791 |
+
"rewards/connected_reward": 0.8970238566398621,
|
| 1792 |
+
"rewards/format_reward": 0.3750000298023224,
|
| 1793 |
+
"rewards/tag_count_reward": 0.6026785969734192,
|
| 1794 |
+
"step": 119
|
| 1795 |
+
},
|
| 1796 |
+
{
|
| 1797 |
+
"clip_ratio": 0.0,
|
| 1798 |
+
"completion_length": 430.7500305175781,
|
| 1799 |
+
"epoch": 0.0192,
|
| 1800 |
+
"grad_norm": 7002.36767578125,
|
| 1801 |
+
"kl": 7.375,
|
| 1802 |
+
"learning_rate": 4.774575140626317e-07,
|
| 1803 |
+
"loss": 0.9658,
|
| 1804 |
+
"reward": 1.498660683631897,
|
| 1805 |
+
"reward_std": 0.9669275879859924,
|
| 1806 |
+
"rewards/connected_reward": 1.0910714864730835,
|
| 1807 |
+
"rewards/format_reward": 0.4107142984867096,
|
| 1808 |
+
"rewards/tag_count_reward": 0.6741071939468384,
|
| 1809 |
+
"step": 120
|
| 1810 |
+
},
|
| 1811 |
+
{
|
| 1812 |
+
"clip_ratio": 0.0,
|
| 1813 |
+
"completion_length": 466.33929443359375,
|
| 1814 |
+
"epoch": 0.01936,
|
| 1815 |
+
"grad_norm": 2396.701416015625,
|
| 1816 |
+
"kl": 6.78125,
|
| 1817 |
+
"learning_rate": 4.4712697716573994e-07,
|
| 1818 |
+
"loss": 0.6327,
|
| 1819 |
+
"reward": 1.1019346714019775,
|
| 1820 |
+
"reward_std": 0.9098219871520996,
|
| 1821 |
+
"rewards/connected_reward": 0.7291667461395264,
|
| 1822 |
+
"rewards/format_reward": 0.3571428656578064,
|
| 1823 |
+
"rewards/tag_count_reward": 0.6473214626312256,
|
| 1824 |
+
"step": 121
|
| 1825 |
+
},
|
| 1826 |
+
{
|
| 1827 |
+
"clip_ratio": 0.0,
|
| 1828 |
+
"completion_length": 441.9464416503906,
|
| 1829 |
+
"epoch": 0.01952,
|
| 1830 |
+
"grad_norm": 18626.265625,
|
| 1831 |
+
"kl": 37.75,
|
| 1832 |
+
"learning_rate": 4.1769689822475147e-07,
|
| 1833 |
+
"loss": 3.2228,
|
| 1834 |
+
"reward": 1.485119104385376,
|
| 1835 |
+
"reward_std": 1.0364267826080322,
|
| 1836 |
+
"rewards/connected_reward": 1.0386905670166016,
|
| 1837 |
+
"rewards/format_reward": 0.5178571939468384,
|
| 1838 |
+
"rewards/tag_count_reward": 0.625,
|
| 1839 |
+
"step": 122
|
| 1840 |
+
},
|
| 1841 |
+
{
|
| 1842 |
+
"clip_ratio": 0.0,
|
| 1843 |
+
"completion_length": 389.2321472167969,
|
| 1844 |
+
"epoch": 0.01968,
|
| 1845 |
+
"grad_norm": 198.896728515625,
|
| 1846 |
+
"kl": 2.34375,
|
| 1847 |
+
"learning_rate": 3.891801862449629e-07,
|
| 1848 |
+
"loss": 0.3945,
|
| 1849 |
+
"reward": 1.3038691282272339,
|
| 1850 |
+
"reward_std": 1.0375421047210693,
|
| 1851 |
+
"rewards/connected_reward": 0.8851190805435181,
|
| 1852 |
+
"rewards/format_reward": 0.4464285969734192,
|
| 1853 |
+
"rewards/tag_count_reward": 0.6517857313156128,
|
| 1854 |
+
"step": 123
|
| 1855 |
+
},
|
| 1856 |
+
{
|
| 1857 |
+
"clip_ratio": 0.0,
|
| 1858 |
+
"completion_length": 429.5714416503906,
|
| 1859 |
+
"epoch": 0.01984,
|
| 1860 |
+
"grad_norm": 180.44760131835938,
|
| 1861 |
+
"kl": 2.265625,
|
| 1862 |
+
"learning_rate": 3.615893495987335e-07,
|
| 1863 |
+
"loss": 0.4959,
|
| 1864 |
+
"reward": 1.2294642925262451,
|
| 1865 |
+
"reward_std": 0.82194584608078,
|
| 1866 |
+
"rewards/connected_reward": 0.7562500238418579,
|
| 1867 |
+
"rewards/format_reward": 0.5178571939468384,
|
| 1868 |
+
"rewards/tag_count_reward": 0.7142857313156128,
|
| 1869 |
+
"step": 124
|
| 1870 |
+
},
|
| 1871 |
+
{
|
| 1872 |
+
"clip_ratio": 0.0,
|
| 1873 |
+
"completion_length": 418.4821472167969,
|
| 1874 |
+
"epoch": 0.02,
|
| 1875 |
+
"grad_norm": 222.1766815185547,
|
| 1876 |
+
"kl": 1.484375,
|
| 1877 |
+
"learning_rate": 3.3493649053890325e-07,
|
| 1878 |
+
"loss": 0.4571,
|
| 1879 |
+
"reward": 1.794196605682373,
|
| 1880 |
+
"reward_std": 0.9758167862892151,
|
| 1881 |
+
"rewards/connected_reward": 1.227678656578064,
|
| 1882 |
+
"rewards/format_reward": 0.6428571939468384,
|
| 1883 |
+
"rewards/tag_count_reward": 0.816964328289032,
|
| 1884 |
+
"step": 125
|
| 1885 |
+
},
|
| 1886 |
+
{
|
| 1887 |
+
"clip_ratio": 0.0,
|
| 1888 |
+
"completion_length": 437.33929443359375,
|
| 1889 |
+
"epoch": 0.02016,
|
| 1890 |
+
"grad_norm": 233.66514587402344,
|
| 1891 |
+
"kl": 1.8828125,
|
| 1892 |
+
"learning_rate": 3.092332998903416e-07,
|
| 1893 |
+
"loss": 0.4417,
|
| 1894 |
+
"reward": 1.347916841506958,
|
| 1895 |
+
"reward_std": 0.7624654173851013,
|
| 1896 |
+
"rewards/connected_reward": 0.8398810625076294,
|
| 1897 |
+
"rewards/format_reward": 0.5714285969734192,
|
| 1898 |
+
"rewards/tag_count_reward": 0.7410714626312256,
|
| 1899 |
+
"step": 126
|
| 1900 |
+
},
|
| 1901 |
+
{
|
| 1902 |
+
"clip_ratio": 0.0,
|
| 1903 |
+
"completion_length": 311.4285888671875,
|
| 1904 |
+
"epoch": 0.02032,
|
| 1905 |
+
"grad_norm": 183.92347717285156,
|
| 1906 |
+
"kl": 2.0625,
|
| 1907 |
+
"learning_rate": 2.844910519219632e-07,
|
| 1908 |
+
"loss": 0.347,
|
| 1909 |
+
"reward": 1.6866072416305542,
|
| 1910 |
+
"reward_std": 1.0179650783538818,
|
| 1911 |
+
"rewards/connected_reward": 1.1875,
|
| 1912 |
+
"rewards/format_reward": 0.5535714626312256,
|
| 1913 |
+
"rewards/tag_count_reward": 0.7410714626312256,
|
| 1914 |
+
"step": 127
|
| 1915 |
+
},
|
| 1916 |
+
{
|
| 1917 |
+
"clip_ratio": 0.0,
|
| 1918 |
+
"completion_length": 340.01788330078125,
|
| 1919 |
+
"epoch": 0.02048,
|
| 1920 |
+
"grad_norm": 557.9811401367188,
|
| 1921 |
+
"kl": 8.25,
|
| 1922 |
+
"learning_rate": 2.6072059940146775e-07,
|
| 1923 |
+
"loss": 0.497,
|
| 1924 |
+
"reward": 1.3771826028823853,
|
| 1925 |
+
"reward_std": 1.0474642515182495,
|
| 1926 |
+
"rewards/connected_reward": 0.8343254327774048,
|
| 1927 |
+
"rewards/format_reward": 0.625,
|
| 1928 |
+
"rewards/tag_count_reward": 0.7678571939468384,
|
| 1929 |
+
"step": 128
|
| 1930 |
+
},
|
| 1931 |
+
{
|
| 1932 |
+
"clip_ratio": 0.0,
|
| 1933 |
+
"completion_length": 359.33929443359375,
|
| 1934 |
+
"epoch": 0.02064,
|
| 1935 |
+
"grad_norm": 96.67167663574219,
|
| 1936 |
+
"kl": 3.03125,
|
| 1937 |
+
"learning_rate": 2.3793236883495164e-07,
|
| 1938 |
+
"loss": 0.378,
|
| 1939 |
+
"reward": 1.5348213911056519,
|
| 1940 |
+
"reward_std": 0.820845365524292,
|
| 1941 |
+
"rewards/connected_reward": 1.0187500715255737,
|
| 1942 |
+
"rewards/format_reward": 0.5714285969734192,
|
| 1943 |
+
"rewards/tag_count_reward": 0.7678571939468384,
|
| 1944 |
+
"step": 129
|
| 1945 |
+
},
|
| 1946 |
+
{
|
| 1947 |
+
"clip_ratio": 0.0,
|
| 1948 |
+
"completion_length": 404.6785888671875,
|
| 1949 |
+
"epoch": 0.0208,
|
| 1950 |
+
"grad_norm": 1207.408935546875,
|
| 1951 |
+
"kl": 15.1875,
|
| 1952 |
+
"learning_rate": 2.1613635589349756e-07,
|
| 1953 |
+
"loss": 1.1967,
|
| 1954 |
+
"reward": 1.7483631372451782,
|
| 1955 |
+
"reward_std": 0.8727819323539734,
|
| 1956 |
+
"rewards/connected_reward": 1.251488208770752,
|
| 1957 |
+
"rewards/format_reward": 0.535714328289032,
|
| 1958 |
+
"rewards/tag_count_reward": 0.7633928656578064,
|
| 1959 |
+
"step": 130
|
| 1960 |
+
},
|
| 1961 |
+
{
|
| 1962 |
+
"clip_ratio": 0.0,
|
| 1963 |
+
"completion_length": 550.1785888671875,
|
| 1964 |
+
"epoch": 0.02096,
|
| 1965 |
+
"grad_norm": 146.79945373535156,
|
| 1966 |
+
"kl": 3.203125,
|
| 1967 |
+
"learning_rate": 1.95342121028749e-07,
|
| 1968 |
+
"loss": 0.4891,
|
| 1969 |
+
"reward": 1.643898844718933,
|
| 1970 |
+
"reward_std": 1.03229558467865,
|
| 1971 |
+
"rewards/connected_reward": 1.2470238208770752,
|
| 1972 |
+
"rewards/format_reward": 0.4107142984867096,
|
| 1973 |
+
"rewards/tag_count_reward": 0.6383928656578064,
|
| 1974 |
+
"step": 131
|
| 1975 |
+
},
|
| 1976 |
+
{
|
| 1977 |
+
"clip_ratio": 0.0,
|
| 1978 |
+
"completion_length": 448.8214416503906,
|
| 1979 |
+
"epoch": 0.02112,
|
| 1980 |
+
"grad_norm": 201.84288024902344,
|
| 1981 |
+
"kl": 2.734375,
|
| 1982 |
+
"learning_rate": 1.7555878527937164e-07,
|
| 1983 |
+
"loss": 0.437,
|
| 1984 |
+
"reward": 1.5223214626312256,
|
| 1985 |
+
"reward_std": 0.8164080381393433,
|
| 1986 |
+
"rewards/connected_reward": 1.0410715341567993,
|
| 1987 |
+
"rewards/format_reward": 0.5178571939468384,
|
| 1988 |
+
"rewards/tag_count_reward": 0.7410714626312256,
|
| 1989 |
+
"step": 132
|
| 1990 |
+
},
|
| 1991 |
+
{
|
| 1992 |
+
"clip_ratio": 0.0,
|
| 1993 |
+
"completion_length": 445.96429443359375,
|
| 1994 |
+
"epoch": 0.02128,
|
| 1995 |
+
"grad_norm": 271.8717956542969,
|
| 1996 |
+
"kl": 2.765625,
|
| 1997 |
+
"learning_rate": 1.567950262702714e-07,
|
| 1998 |
+
"loss": 0.7114,
|
| 1999 |
+
"reward": 1.504166841506958,
|
| 2000 |
+
"reward_std": 0.8897882103919983,
|
| 2001 |
+
"rewards/connected_reward": 1.0604166984558105,
|
| 2002 |
+
"rewards/format_reward": 0.4642857313156128,
|
| 2003 |
+
"rewards/tag_count_reward": 0.7053571939468384,
|
| 2004 |
+
"step": 133
|
| 2005 |
+
},
|
| 2006 |
+
{
|
| 2007 |
+
"clip_ratio": 0.0,
|
| 2008 |
+
"completion_length": 549.607177734375,
|
| 2009 |
+
"epoch": 0.02144,
|
| 2010 |
+
"grad_norm": 262.3206787109375,
|
| 2011 |
+
"kl": 3.6875,
|
| 2012 |
+
"learning_rate": 1.3905907440629752e-07,
|
| 2013 |
+
"loss": 0.6148,
|
| 2014 |
+
"reward": 1.5785714387893677,
|
| 2015 |
+
"reward_std": 0.9023001194000244,
|
| 2016 |
+
"rewards/connected_reward": 1.133928656578064,
|
| 2017 |
+
"rewards/format_reward": 0.4821428656578064,
|
| 2018 |
+
"rewards/tag_count_reward": 0.6785714626312256,
|
| 2019 |
+
"step": 134
|
| 2020 |
+
},
|
| 2021 |
+
{
|
| 2022 |
+
"clip_ratio": 0.0,
|
| 2023 |
+
"completion_length": 370.26788330078125,
|
| 2024 |
+
"epoch": 0.0216,
|
| 2025 |
+
"grad_norm": 199.73208618164062,
|
| 2026 |
+
"kl": 3.75,
|
| 2027 |
+
"learning_rate": 1.223587092621162e-07,
|
| 2028 |
+
"loss": 0.5675,
|
| 2029 |
+
"reward": 1.5458333492279053,
|
| 2030 |
+
"reward_std": 0.9446250796318054,
|
| 2031 |
+
"rewards/connected_reward": 1.1110119819641113,
|
| 2032 |
+
"rewards/format_reward": 0.4464285969734192,
|
| 2033 |
+
"rewards/tag_count_reward": 0.7053571939468384,
|
| 2034 |
+
"step": 135
|
| 2035 |
+
},
|
| 2036 |
+
{
|
| 2037 |
+
"clip_ratio": 0.0,
|
| 2038 |
+
"completion_length": 455.6785888671875,
|
| 2039 |
+
"epoch": 0.02176,
|
| 2040 |
+
"grad_norm": 119.2493667602539,
|
| 2041 |
+
"kl": 1.4765625,
|
| 2042 |
+
"learning_rate": 1.067012561698319e-07,
|
| 2043 |
+
"loss": 0.4643,
|
| 2044 |
+
"reward": 1.4904762506484985,
|
| 2045 |
+
"reward_std": 0.9661902189254761,
|
| 2046 |
+
"rewards/connected_reward": 0.985119104385376,
|
| 2047 |
+
"rewards/format_reward": 0.5714285969734192,
|
| 2048 |
+
"rewards/tag_count_reward": 0.7321428656578064,
|
| 2049 |
+
"step": 136
|
| 2050 |
+
},
|
| 2051 |
+
{
|
| 2052 |
+
"clip_ratio": 0.0,
|
| 2053 |
+
"completion_length": 413.1607360839844,
|
| 2054 |
+
"epoch": 0.02192,
|
| 2055 |
+
"grad_norm": 189.31283569335938,
|
| 2056 |
+
"kl": 1.5390625,
|
| 2057 |
+
"learning_rate": 9.209358300585474e-08,
|
| 2058 |
+
"loss": 0.5992,
|
| 2059 |
+
"reward": 1.70119047164917,
|
| 2060 |
+
"reward_std": 1.0350253582000732,
|
| 2061 |
+
"rewards/connected_reward": 1.2422618865966797,
|
| 2062 |
+
"rewards/format_reward": 0.5,
|
| 2063 |
+
"rewards/tag_count_reward": 0.6964285969734192,
|
| 2064 |
+
"step": 137
|
| 2065 |
+
},
|
| 2066 |
+
{
|
| 2067 |
+
"clip_ratio": 0.0,
|
| 2068 |
+
"completion_length": 396.64288330078125,
|
| 2069 |
+
"epoch": 0.02208,
|
| 2070 |
+
"grad_norm": 312.1037292480469,
|
| 2071 |
+
"kl": 3.28125,
|
| 2072 |
+
"learning_rate": 7.854209717842231e-08,
|
| 2073 |
+
"loss": 0.2851,
|
| 2074 |
+
"reward": 1.2653274536132812,
|
| 2075 |
+
"reward_std": 0.9041686654090881,
|
| 2076 |
+
"rewards/connected_reward": 0.9273810386657715,
|
| 2077 |
+
"rewards/format_reward": 0.3035714328289032,
|
| 2078 |
+
"rewards/tag_count_reward": 0.6205357313156128,
|
| 2079 |
+
"step": 138
|
| 2080 |
+
},
|
| 2081 |
+
{
|
| 2082 |
+
"clip_ratio": 0.0,
|
| 2083 |
+
"completion_length": 428.6607360839844,
|
| 2084 |
+
"epoch": 0.02224,
|
| 2085 |
+
"grad_norm": 96.09481048583984,
|
| 2086 |
+
"kl": 3.234375,
|
| 2087 |
+
"learning_rate": 6.605274281709929e-08,
|
| 2088 |
+
"loss": 0.6046,
|
| 2089 |
+
"reward": 1.3367561101913452,
|
| 2090 |
+
"reward_std": 0.9484272599220276,
|
| 2091 |
+
"rewards/connected_reward": 1.021130919456482,
|
| 2092 |
+
"rewards/format_reward": 0.2857142984867096,
|
| 2093 |
+
"rewards/tag_count_reward": 0.5758928656578064,
|
| 2094 |
+
"step": 139
|
| 2095 |
+
},
|
| 2096 |
+
{
|
| 2097 |
+
"clip_ratio": 0.0,
|
| 2098 |
+
"completion_length": 427.8035888671875,
|
| 2099 |
+
"epoch": 0.0224,
|
| 2100 |
+
"grad_norm": 112.52371978759766,
|
| 2101 |
+
"kl": 2.671875,
|
| 2102 |
+
"learning_rate": 5.463099816548578e-08,
|
| 2103 |
+
"loss": 0.4703,
|
| 2104 |
+
"reward": 1.6272321939468384,
|
| 2105 |
+
"reward_std": 0.9923813939094543,
|
| 2106 |
+
"rewards/connected_reward": 1.2071430683135986,
|
| 2107 |
+
"rewards/format_reward": 0.4464285969734192,
|
| 2108 |
+
"rewards/tag_count_reward": 0.65625,
|
| 2109 |
+
"step": 140
|
| 2110 |
+
},
|
| 2111 |
+
{
|
| 2112 |
+
"clip_ratio": 0.0,
|
| 2113 |
+
"completion_length": 442.1607360839844,
|
| 2114 |
+
"epoch": 0.02256,
|
| 2115 |
+
"grad_norm": 116.0601577758789,
|
| 2116 |
+
"kl": 3.25,
|
| 2117 |
+
"learning_rate": 4.428187317827848e-08,
|
| 2118 |
+
"loss": 0.5996,
|
| 2119 |
+
"reward": 1.4907739162445068,
|
| 2120 |
+
"reward_std": 0.8477200269699097,
|
| 2121 |
+
"rewards/connected_reward": 1.0898810625076294,
|
| 2122 |
+
"rewards/format_reward": 0.4107142984867096,
|
| 2123 |
+
"rewards/tag_count_reward": 0.6517857313156128,
|
| 2124 |
+
"step": 141
|
| 2125 |
+
},
|
| 2126 |
+
{
|
| 2127 |
+
"clip_ratio": 0.0,
|
| 2128 |
+
"completion_length": 394.21429443359375,
|
| 2129 |
+
"epoch": 0.02272,
|
| 2130 |
+
"grad_norm": 1539.869384765625,
|
| 2131 |
+
"kl": 6.03125,
|
| 2132 |
+
"learning_rate": 3.5009907323737826e-08,
|
| 2133 |
+
"loss": 1.3606,
|
| 2134 |
+
"reward": 1.6227679252624512,
|
| 2135 |
+
"reward_std": 1.0998399257659912,
|
| 2136 |
+
"rewards/connected_reward": 1.0982143878936768,
|
| 2137 |
+
"rewards/format_reward": 0.6071428656578064,
|
| 2138 |
+
"rewards/tag_count_reward": 0.7366071939468384,
|
| 2139 |
+
"step": 142
|
| 2140 |
+
},
|
| 2141 |
+
{
|
| 2142 |
+
"clip_ratio": 0.0,
|
| 2143 |
+
"completion_length": 361.9285888671875,
|
| 2144 |
+
"epoch": 0.02288,
|
| 2145 |
+
"grad_norm": 60.153076171875,
|
| 2146 |
+
"kl": 2.84375,
|
| 2147 |
+
"learning_rate": 2.681916759252917e-08,
|
| 2148 |
+
"loss": 0.4273,
|
| 2149 |
+
"reward": 1.530272126197815,
|
| 2150 |
+
"reward_std": 1.0186634063720703,
|
| 2151 |
+
"rewards/connected_reward": 1.1356292963027954,
|
| 2152 |
+
"rewards/format_reward": 0.392857164144516,
|
| 2153 |
+
"rewards/tag_count_reward": 0.660714328289032,
|
| 2154 |
+
"step": 143
|
| 2155 |
+
},
|
| 2156 |
+
{
|
| 2157 |
+
"clip_ratio": 0.0,
|
| 2158 |
+
"completion_length": 420.0000305175781,
|
| 2159 |
+
"epoch": 0.02304,
|
| 2160 |
+
"grad_norm": 57.555294036865234,
|
| 2161 |
+
"kl": 2.4375,
|
| 2162 |
+
"learning_rate": 1.9713246713805588e-08,
|
| 2163 |
+
"loss": 0.4188,
|
| 2164 |
+
"reward": 1.5285714864730835,
|
| 2165 |
+
"reward_std": 1.0183099508285522,
|
| 2166 |
+
"rewards/connected_reward": 1.0892857313156128,
|
| 2167 |
+
"rewards/format_reward": 0.4821428656578064,
|
| 2168 |
+
"rewards/tag_count_reward": 0.660714328289032,
|
| 2169 |
+
"step": 144
|
| 2170 |
+
},
|
| 2171 |
+
{
|
| 2172 |
+
"clip_ratio": 0.0,
|
| 2173 |
+
"completion_length": 327.7321472167969,
|
| 2174 |
+
"epoch": 0.0232,
|
| 2175 |
+
"grad_norm": 92.85120391845703,
|
| 2176 |
+
"kl": 1.3828125,
|
| 2177 |
+
"learning_rate": 1.3695261579316776e-08,
|
| 2178 |
+
"loss": 0.2957,
|
| 2179 |
+
"reward": 1.6851191520690918,
|
| 2180 |
+
"reward_std": 0.9654322266578674,
|
| 2181 |
+
"rewards/connected_reward": 1.2351192235946655,
|
| 2182 |
+
"rewards/format_reward": 0.4821428656578064,
|
| 2183 |
+
"rewards/tag_count_reward": 0.6964285969734192,
|
| 2184 |
+
"step": 145
|
| 2185 |
+
},
|
| 2186 |
+
{
|
| 2187 |
+
"clip_ratio": 0.0,
|
| 2188 |
+
"completion_length": 374.5000305175781,
|
| 2189 |
+
"epoch": 0.02336,
|
| 2190 |
+
"grad_norm": 448.3329772949219,
|
| 2191 |
+
"kl": 3.171875,
|
| 2192 |
+
"learning_rate": 8.767851876239075e-09,
|
| 2193 |
+
"loss": 0.5714,
|
| 2194 |
+
"reward": 1.5226191282272339,
|
| 2195 |
+
"reward_std": 0.7929543852806091,
|
| 2196 |
+
"rewards/connected_reward": 1.0904762744903564,
|
| 2197 |
+
"rewards/format_reward": 0.4464285969734192,
|
| 2198 |
+
"rewards/tag_count_reward": 0.6964285969734192,
|
| 2199 |
+
"step": 146
|
| 2200 |
+
},
|
| 2201 |
+
{
|
| 2202 |
+
"clip_ratio": 0.0,
|
| 2203 |
+
"completion_length": 430.26788330078125,
|
| 2204 |
+
"epoch": 0.02352,
|
| 2205 |
+
"grad_norm": 171.38597106933594,
|
| 2206 |
+
"kl": 2.390625,
|
| 2207 |
+
"learning_rate": 4.933178929321103e-09,
|
| 2208 |
+
"loss": 0.4569,
|
| 2209 |
+
"reward": 1.3949406147003174,
|
| 2210 |
+
"reward_std": 0.719992458820343,
|
| 2211 |
+
"rewards/connected_reward": 0.985119104385376,
|
| 2212 |
+
"rewards/format_reward": 0.4285714626312256,
|
| 2213 |
+
"rewards/tag_count_reward": 0.6517857313156128,
|
| 2214 |
+
"step": 147
|
| 2215 |
+
},
|
| 2216 |
+
{
|
| 2217 |
+
"clip_ratio": 0.0,
|
| 2218 |
+
"completion_length": 370.5000305175781,
|
| 2219 |
+
"epoch": 0.02368,
|
| 2220 |
+
"grad_norm": 80.20691680908203,
|
| 2221 |
+
"kl": 1.7421875,
|
| 2222 |
+
"learning_rate": 2.192924752854042e-09,
|
| 2223 |
+
"loss": 0.2907,
|
| 2224 |
+
"reward": 1.4619261026382446,
|
| 2225 |
+
"reward_std": 0.9718099236488342,
|
| 2226 |
+
"rewards/connected_reward": 1.0248724222183228,
|
| 2227 |
+
"rewards/format_reward": 0.4642857313156128,
|
| 2228 |
+
"rewards/tag_count_reward": 0.6830357313156128,
|
| 2229 |
+
"step": 148
|
| 2230 |
+
},
|
| 2231 |
+
{
|
| 2232 |
+
"clip_ratio": 0.0,
|
| 2233 |
+
"completion_length": 345.5714416503906,
|
| 2234 |
+
"epoch": 0.02384,
|
| 2235 |
+
"grad_norm": 57.89230728149414,
|
| 2236 |
+
"kl": 0.96875,
|
| 2237 |
+
"learning_rate": 5.48291312886251e-10,
|
| 2238 |
+
"loss": 0.3204,
|
| 2239 |
+
"reward": 1.9949404001235962,
|
| 2240 |
+
"reward_std": 0.6536844968795776,
|
| 2241 |
+
"rewards/connected_reward": 1.5922620296478271,
|
| 2242 |
+
"rewards/format_reward": 0.392857164144516,
|
| 2243 |
+
"rewards/tag_count_reward": 0.6875000596046448,
|
| 2244 |
+
"step": 149
|
| 2245 |
+
},
|
| 2246 |
+
{
|
| 2247 |
+
"clip_ratio": 0.0,
|
| 2248 |
+
"completion_length": 481.607177734375,
|
| 2249 |
+
"epoch": 0.024,
|
| 2250 |
+
"grad_norm": 70.68065643310547,
|
| 2251 |
+
"kl": 2.9375,
|
| 2252 |
+
"learning_rate": 0.0,
|
| 2253 |
+
"loss": 0.3935,
|
| 2254 |
+
"reward": 1.4215773344039917,
|
| 2255 |
+
"reward_std": 1.039550542831421,
|
| 2256 |
+
"rewards/connected_reward": 1.1175596714019775,
|
| 2257 |
+
"rewards/format_reward": 0.267857164144516,
|
| 2258 |
+
"rewards/tag_count_reward": 0.566964328289032,
|
| 2259 |
+
"step": 150
|
| 2260 |
+
},
|
| 2261 |
+
{
|
| 2262 |
+
"epoch": 0.024,
|
| 2263 |
+
"step": 150,
|
| 2264 |
+
"total_flos": 0.0,
|
| 2265 |
+
"train_loss": 0.31898806129537965,
|
| 2266 |
+
"train_runtime": 4133.207,
|
| 2267 |
+
"train_samples_per_second": 2.032,
|
| 2268 |
+
"train_steps_per_second": 0.036
|
| 2269 |
+
}
|
| 2270 |
+
],
|
| 2271 |
+
"logging_steps": 1,
|
| 2272 |
+
"max_steps": 150,
|
| 2273 |
+
"num_input_tokens_seen": 0,
|
| 2274 |
+
"num_train_epochs": 1,
|
| 2275 |
+
"save_steps": 500,
|
| 2276 |
+
"stateful_callbacks": {
|
| 2277 |
+
"TrainerControl": {
|
| 2278 |
+
"args": {
|
| 2279 |
+
"should_epoch_stop": false,
|
| 2280 |
+
"should_evaluate": false,
|
| 2281 |
+
"should_log": false,
|
| 2282 |
+
"should_save": true,
|
| 2283 |
+
"should_training_stop": true
|
| 2284 |
+
},
|
| 2285 |
+
"attributes": {}
|
| 2286 |
+
}
|
| 2287 |
+
},
|
| 2288 |
+
"total_flos": 0.0,
|
| 2289 |
+
"train_batch_size": 8,
|
| 2290 |
+
"trial_name": null,
|
| 2291 |
+
"trial_params": null
|
| 2292 |
+
}
|