Model save
Browse files- README.md +68 -0
- all_results.json +8 -0
- generation_config.json +9 -0
- train_results.json +8 -0
- trainer_state.json +2779 -0
README.md
ADDED
|
@@ -0,0 +1,68 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
base_model: akhauriyash/DeepSeek-R1-Distill-Qwen-1.5B-SpeculativeReasoner
|
| 3 |
+
library_name: transformers
|
| 4 |
+
model_name: 1.5B_E2EGRPO_8B_128Gen_4K_2xAcc
|
| 5 |
+
tags:
|
| 6 |
+
- generated_from_trainer
|
| 7 |
+
- trl
|
| 8 |
+
- grpo
|
| 9 |
+
licence: license
|
| 10 |
+
---
|
| 11 |
+
|
| 12 |
+
# Model Card for 1.5B_E2EGRPO_8B_128Gen_4K_2xAcc
|
| 13 |
+
|
| 14 |
+
This model is a fine-tuned version of [akhauriyash/DeepSeek-R1-Distill-Qwen-1.5B-SpeculativeReasoner](https://huggingface.co/akhauriyash/DeepSeek-R1-Distill-Qwen-1.5B-SpeculativeReasoner).
|
| 15 |
+
It has been trained using [TRL](https://github.com/huggingface/trl).
|
| 16 |
+
|
| 17 |
+
## Quick start
|
| 18 |
+
|
| 19 |
+
```python
|
| 20 |
+
from transformers import pipeline
|
| 21 |
+
|
| 22 |
+
question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
|
| 23 |
+
generator = pipeline("text-generation", model="lisali126/1.5B_E2EGRPO_8B_128Gen_4K_2xAcc", device="cuda")
|
| 24 |
+
output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
|
| 25 |
+
print(output["generated_text"])
|
| 26 |
+
```
|
| 27 |
+
|
| 28 |
+
## Training procedure
|
| 29 |
+
|
| 30 |
+
[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/splitreason/SpeculativeReasoning/runs/fhqi5anv)
|
| 31 |
+
|
| 32 |
+
|
| 33 |
+
This model was trained with GRPO, a method introduced in [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://huggingface.co/papers/2402.03300).
|
| 34 |
+
|
| 35 |
+
### Framework versions
|
| 36 |
+
|
| 37 |
+
- TRL: 0.16.0
|
| 38 |
+
- Transformers: 4.50.0
|
| 39 |
+
- Pytorch: 2.5.1
|
| 40 |
+
- Datasets: 3.6.0
|
| 41 |
+
- Tokenizers: 0.21.1
|
| 42 |
+
|
| 43 |
+
## Citations
|
| 44 |
+
|
| 45 |
+
Cite GRPO as:
|
| 46 |
+
|
| 47 |
+
```bibtex
|
| 48 |
+
@article{zhihong2024deepseekmath,
|
| 49 |
+
title = {{DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models}},
|
| 50 |
+
author = {Zhihong Shao and Peiyi Wang and Qihao Zhu and Runxin Xu and Junxiao Song and Mingchuan Zhang and Y. K. Li and Y. Wu and Daya Guo},
|
| 51 |
+
year = 2024,
|
| 52 |
+
eprint = {arXiv:2402.03300},
|
| 53 |
+
}
|
| 54 |
+
|
| 55 |
+
```
|
| 56 |
+
|
| 57 |
+
Cite TRL as:
|
| 58 |
+
|
| 59 |
+
```bibtex
|
| 60 |
+
@misc{vonwerra2022trl,
|
| 61 |
+
title = {{TRL: Transformer Reinforcement Learning}},
|
| 62 |
+
author = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang and Kashif Rasul and Quentin Gallouédec},
|
| 63 |
+
year = 2020,
|
| 64 |
+
journal = {GitHub repository},
|
| 65 |
+
publisher = {GitHub},
|
| 66 |
+
howpublished = {\url{https://github.com/huggingface/trl}}
|
| 67 |
+
}
|
| 68 |
+
```
|
all_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"total_flos": 0.0,
|
| 3 |
+
"train_loss": 2.9861142760828923e-07,
|
| 4 |
+
"train_runtime": 124.7789,
|
| 5 |
+
"train_samples": 93733,
|
| 6 |
+
"train_samples_per_second": 38.468,
|
| 7 |
+
"train_steps_per_second": 1.202
|
| 8 |
+
}
|
generation_config.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"bos_token_id": 151646,
|
| 4 |
+
"do_sample": true,
|
| 5 |
+
"eos_token_id": 151643,
|
| 6 |
+
"temperature": 0.6,
|
| 7 |
+
"top_p": 0.95,
|
| 8 |
+
"transformers_version": "4.50.0"
|
| 9 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"total_flos": 0.0,
|
| 3 |
+
"train_loss": 2.9861142760828923e-07,
|
| 4 |
+
"train_runtime": 124.7789,
|
| 5 |
+
"train_samples": 93733,
|
| 6 |
+
"train_samples_per_second": 38.468,
|
| 7 |
+
"train_steps_per_second": 1.202
|
| 8 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,2779 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_global_step": null,
|
| 3 |
+
"best_metric": null,
|
| 4 |
+
"best_model_checkpoint": null,
|
| 5 |
+
"epoch": 0.0036486616239744807,
|
| 6 |
+
"eval_steps": 500,
|
| 7 |
+
"global_step": 171,
|
| 8 |
+
"is_hyper_param_search": false,
|
| 9 |
+
"is_local_process_zero": true,
|
| 10 |
+
"is_world_process_zero": true,
|
| 11 |
+
"log_history": [
|
| 12 |
+
{
|
| 13 |
+
"clip_ratio": 0.0,
|
| 14 |
+
"completion_length": 2031.09375,
|
| 15 |
+
"epoch": 4.267394968741332e-05,
|
| 16 |
+
"grad_norm": 0.3983817873336318,
|
| 17 |
+
"kl": 0.0010356903076171875,
|
| 18 |
+
"learning_rate": 5e-08,
|
| 19 |
+
"loss": 0.0026,
|
| 20 |
+
"num_tokens": 71211.0,
|
| 21 |
+
"reward": 1.8085778951644897,
|
| 22 |
+
"reward_std": 0.6016984581947327,
|
| 23 |
+
"rewards/accuracy_reward": 0.0625,
|
| 24 |
+
"rewards/format_reward": 0.90625,
|
| 25 |
+
"rewards/tag_count_reward": 0.7773278653621674,
|
| 26 |
+
"step": 1
|
| 27 |
+
},
|
| 28 |
+
{
|
| 29 |
+
"clip_ratio": 0.0,
|
| 30 |
+
"completion_length": 2048.0,
|
| 31 |
+
"epoch": 8.534789937482664e-05,
|
| 32 |
+
"grad_norm": 0.3251471020810454,
|
| 33 |
+
"kl": 0.0007724761962890625,
|
| 34 |
+
"learning_rate": 1e-07,
|
| 35 |
+
"loss": 0.0,
|
| 36 |
+
"num_tokens": 141691.0,
|
| 37 |
+
"reward": 1.8670850992202759,
|
| 38 |
+
"reward_std": 0.7979374527931213,
|
| 39 |
+
"rewards/accuracy_reward": 0.0,
|
| 40 |
+
"rewards/format_reward": 0.96875,
|
| 41 |
+
"rewards/tag_count_reward": 0.8983351588249207,
|
| 42 |
+
"step": 2
|
| 43 |
+
},
|
| 44 |
+
{
|
| 45 |
+
"clip_ratio": 0.0,
|
| 46 |
+
"completion_length": 2048.0,
|
| 47 |
+
"epoch": 0.00012802184906223995,
|
| 48 |
+
"grad_norm": 0.3287029570359897,
|
| 49 |
+
"kl": 0.0007266998291015625,
|
| 50 |
+
"learning_rate": 1.5e-07,
|
| 51 |
+
"loss": 0.0,
|
| 52 |
+
"num_tokens": 213243.0,
|
| 53 |
+
"reward": 1.8505700826644897,
|
| 54 |
+
"reward_std": 1.1942417919635773,
|
| 55 |
+
"rewards/accuracy_reward": 0.03125,
|
| 56 |
+
"rewards/format_reward": 0.8125,
|
| 57 |
+
"rewards/tag_count_reward": 0.9755701720714569,
|
| 58 |
+
"step": 3
|
| 59 |
+
},
|
| 60 |
+
{
|
| 61 |
+
"clip_ratio": 0.0,
|
| 62 |
+
"completion_length": 2048.0,
|
| 63 |
+
"epoch": 0.00017069579874965328,
|
| 64 |
+
"grad_norm": 0.3344236699335657,
|
| 65 |
+
"kl": 0.000820159912109375,
|
| 66 |
+
"learning_rate": 2e-07,
|
| 67 |
+
"loss": 0.0,
|
| 68 |
+
"num_tokens": 285219.0,
|
| 69 |
+
"reward": 1.4784181118011475,
|
| 70 |
+
"reward_std": 0.9034450650215149,
|
| 71 |
+
"rewards/accuracy_reward": 0.0,
|
| 72 |
+
"rewards/format_reward": 0.90625,
|
| 73 |
+
"rewards/tag_count_reward": 0.5721680819988251,
|
| 74 |
+
"step": 4
|
| 75 |
+
},
|
| 76 |
+
{
|
| 77 |
+
"clip_ratio": 0.0,
|
| 78 |
+
"completion_length": 1931.0625,
|
| 79 |
+
"epoch": 0.0002133697484370666,
|
| 80 |
+
"grad_norm": 0.3987385108593764,
|
| 81 |
+
"kl": 0.0009784698486328125,
|
| 82 |
+
"learning_rate": 2.5e-07,
|
| 83 |
+
"loss": 0.0015,
|
| 84 |
+
"num_tokens": 352309.0,
|
| 85 |
+
"reward": 3.186710834503174,
|
| 86 |
+
"reward_std": 1.3371617197990417,
|
| 87 |
+
"rewards/accuracy_reward": 0.25,
|
| 88 |
+
"rewards/format_reward": 0.96875,
|
| 89 |
+
"rewards/tag_count_reward": 1.717960774898529,
|
| 90 |
+
"step": 5
|
| 91 |
+
},
|
| 92 |
+
{
|
| 93 |
+
"clip_ratio": 0.0,
|
| 94 |
+
"completion_length": 2035.375,
|
| 95 |
+
"epoch": 0.0002560436981244799,
|
| 96 |
+
"grad_norm": 0.35582728944677733,
|
| 97 |
+
"kl": 0.000858306884765625,
|
| 98 |
+
"learning_rate": 3e-07,
|
| 99 |
+
"loss": 0.0104,
|
| 100 |
+
"num_tokens": 423273.0,
|
| 101 |
+
"reward": 1.8892364501953125,
|
| 102 |
+
"reward_std": 0.8221084475517273,
|
| 103 |
+
"rewards/accuracy_reward": 0.03125,
|
| 104 |
+
"rewards/format_reward": 1.0,
|
| 105 |
+
"rewards/tag_count_reward": 0.8267365097999573,
|
| 106 |
+
"step": 6
|
| 107 |
+
},
|
| 108 |
+
{
|
| 109 |
+
"clip_ratio": 0.0,
|
| 110 |
+
"completion_length": 2048.0,
|
| 111 |
+
"epoch": 0.0002987176478118932,
|
| 112 |
+
"grad_norm": 0.32058515004177995,
|
| 113 |
+
"kl": 0.00078582763671875,
|
| 114 |
+
"learning_rate": 3.5e-07,
|
| 115 |
+
"loss": 0.0,
|
| 116 |
+
"num_tokens": 494601.0,
|
| 117 |
+
"reward": 2.4088869094848633,
|
| 118 |
+
"reward_std": 0.9408210217952728,
|
| 119 |
+
"rewards/accuracy_reward": 0.0625,
|
| 120 |
+
"rewards/format_reward": 0.96875,
|
| 121 |
+
"rewards/tag_count_reward": 1.3151369094848633,
|
| 122 |
+
"step": 7
|
| 123 |
+
},
|
| 124 |
+
{
|
| 125 |
+
"clip_ratio": 0.0,
|
| 126 |
+
"completion_length": 2048.0,
|
| 127 |
+
"epoch": 0.00034139159749930657,
|
| 128 |
+
"grad_norm": 0.34435846989411917,
|
| 129 |
+
"kl": 0.000797271728515625,
|
| 130 |
+
"learning_rate": 4e-07,
|
| 131 |
+
"loss": 0.0,
|
| 132 |
+
"num_tokens": 566841.0,
|
| 133 |
+
"reward": 1.951629400253296,
|
| 134 |
+
"reward_std": 1.0791807174682617,
|
| 135 |
+
"rewards/accuracy_reward": 0.0,
|
| 136 |
+
"rewards/format_reward": 0.96875,
|
| 137 |
+
"rewards/tag_count_reward": 0.9828794598579407,
|
| 138 |
+
"step": 8
|
| 139 |
+
},
|
| 140 |
+
{
|
| 141 |
+
"clip_ratio": 0.0,
|
| 142 |
+
"completion_length": 2048.0,
|
| 143 |
+
"epoch": 0.0003840655471867199,
|
| 144 |
+
"grad_norm": 0.32775594661215424,
|
| 145 |
+
"kl": 0.000762939453125,
|
| 146 |
+
"learning_rate": 4.5e-07,
|
| 147 |
+
"loss": 0.0,
|
| 148 |
+
"num_tokens": 636993.0,
|
| 149 |
+
"reward": 2.0051379203796387,
|
| 150 |
+
"reward_std": 0.9216841459274292,
|
| 151 |
+
"rewards/accuracy_reward": 0.0,
|
| 152 |
+
"rewards/format_reward": 0.9375,
|
| 153 |
+
"rewards/tag_count_reward": 1.0676379799842834,
|
| 154 |
+
"step": 9
|
| 155 |
+
},
|
| 156 |
+
{
|
| 157 |
+
"clip_ratio": 0.0,
|
| 158 |
+
"completion_length": 1740.6875,
|
| 159 |
+
"epoch": 0.0004267394968741332,
|
| 160 |
+
"grad_norm": 0.39153346737627215,
|
| 161 |
+
"kl": 0.0007114410400390625,
|
| 162 |
+
"learning_rate": 5e-07,
|
| 163 |
+
"loss": -0.004,
|
| 164 |
+
"num_tokens": 697599.0,
|
| 165 |
+
"reward": 2.7380948066711426,
|
| 166 |
+
"reward_std": 1.168121099472046,
|
| 167 |
+
"rewards/accuracy_reward": 0.0625,
|
| 168 |
+
"rewards/format_reward": 0.875,
|
| 169 |
+
"rewards/tag_count_reward": 1.738094836473465,
|
| 170 |
+
"step": 10
|
| 171 |
+
},
|
| 172 |
+
{
|
| 173 |
+
"clip_ratio": 0.0,
|
| 174 |
+
"completion_length": 2046.25,
|
| 175 |
+
"epoch": 0.0004694134465615465,
|
| 176 |
+
"grad_norm": 0.32659464326340937,
|
| 177 |
+
"kl": 0.00074005126953125,
|
| 178 |
+
"learning_rate": 5.5e-07,
|
| 179 |
+
"loss": 0.0005,
|
| 180 |
+
"num_tokens": 71104.0,
|
| 181 |
+
"reward": 1.7140210270881653,
|
| 182 |
+
"reward_std": 0.9808962345123291,
|
| 183 |
+
"rewards/accuracy_reward": 0.0,
|
| 184 |
+
"rewards/format_reward": 0.875,
|
| 185 |
+
"rewards/tag_count_reward": 0.8390210568904877,
|
| 186 |
+
"step": 11
|
| 187 |
+
},
|
| 188 |
+
{
|
| 189 |
+
"clip_ratio": 0.0,
|
| 190 |
+
"completion_length": 1901.375,
|
| 191 |
+
"epoch": 0.0005120873962489598,
|
| 192 |
+
"grad_norm": 0.3575649163379762,
|
| 193 |
+
"kl": 0.00080108642578125,
|
| 194 |
+
"learning_rate": 6e-07,
|
| 195 |
+
"loss": -0.0252,
|
| 196 |
+
"num_tokens": 138508.0,
|
| 197 |
+
"reward": 1.9060912132263184,
|
| 198 |
+
"reward_std": 1.0913816094398499,
|
| 199 |
+
"rewards/accuracy_reward": 0.0,
|
| 200 |
+
"rewards/format_reward": 0.875,
|
| 201 |
+
"rewards/tag_count_reward": 1.0310912430286407,
|
| 202 |
+
"step": 12
|
| 203 |
+
},
|
| 204 |
+
{
|
| 205 |
+
"clip_ratio": 0.0,
|
| 206 |
+
"completion_length": 2048.0,
|
| 207 |
+
"epoch": 0.0005547613459363732,
|
| 208 |
+
"grad_norm": 0.3223149440544977,
|
| 209 |
+
"kl": 0.00067901611328125,
|
| 210 |
+
"learning_rate": 6.5e-07,
|
| 211 |
+
"loss": 0.0,
|
| 212 |
+
"num_tokens": 209404.0,
|
| 213 |
+
"reward": 2.176819324493408,
|
| 214 |
+
"reward_std": 0.7264281809329987,
|
| 215 |
+
"rewards/accuracy_reward": 0.0,
|
| 216 |
+
"rewards/format_reward": 1.0,
|
| 217 |
+
"rewards/tag_count_reward": 1.1768193244934082,
|
| 218 |
+
"step": 13
|
| 219 |
+
},
|
| 220 |
+
{
|
| 221 |
+
"clip_ratio": 0.0,
|
| 222 |
+
"completion_length": 2004.9375,
|
| 223 |
+
"epoch": 0.0005974352956237864,
|
| 224 |
+
"grad_norm": 0.38930011034204937,
|
| 225 |
+
"kl": 0.000835418701171875,
|
| 226 |
+
"learning_rate": 7e-07,
|
| 227 |
+
"loss": 0.0198,
|
| 228 |
+
"num_tokens": 278546.0,
|
| 229 |
+
"reward": 1.4568162560462952,
|
| 230 |
+
"reward_std": 1.36779123544693,
|
| 231 |
+
"rewards/accuracy_reward": 0.0625,
|
| 232 |
+
"rewards/format_reward": 0.8125,
|
| 233 |
+
"rewards/tag_count_reward": 0.5193163156509399,
|
| 234 |
+
"step": 14
|
| 235 |
+
},
|
| 236 |
+
{
|
| 237 |
+
"clip_ratio": 0.0,
|
| 238 |
+
"completion_length": 2048.0,
|
| 239 |
+
"epoch": 0.0006401092453111998,
|
| 240 |
+
"grad_norm": 0.36894987673005736,
|
| 241 |
+
"kl": 0.000782012939453125,
|
| 242 |
+
"learning_rate": 7.5e-07,
|
| 243 |
+
"loss": 0.0,
|
| 244 |
+
"num_tokens": 350138.0,
|
| 245 |
+
"reward": 1.668050229549408,
|
| 246 |
+
"reward_std": 1.214027762413025,
|
| 247 |
+
"rewards/accuracy_reward": 0.0,
|
| 248 |
+
"rewards/format_reward": 0.84375,
|
| 249 |
+
"rewards/tag_count_reward": 0.8243001699447632,
|
| 250 |
+
"step": 15
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"clip_ratio": 0.0,
|
| 254 |
+
"completion_length": 2048.0,
|
| 255 |
+
"epoch": 0.0006827831949986131,
|
| 256 |
+
"grad_norm": 0.37575345114098535,
|
| 257 |
+
"kl": 0.000896453857421875,
|
| 258 |
+
"learning_rate": 8e-07,
|
| 259 |
+
"loss": 0.0,
|
| 260 |
+
"num_tokens": 422122.0,
|
| 261 |
+
"reward": 1.5416027903556824,
|
| 262 |
+
"reward_std": 1.0562286078929901,
|
| 263 |
+
"rewards/accuracy_reward": 0.0,
|
| 264 |
+
"rewards/format_reward": 0.75,
|
| 265 |
+
"rewards/tag_count_reward": 0.7916028201580048,
|
| 266 |
+
"step": 16
|
| 267 |
+
},
|
| 268 |
+
{
|
| 269 |
+
"clip_ratio": 0.0,
|
| 270 |
+
"completion_length": 2042.25,
|
| 271 |
+
"epoch": 0.0007254571446860264,
|
| 272 |
+
"grad_norm": 0.2993799169994233,
|
| 273 |
+
"kl": 0.0006103515625,
|
| 274 |
+
"learning_rate": 8.499999999999999e-07,
|
| 275 |
+
"loss": 0.0026,
|
| 276 |
+
"num_tokens": 493314.0,
|
| 277 |
+
"reward": 2.7380008697509766,
|
| 278 |
+
"reward_std": 1.1992889642715454,
|
| 279 |
+
"rewards/accuracy_reward": 0.1875,
|
| 280 |
+
"rewards/format_reward": 0.9375,
|
| 281 |
+
"rewards/tag_count_reward": 1.4255009293556213,
|
| 282 |
+
"step": 17
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"clip_ratio": 0.0,
|
| 286 |
+
"completion_length": 2009.25,
|
| 287 |
+
"epoch": 0.0007681310943734397,
|
| 288 |
+
"grad_norm": 0.3360225704032281,
|
| 289 |
+
"kl": 0.0007991790771484375,
|
| 290 |
+
"learning_rate": 9e-07,
|
| 291 |
+
"loss": -0.0085,
|
| 292 |
+
"num_tokens": 563994.0,
|
| 293 |
+
"reward": 2.2660080194473267,
|
| 294 |
+
"reward_std": 1.3002614974975586,
|
| 295 |
+
"rewards/accuracy_reward": 0.09375,
|
| 296 |
+
"rewards/format_reward": 0.8125,
|
| 297 |
+
"rewards/tag_count_reward": 1.2660080194473267,
|
| 298 |
+
"step": 18
|
| 299 |
+
},
|
| 300 |
+
{
|
| 301 |
+
"clip_ratio": 0.0,
|
| 302 |
+
"completion_length": 2037.40625,
|
| 303 |
+
"epoch": 0.000810805044060853,
|
| 304 |
+
"grad_norm": 0.34256891205027545,
|
| 305 |
+
"kl": 0.0008869171142578125,
|
| 306 |
+
"learning_rate": 9.499999999999999e-07,
|
| 307 |
+
"loss": 0.0063,
|
| 308 |
+
"num_tokens": 634519.0,
|
| 309 |
+
"reward": 2.343594551086426,
|
| 310 |
+
"reward_std": 0.8207766711711884,
|
| 311 |
+
"rewards/accuracy_reward": 0.125,
|
| 312 |
+
"rewards/format_reward": 0.96875,
|
| 313 |
+
"rewards/tag_count_reward": 1.124844491481781,
|
| 314 |
+
"step": 19
|
| 315 |
+
},
|
| 316 |
+
{
|
| 317 |
+
"clip_ratio": 0.0,
|
| 318 |
+
"completion_length": 1908.03125,
|
| 319 |
+
"epoch": 0.0008534789937482664,
|
| 320 |
+
"grad_norm": 0.3006084532734205,
|
| 321 |
+
"kl": 0.0005855560302734375,
|
| 322 |
+
"learning_rate": 1e-06,
|
| 323 |
+
"loss": -0.0068,
|
| 324 |
+
"num_tokens": 701096.0,
|
| 325 |
+
"reward": 2.1461480855941772,
|
| 326 |
+
"reward_std": 1.2458754181861877,
|
| 327 |
+
"rewards/accuracy_reward": 0.0625,
|
| 328 |
+
"rewards/format_reward": 0.8125,
|
| 329 |
+
"rewards/tag_count_reward": 1.2086480855941772,
|
| 330 |
+
"step": 20
|
| 331 |
+
},
|
| 332 |
+
{
|
| 333 |
+
"clip_ratio": 0.0,
|
| 334 |
+
"completion_length": 2025.6875,
|
| 335 |
+
"epoch": 0.0008961529434356797,
|
| 336 |
+
"grad_norm": 0.36384433442763486,
|
| 337 |
+
"kl": 0.000881195068359375,
|
| 338 |
+
"learning_rate": 9.99931462820376e-07,
|
| 339 |
+
"loss": 0.0126,
|
| 340 |
+
"num_tokens": 773430.0,
|
| 341 |
+
"reward": 1.7199882864952087,
|
| 342 |
+
"reward_std": 1.6680212020874023,
|
| 343 |
+
"rewards/accuracy_reward": 0.03125,
|
| 344 |
+
"rewards/format_reward": 0.75,
|
| 345 |
+
"rewards/tag_count_reward": 0.9074882864952087,
|
| 346 |
+
"step": 21
|
| 347 |
+
},
|
| 348 |
+
{
|
| 349 |
+
"clip_ratio": 0.0,
|
| 350 |
+
"completion_length": 2009.40625,
|
| 351 |
+
"epoch": 0.000938826893123093,
|
| 352 |
+
"grad_norm": 0.3616488312640992,
|
| 353 |
+
"kl": 0.000823974609375,
|
| 354 |
+
"learning_rate": 9.997258721585931e-07,
|
| 355 |
+
"loss": 0.024,
|
| 356 |
+
"num_tokens": 845635.0,
|
| 357 |
+
"reward": 2.135270118713379,
|
| 358 |
+
"reward_std": 0.9850693941116333,
|
| 359 |
+
"rewards/accuracy_reward": 0.0,
|
| 360 |
+
"rewards/format_reward": 0.875,
|
| 361 |
+
"rewards/tag_count_reward": 1.2602701783180237,
|
| 362 |
+
"step": 22
|
| 363 |
+
},
|
| 364 |
+
{
|
| 365 |
+
"clip_ratio": 0.0,
|
| 366 |
+
"completion_length": 2029.75,
|
| 367 |
+
"epoch": 0.0009815008428105062,
|
| 368 |
+
"grad_norm": 0.30599912955907155,
|
| 369 |
+
"kl": 0.00061798095703125,
|
| 370 |
+
"learning_rate": 9.993832906395582e-07,
|
| 371 |
+
"loss": 0.0103,
|
| 372 |
+
"num_tokens": 915331.0,
|
| 373 |
+
"reward": 2.809187412261963,
|
| 374 |
+
"reward_std": 1.0471147298812866,
|
| 375 |
+
"rewards/accuracy_reward": 0.21875,
|
| 376 |
+
"rewards/format_reward": 0.9375,
|
| 377 |
+
"rewards/tag_count_reward": 1.4341875314712524,
|
| 378 |
+
"step": 23
|
| 379 |
+
},
|
| 380 |
+
{
|
| 381 |
+
"clip_ratio": 0.0,
|
| 382 |
+
"completion_length": 2048.0,
|
| 383 |
+
"epoch": 0.0010241747924979196,
|
| 384 |
+
"grad_norm": 0.3904882439942461,
|
| 385 |
+
"kl": 0.00095367431640625,
|
| 386 |
+
"learning_rate": 9.989038226169207e-07,
|
| 387 |
+
"loss": 0.0,
|
| 388 |
+
"num_tokens": 986507.0,
|
| 389 |
+
"reward": 1.802339255809784,
|
| 390 |
+
"reward_std": 0.9541150331497192,
|
| 391 |
+
"rewards/accuracy_reward": 0.0,
|
| 392 |
+
"rewards/format_reward": 0.84375,
|
| 393 |
+
"rewards/tag_count_reward": 0.9585893154144287,
|
| 394 |
+
"step": 24
|
| 395 |
+
},
|
| 396 |
+
{
|
| 397 |
+
"clip_ratio": 0.0,
|
| 398 |
+
"completion_length": 2016.5625,
|
| 399 |
+
"epoch": 0.001066848742185333,
|
| 400 |
+
"grad_norm": 0.32161813853306614,
|
| 401 |
+
"kl": 0.00067138671875,
|
| 402 |
+
"learning_rate": 9.982876141412855e-07,
|
| 403 |
+
"loss": 0.014,
|
| 404 |
+
"num_tokens": 1056101.0,
|
| 405 |
+
"reward": 2.8008618354797363,
|
| 406 |
+
"reward_std": 0.8961454629898071,
|
| 407 |
+
"rewards/accuracy_reward": 0.1875,
|
| 408 |
+
"rewards/format_reward": 0.9375,
|
| 409 |
+
"rewards/tag_count_reward": 1.4883618354797363,
|
| 410 |
+
"step": 25
|
| 411 |
+
},
|
| 412 |
+
{
|
| 413 |
+
"clip_ratio": 0.0,
|
| 414 |
+
"completion_length": 2042.71875,
|
| 415 |
+
"epoch": 0.0011095226918727463,
|
| 416 |
+
"grad_norm": 0.33527075865177447,
|
| 417 |
+
"kl": 0.000720977783203125,
|
| 418 |
+
"learning_rate": 9.975348529157229e-07,
|
| 419 |
+
"loss": 0.0048,
|
| 420 |
+
"num_tokens": 1129244.0,
|
| 421 |
+
"reward": 1.8309496641159058,
|
| 422 |
+
"reward_std": 1.159160703420639,
|
| 423 |
+
"rewards/accuracy_reward": 0.03125,
|
| 424 |
+
"rewards/format_reward": 0.8125,
|
| 425 |
+
"rewards/tag_count_reward": 0.955949604511261,
|
| 426 |
+
"step": 26
|
| 427 |
+
},
|
| 428 |
+
{
|
| 429 |
+
"clip_ratio": 0.0,
|
| 430 |
+
"completion_length": 2048.0,
|
| 431 |
+
"epoch": 0.0011521966415601597,
|
| 432 |
+
"grad_norm": 0.32908808477356694,
|
| 433 |
+
"kl": 0.0007114410400390625,
|
| 434 |
+
"learning_rate": 9.96645768238595e-07,
|
| 435 |
+
"loss": 0.0,
|
| 436 |
+
"num_tokens": 1200028.0,
|
| 437 |
+
"reward": 2.0427041053771973,
|
| 438 |
+
"reward_std": 0.8570608794689178,
|
| 439 |
+
"rewards/accuracy_reward": 0.0,
|
| 440 |
+
"rewards/format_reward": 0.9375,
|
| 441 |
+
"rewards/tag_count_reward": 1.1052041053771973,
|
| 442 |
+
"step": 27
|
| 443 |
+
},
|
| 444 |
+
{
|
| 445 |
+
"clip_ratio": 0.0,
|
| 446 |
+
"completion_length": 2048.0,
|
| 447 |
+
"epoch": 0.0011948705912475728,
|
| 448 |
+
"grad_norm": 0.3620332658267825,
|
| 449 |
+
"kl": 0.000873565673828125,
|
| 450 |
+
"learning_rate": 9.956206309337066e-07,
|
| 451 |
+
"loss": 0.0,
|
| 452 |
+
"num_tokens": 1273572.0,
|
| 453 |
+
"reward": 1.771780788898468,
|
| 454 |
+
"reward_std": 0.40034283697605133,
|
| 455 |
+
"rewards/accuracy_reward": 0.0,
|
| 456 |
+
"rewards/format_reward": 0.96875,
|
| 457 |
+
"rewards/tag_count_reward": 0.8030306994915009,
|
| 458 |
+
"step": 28
|
| 459 |
+
},
|
| 460 |
+
{
|
| 461 |
+
"clip_ratio": 0.0,
|
| 462 |
+
"completion_length": 2031.0625,
|
| 463 |
+
"epoch": 0.0012375445409349862,
|
| 464 |
+
"grad_norm": 0.34069596272177766,
|
| 465 |
+
"kl": 0.00084686279296875,
|
| 466 |
+
"learning_rate": 9.944597532678119e-07,
|
| 467 |
+
"loss": -0.0057,
|
| 468 |
+
"num_tokens": 1343238.0,
|
| 469 |
+
"reward": 1.7035346031188965,
|
| 470 |
+
"reward_std": 0.9985750615596771,
|
| 471 |
+
"rewards/accuracy_reward": 0.0,
|
| 472 |
+
"rewards/format_reward": 0.84375,
|
| 473 |
+
"rewards/tag_count_reward": 0.8597846627235413,
|
| 474 |
+
"step": 29
|
| 475 |
+
},
|
| 476 |
+
{
|
| 477 |
+
"clip_ratio": 0.0,
|
| 478 |
+
"completion_length": 2048.0,
|
| 479 |
+
"epoch": 0.0012802184906223995,
|
| 480 |
+
"grad_norm": 0.35349092982266106,
|
| 481 |
+
"kl": 0.0009021759033203125,
|
| 482 |
+
"learning_rate": 9.931634888554935e-07,
|
| 483 |
+
"loss": 0.0,
|
| 484 |
+
"num_tokens": 1414750.0,
|
| 485 |
+
"reward": 1.8466508388519287,
|
| 486 |
+
"reward_std": 0.9060315191745758,
|
| 487 |
+
"rewards/accuracy_reward": 0.0,
|
| 488 |
+
"rewards/format_reward": 0.9375,
|
| 489 |
+
"rewards/tag_count_reward": 0.9091507792472839,
|
| 490 |
+
"step": 30
|
| 491 |
+
},
|
| 492 |
+
{
|
| 493 |
+
"clip_ratio": 0.0,
|
| 494 |
+
"completion_length": 2048.0,
|
| 495 |
+
"epoch": 0.001322892440309813,
|
| 496 |
+
"grad_norm": 0.38450317214942537,
|
| 497 |
+
"kl": 0.00091552734375,
|
| 498 |
+
"learning_rate": 9.917322325514487e-07,
|
| 499 |
+
"loss": 0.0,
|
| 500 |
+
"num_tokens": 1486590.0,
|
| 501 |
+
"reward": 1.9213616251945496,
|
| 502 |
+
"reward_std": 0.9988719820976257,
|
| 503 |
+
"rewards/accuracy_reward": 0.0,
|
| 504 |
+
"rewards/format_reward": 0.875,
|
| 505 |
+
"rewards/tag_count_reward": 1.0463614761829376,
|
| 506 |
+
"step": 31
|
| 507 |
+
},
|
| 508 |
+
{
|
| 509 |
+
"clip_ratio": 0.0,
|
| 510 |
+
"completion_length": 1984.6875,
|
| 511 |
+
"epoch": 0.0013655663899972263,
|
| 512 |
+
"grad_norm": 0.3058960875298296,
|
| 513 |
+
"kl": 0.0005664825439453125,
|
| 514 |
+
"learning_rate": 9.901664203302124e-07,
|
| 515 |
+
"loss": 0.0082,
|
| 516 |
+
"num_tokens": 1555188.0,
|
| 517 |
+
"reward": 2.964768171310425,
|
| 518 |
+
"reward_std": 1.3874636888504028,
|
| 519 |
+
"rewards/accuracy_reward": 0.21875,
|
| 520 |
+
"rewards/format_reward": 0.90625,
|
| 521 |
+
"rewards/tag_count_reward": 1.6210182905197144,
|
| 522 |
+
"step": 32
|
| 523 |
+
},
|
| 524 |
+
{
|
| 525 |
+
"clip_ratio": 0.0,
|
| 526 |
+
"completion_length": 1987.71875,
|
| 527 |
+
"epoch": 0.0014082403396846394,
|
| 528 |
+
"grad_norm": 0.3346926556705837,
|
| 529 |
+
"kl": 0.000774383544921875,
|
| 530 |
+
"learning_rate": 9.88466529153356e-07,
|
| 531 |
+
"loss": 0.0126,
|
| 532 |
+
"num_tokens": 1624083.0,
|
| 533 |
+
"reward": 2.5380399227142334,
|
| 534 |
+
"reward_std": 1.2484303712844849,
|
| 535 |
+
"rewards/accuracy_reward": 0.15625,
|
| 536 |
+
"rewards/format_reward": 0.90625,
|
| 537 |
+
"rewards/tag_count_reward": 1.3192899227142334,
|
| 538 |
+
"step": 33
|
| 539 |
+
},
|
| 540 |
+
{
|
| 541 |
+
"clip_ratio": 0.0,
|
| 542 |
+
"completion_length": 2041.5,
|
| 543 |
+
"epoch": 0.0014509142893720528,
|
| 544 |
+
"grad_norm": 0.33957311973478305,
|
| 545 |
+
"kl": 0.0008716583251953125,
|
| 546 |
+
"learning_rate": 9.866330768241983e-07,
|
| 547 |
+
"loss": 0.0059,
|
| 548 |
+
"num_tokens": 1697067.0,
|
| 549 |
+
"reward": 2.0916172862052917,
|
| 550 |
+
"reward_std": 1.0687910616397858,
|
| 551 |
+
"rewards/accuracy_reward": 0.125,
|
| 552 |
+
"rewards/format_reward": 0.90625,
|
| 553 |
+
"rewards/tag_count_reward": 0.9353671967983246,
|
| 554 |
+
"step": 34
|
| 555 |
+
},
|
| 556 |
+
{
|
| 557 |
+
"clip_ratio": 0.0,
|
| 558 |
+
"completion_length": 1583.78125,
|
| 559 |
+
"epoch": 0.0014935882390594661,
|
| 560 |
+
"grad_norm": 0.4056886840426401,
|
| 561 |
+
"kl": 0.0007305145263671875,
|
| 562 |
+
"learning_rate": 9.846666218300807e-07,
|
| 563 |
+
"loss": 0.0221,
|
| 564 |
+
"num_tokens": 1752476.0,
|
| 565 |
+
"reward": 2.631199359893799,
|
| 566 |
+
"reward_std": 0.7480657994747162,
|
| 567 |
+
"rewards/accuracy_reward": 0.0,
|
| 568 |
+
"rewards/format_reward": 1.0,
|
| 569 |
+
"rewards/tag_count_reward": 1.6311993300914764,
|
| 570 |
+
"step": 35
|
| 571 |
+
},
|
| 572 |
+
{
|
| 573 |
+
"clip_ratio": 0.0,
|
| 574 |
+
"completion_length": 1915.21875,
|
| 575 |
+
"epoch": 0.0015362621887468795,
|
| 576 |
+
"grad_norm": 0.4000863473870498,
|
| 577 |
+
"kl": 0.0006809234619140625,
|
| 578 |
+
"learning_rate": 9.825677631722435e-07,
|
| 579 |
+
"loss": 0.0169,
|
| 580 |
+
"num_tokens": 1819779.0,
|
| 581 |
+
"reward": 3.2577226161956787,
|
| 582 |
+
"reward_std": 1.5010907649993896,
|
| 583 |
+
"rewards/accuracy_reward": 0.40625,
|
| 584 |
+
"rewards/format_reward": 0.90625,
|
| 585 |
+
"rewards/tag_count_reward": 1.5389727354049683,
|
| 586 |
+
"step": 36
|
| 587 |
+
},
|
| 588 |
+
{
|
| 589 |
+
"clip_ratio": 0.0,
|
| 590 |
+
"completion_length": 2048.0,
|
| 591 |
+
"epoch": 0.0015789361384342929,
|
| 592 |
+
"grad_norm": 0.3571328945597946,
|
| 593 |
+
"kl": 0.000835418701171875,
|
| 594 |
+
"learning_rate": 9.80337140183366e-07,
|
| 595 |
+
"loss": 0.0,
|
| 596 |
+
"num_tokens": 1890907.0,
|
| 597 |
+
"reward": 1.9089466333389282,
|
| 598 |
+
"reward_std": 1.2897401452064514,
|
| 599 |
+
"rewards/accuracy_reward": 0.0,
|
| 600 |
+
"rewards/format_reward": 0.90625,
|
| 601 |
+
"rewards/tag_count_reward": 1.0026965141296387,
|
| 602 |
+
"step": 37
|
| 603 |
+
},
|
| 604 |
+
{
|
| 605 |
+
"clip_ratio": 0.0,
|
| 606 |
+
"completion_length": 1984.65625,
|
| 607 |
+
"epoch": 0.001621610088121706,
|
| 608 |
+
"grad_norm": 0.34826551897125985,
|
| 609 |
+
"kl": 0.000888824462890625,
|
| 610 |
+
"learning_rate": 9.779754323328192e-07,
|
| 611 |
+
"loss": -0.0048,
|
| 612 |
+
"num_tokens": 1962112.0,
|
| 613 |
+
"reward": 2.465604305267334,
|
| 614 |
+
"reward_std": 0.6719761490821838,
|
| 615 |
+
"rewards/accuracy_reward": 0.15625,
|
| 616 |
+
"rewards/format_reward": 0.96875,
|
| 617 |
+
"rewards/tag_count_reward": 1.184354305267334,
|
| 618 |
+
"step": 38
|
| 619 |
+
},
|
| 620 |
+
{
|
| 621 |
+
"clip_ratio": 0.0,
|
| 622 |
+
"completion_length": 1986.3125,
|
| 623 |
+
"epoch": 0.0016642840378091194,
|
| 624 |
+
"grad_norm": 0.3626473077888914,
|
| 625 |
+
"kl": 0.0008087158203125,
|
| 626 |
+
"learning_rate": 9.754833590196926e-07,
|
| 627 |
+
"loss": 0.0128,
|
| 628 |
+
"num_tokens": 2031394.0,
|
| 629 |
+
"reward": 2.3278441429138184,
|
| 630 |
+
"reward_std": 1.3251615166664124,
|
| 631 |
+
"rewards/accuracy_reward": 0.15625,
|
| 632 |
+
"rewards/format_reward": 0.875,
|
| 633 |
+
"rewards/tag_count_reward": 1.1403440535068512,
|
| 634 |
+
"step": 39
|
| 635 |
+
},
|
| 636 |
+
{
|
| 637 |
+
"clip_ratio": 0.0,
|
| 638 |
+
"completion_length": 1997.78125,
|
| 639 |
+
"epoch": 0.0017069579874965327,
|
| 640 |
+
"grad_norm": 0.37047498961390796,
|
| 641 |
+
"kl": 0.000820159912109375,
|
| 642 |
+
"learning_rate": 9.728616793536587e-07,
|
| 643 |
+
"loss": 0.0037,
|
| 644 |
+
"num_tokens": 2101315.0,
|
| 645 |
+
"reward": 2.664921522140503,
|
| 646 |
+
"reward_std": 0.903692215681076,
|
| 647 |
+
"rewards/accuracy_reward": 0.25,
|
| 648 |
+
"rewards/format_reward": 1.0,
|
| 649 |
+
"rewards/tag_count_reward": 1.4149215817451477,
|
| 650 |
+
"step": 40
|
| 651 |
+
},
|
| 652 |
+
{
|
| 653 |
+
"clip_ratio": 0.0,
|
| 654 |
+
"completion_length": 2015.375,
|
| 655 |
+
"epoch": 0.001749631937183946,
|
| 656 |
+
"grad_norm": 0.3377579072159423,
|
| 657 |
+
"kl": 0.000823974609375,
|
| 658 |
+
"learning_rate": 9.701111919237408e-07,
|
| 659 |
+
"loss": 0.0138,
|
| 660 |
+
"num_tokens": 2171239.0,
|
| 661 |
+
"reward": 2.366754174232483,
|
| 662 |
+
"reward_std": 1.4368875622749329,
|
| 663 |
+
"rewards/accuracy_reward": 0.1875,
|
| 664 |
+
"rewards/format_reward": 0.84375,
|
| 665 |
+
"rewards/tag_count_reward": 1.1480042934417725,
|
| 666 |
+
"step": 41
|
| 667 |
+
},
|
| 668 |
+
{
|
| 669 |
+
"clip_ratio": 0.0,
|
| 670 |
+
"completion_length": 2005.6875,
|
| 671 |
+
"epoch": 0.0017923058868713595,
|
| 672 |
+
"grad_norm": 0.37113198614157994,
|
| 673 |
+
"kl": 0.000865936279296875,
|
| 674 |
+
"learning_rate": 9.672327345550543e-07,
|
| 675 |
+
"loss": 0.0094,
|
| 676 |
+
"num_tokens": 2241533.0,
|
| 677 |
+
"reward": 2.2223082780838013,
|
| 678 |
+
"reward_std": 1.1564024090766907,
|
| 679 |
+
"rewards/accuracy_reward": 0.1875,
|
| 680 |
+
"rewards/format_reward": 0.875,
|
| 681 |
+
"rewards/tag_count_reward": 0.9723082184791565,
|
| 682 |
+
"step": 42
|
| 683 |
+
},
|
| 684 |
+
{
|
| 685 |
+
"clip_ratio": 0.0,
|
| 686 |
+
"completion_length": 2011.71875,
|
| 687 |
+
"epoch": 0.0018349798365587726,
|
| 688 |
+
"grad_norm": 0.3307008971877644,
|
| 689 |
+
"kl": 0.0007457733154296875,
|
| 690 |
+
"learning_rate": 9.64227184053598e-07,
|
| 691 |
+
"loss": 0.0161,
|
| 692 |
+
"num_tokens": 2315620.0,
|
| 693 |
+
"reward": 2.1009689569473267,
|
| 694 |
+
"reward_std": 1.3331063985824585,
|
| 695 |
+
"rewards/accuracy_reward": 0.125,
|
| 696 |
+
"rewards/format_reward": 0.78125,
|
| 697 |
+
"rewards/tag_count_reward": 1.0697189569473267,
|
| 698 |
+
"step": 43
|
| 699 |
+
},
|
| 700 |
+
{
|
| 701 |
+
"clip_ratio": 0.0,
|
| 702 |
+
"completion_length": 2048.0,
|
| 703 |
+
"epoch": 0.001877653786246186,
|
| 704 |
+
"grad_norm": 0.368962193689309,
|
| 705 |
+
"kl": 0.000934600830078125,
|
| 706 |
+
"learning_rate": 9.610954559391704e-07,
|
| 707 |
+
"loss": 0.0,
|
| 708 |
+
"num_tokens": 2386116.0,
|
| 709 |
+
"reward": 1.7588238716125488,
|
| 710 |
+
"reward_std": 0.8519691824913025,
|
| 711 |
+
"rewards/accuracy_reward": 0.0,
|
| 712 |
+
"rewards/format_reward": 0.9375,
|
| 713 |
+
"rewards/tag_count_reward": 0.8213238716125488,
|
| 714 |
+
"step": 44
|
| 715 |
+
},
|
| 716 |
+
{
|
| 717 |
+
"clip_ratio": 0.0,
|
| 718 |
+
"completion_length": 2001.9375,
|
| 719 |
+
"epoch": 0.0019203277359335993,
|
| 720 |
+
"grad_norm": 0.34146280705998855,
|
| 721 |
+
"kl": 0.000732421875,
|
| 722 |
+
"learning_rate": 9.578385041664925e-07,
|
| 723 |
+
"loss": 0.0091,
|
| 724 |
+
"num_tokens": 2456554.0,
|
| 725 |
+
"reward": 2.2176711559295654,
|
| 726 |
+
"reward_std": 1.2107892632484436,
|
| 727 |
+
"rewards/accuracy_reward": 0.0,
|
| 728 |
+
"rewards/format_reward": 0.875,
|
| 729 |
+
"rewards/tag_count_reward": 1.3426711559295654,
|
| 730 |
+
"step": 45
|
| 731 |
+
},
|
| 732 |
+
{
|
| 733 |
+
"clip_ratio": 0.0,
|
| 734 |
+
"completion_length": 1962.46875,
|
| 735 |
+
"epoch": 0.0019630016856210125,
|
| 736 |
+
"grad_norm": 0.377746839715635,
|
| 737 |
+
"kl": 0.0009136199951171875,
|
| 738 |
+
"learning_rate": 9.54457320834625e-07,
|
| 739 |
+
"loss": -0.0045,
|
| 740 |
+
"num_tokens": 2525593.0,
|
| 741 |
+
"reward": 1.9368326663970947,
|
| 742 |
+
"reward_std": 1.222200632095337,
|
| 743 |
+
"rewards/accuracy_reward": 0.0,
|
| 744 |
+
"rewards/format_reward": 0.84375,
|
| 745 |
+
"rewards/tag_count_reward": 1.09308260679245,
|
| 746 |
+
"step": 46
|
| 747 |
+
},
|
| 748 |
+
{
|
| 749 |
+
"clip_ratio": 0.0,
|
| 750 |
+
"completion_length": 2048.0,
|
| 751 |
+
"epoch": 0.002005675635308426,
|
| 752 |
+
"grad_norm": 0.32785782132503205,
|
| 753 |
+
"kl": 0.0006885528564453125,
|
| 754 |
+
"learning_rate": 9.509529358847654e-07,
|
| 755 |
+
"loss": 0.0,
|
| 756 |
+
"num_tokens": 2597745.0,
|
| 757 |
+
"reward": 1.5949768424034119,
|
| 758 |
+
"reward_std": 0.6964653581380844,
|
| 759 |
+
"rewards/accuracy_reward": 0.0,
|
| 760 |
+
"rewards/format_reward": 0.90625,
|
| 761 |
+
"rewards/tag_count_reward": 0.6887268722057343,
|
| 762 |
+
"step": 47
|
| 763 |
+
},
|
| 764 |
+
{
|
| 765 |
+
"clip_ratio": 0.0,
|
| 766 |
+
"completion_length": 2048.0,
|
| 767 |
+
"epoch": 0.002048349584995839,
|
| 768 |
+
"grad_norm": 0.3570518580896509,
|
| 769 |
+
"kl": 0.0008258819580078125,
|
| 770 |
+
"learning_rate": 9.473264167865171e-07,
|
| 771 |
+
"loss": 0.0,
|
| 772 |
+
"num_tokens": 2669161.0,
|
| 773 |
+
"reward": 1.5660050511360168,
|
| 774 |
+
"reward_std": 0.6446836590766907,
|
| 775 |
+
"rewards/accuracy_reward": 0.0,
|
| 776 |
+
"rewards/format_reward": 0.9375,
|
| 777 |
+
"rewards/tag_count_reward": 0.6285050213336945,
|
| 778 |
+
"step": 48
|
| 779 |
+
},
|
| 780 |
+
{
|
| 781 |
+
"clip_ratio": 0.0,
|
| 782 |
+
"completion_length": 2048.0,
|
| 783 |
+
"epoch": 0.0020910235346832525,
|
| 784 |
+
"grad_norm": 0.31954678676989506,
|
| 785 |
+
"kl": 0.0007266998291015625,
|
| 786 |
+
"learning_rate": 9.43578868212728e-07,
|
| 787 |
+
"loss": 0.0,
|
| 788 |
+
"num_tokens": 2739745.0,
|
| 789 |
+
"reward": 1.9797396659851074,
|
| 790 |
+
"reward_std": 1.0861598253250122,
|
| 791 |
+
"rewards/accuracy_reward": 0.0,
|
| 792 |
+
"rewards/format_reward": 0.875,
|
| 793 |
+
"rewards/tag_count_reward": 1.104739785194397,
|
| 794 |
+
"step": 49
|
| 795 |
+
},
|
| 796 |
+
{
|
| 797 |
+
"clip_ratio": 0.0,
|
| 798 |
+
"completion_length": 2048.0,
|
| 799 |
+
"epoch": 0.002133697484370666,
|
| 800 |
+
"grad_norm": 0.3295337166593159,
|
| 801 |
+
"kl": 0.000766754150390625,
|
| 802 |
+
"learning_rate": 9.397114317029974e-07,
|
| 803 |
+
"loss": 0.0,
|
| 804 |
+
"num_tokens": 2810617.0,
|
| 805 |
+
"reward": 1.879860758781433,
|
| 806 |
+
"reward_std": 0.8230662941932678,
|
| 807 |
+
"rewards/accuracy_reward": 0.0,
|
| 808 |
+
"rewards/format_reward": 0.875,
|
| 809 |
+
"rewards/tag_count_reward": 1.004860758781433,
|
| 810 |
+
"step": 50
|
| 811 |
+
},
|
| 812 |
+
{
|
| 813 |
+
"clip_ratio": 0.0,
|
| 814 |
+
"completion_length": 1985.875,
|
| 815 |
+
"epoch": 0.0021763714340580793,
|
| 816 |
+
"grad_norm": 0.332025052005317,
|
| 817 |
+
"kl": 0.0007190704345703125,
|
| 818 |
+
"learning_rate": 9.357252853159505e-07,
|
| 819 |
+
"loss": 0.0128,
|
| 820 |
+
"num_tokens": 2880069.0,
|
| 821 |
+
"reward": 2.3542251586914062,
|
| 822 |
+
"reward_std": 1.0882834196090698,
|
| 823 |
+
"rewards/accuracy_reward": 0.1875,
|
| 824 |
+
"rewards/format_reward": 0.84375,
|
| 825 |
+
"rewards/tag_count_reward": 1.1354751586914062,
|
| 826 |
+
"step": 51
|
| 827 |
+
},
|
| 828 |
+
{
|
| 829 |
+
"clip_ratio": 0.0,
|
| 830 |
+
"completion_length": 2048.0,
|
| 831 |
+
"epoch": 0.0022190453837454926,
|
| 832 |
+
"grad_norm": 0.3675266539339998,
|
| 833 |
+
"kl": 0.0009326934814453125,
|
| 834 |
+
"learning_rate": 9.316216432703916e-07,
|
| 835 |
+
"loss": 0.0,
|
| 836 |
+
"num_tokens": 2950637.0,
|
| 837 |
+
"reward": 1.886525273323059,
|
| 838 |
+
"reward_std": 0.965412437915802,
|
| 839 |
+
"rewards/accuracy_reward": 0.0,
|
| 840 |
+
"rewards/format_reward": 0.8125,
|
| 841 |
+
"rewards/tag_count_reward": 1.0740252435207367,
|
| 842 |
+
"step": 52
|
| 843 |
+
},
|
| 844 |
+
{
|
| 845 |
+
"clip_ratio": 0.0,
|
| 846 |
+
"completion_length": 1954.4375,
|
| 847 |
+
"epoch": 0.002261719333432906,
|
| 848 |
+
"grad_norm": 0.3350129847310395,
|
| 849 |
+
"kl": 0.0007534027099609375,
|
| 850 |
+
"learning_rate": 9.274017555754407e-07,
|
| 851 |
+
"loss": -0.0006,
|
| 852 |
+
"num_tokens": 3019339.0,
|
| 853 |
+
"reward": 1.9703275561332703,
|
| 854 |
+
"reward_std": 0.944995105266571,
|
| 855 |
+
"rewards/accuracy_reward": 0.0,
|
| 856 |
+
"rewards/format_reward": 0.90625,
|
| 857 |
+
"rewards/tag_count_reward": 1.0640775859355927,
|
| 858 |
+
"step": 53
|
| 859 |
+
},
|
| 860 |
+
{
|
| 861 |
+
"clip_ratio": 0.0,
|
| 862 |
+
"completion_length": 2031.125,
|
| 863 |
+
"epoch": 0.0023043932831203194,
|
| 864 |
+
"grad_norm": 0.3652644651974555,
|
| 865 |
+
"kl": 0.0009059906005859375,
|
| 866 |
+
"learning_rate": 9.230669076497687e-07,
|
| 867 |
+
"loss": 0.0028,
|
| 868 |
+
"num_tokens": 3089999.0,
|
| 869 |
+
"reward": 2.2546380162239075,
|
| 870 |
+
"reward_std": 0.7995705902576447,
|
| 871 |
+
"rewards/accuracy_reward": 0.0625,
|
| 872 |
+
"rewards/format_reward": 0.9375,
|
| 873 |
+
"rewards/tag_count_reward": 1.192137897014618,
|
| 874 |
+
"step": 54
|
| 875 |
+
},
|
| 876 |
+
{
|
| 877 |
+
"clip_ratio": 0.0,
|
| 878 |
+
"completion_length": 2043.5,
|
| 879 |
+
"epoch": 0.0023470672328077327,
|
| 880 |
+
"grad_norm": 0.3072691210342616,
|
| 881 |
+
"kl": 0.0006389617919921875,
|
| 882 |
+
"learning_rate": 9.186184199300463e-07,
|
| 883 |
+
"loss": 0.0002,
|
| 884 |
+
"num_tokens": 3159847.0,
|
| 885 |
+
"reward": 2.0708298683166504,
|
| 886 |
+
"reward_std": 1.0467402040958405,
|
| 887 |
+
"rewards/accuracy_reward": 0.0,
|
| 888 |
+
"rewards/format_reward": 0.9375,
|
| 889 |
+
"rewards/tag_count_reward": 1.1333298683166504,
|
| 890 |
+
"step": 55
|
| 891 |
+
},
|
| 892 |
+
{
|
| 893 |
+
"clip_ratio": 0.0,
|
| 894 |
+
"completion_length": 1974.4375,
|
| 895 |
+
"epoch": 0.0023897411824951456,
|
| 896 |
+
"grad_norm": 0.3926736400093857,
|
| 897 |
+
"kl": 0.001049041748046875,
|
| 898 |
+
"learning_rate": 9.140576474687263e-07,
|
| 899 |
+
"loss": 0.0668,
|
| 900 |
+
"num_tokens": 3229557.0,
|
| 901 |
+
"reward": 1.809758186340332,
|
| 902 |
+
"reward_std": 0.7233555614948273,
|
| 903 |
+
"rewards/accuracy_reward": 0.0,
|
| 904 |
+
"rewards/format_reward": 0.96875,
|
| 905 |
+
"rewards/tag_count_reward": 0.8410081565380096,
|
| 906 |
+
"step": 56
|
| 907 |
+
},
|
| 908 |
+
{
|
| 909 |
+
"clip_ratio": 0.0,
|
| 910 |
+
"completion_length": 1960.84375,
|
| 911 |
+
"epoch": 0.002432415132182559,
|
| 912 |
+
"grad_norm": 0.2952096599022392,
|
| 913 |
+
"kl": 0.0006866455078125,
|
| 914 |
+
"learning_rate": 9.093859795212817e-07,
|
| 915 |
+
"loss": 0.0243,
|
| 916 |
+
"num_tokens": 3297232.0,
|
| 917 |
+
"reward": 2.242687225341797,
|
| 918 |
+
"reward_std": 0.8501459658145905,
|
| 919 |
+
"rewards/accuracy_reward": 0.0,
|
| 920 |
+
"rewards/format_reward": 0.8125,
|
| 921 |
+
"rewards/tag_count_reward": 1.430187165737152,
|
| 922 |
+
"step": 57
|
| 923 |
+
},
|
| 924 |
+
{
|
| 925 |
+
"clip_ratio": 0.0,
|
| 926 |
+
"completion_length": 1941.34375,
|
| 927 |
+
"epoch": 0.0024750890818699724,
|
| 928 |
+
"grad_norm": 0.37352162709303555,
|
| 929 |
+
"kl": 0.0009174346923828125,
|
| 930 |
+
"learning_rate": 9.046048391230247e-07,
|
| 931 |
+
"loss": 0.0412,
|
| 932 |
+
"num_tokens": 3365707.0,
|
| 933 |
+
"reward": 2.8109254837036133,
|
| 934 |
+
"reward_std": 1.9174948930740356,
|
| 935 |
+
"rewards/accuracy_reward": 0.21875,
|
| 936 |
+
"rewards/format_reward": 0.90625,
|
| 937 |
+
"rewards/tag_count_reward": 1.467175543308258,
|
| 938 |
+
"step": 58
|
| 939 |
+
},
|
| 940 |
+
{
|
| 941 |
+
"clip_ratio": 0.0,
|
| 942 |
+
"completion_length": 2030.625,
|
| 943 |
+
"epoch": 0.0025177630315573857,
|
| 944 |
+
"grad_norm": 0.34415114868612157,
|
| 945 |
+
"kl": 0.0008220672607421875,
|
| 946 |
+
"learning_rate": 8.997156826556369e-07,
|
| 947 |
+
"loss": 0.0082,
|
| 948 |
+
"num_tokens": 3436919.0,
|
| 949 |
+
"reward": 2.0681787729263306,
|
| 950 |
+
"reward_std": 1.2899426221847534,
|
| 951 |
+
"rewards/accuracy_reward": 0.0,
|
| 952 |
+
"rewards/format_reward": 0.8125,
|
| 953 |
+
"rewards/tag_count_reward": 1.2556788325309753,
|
| 954 |
+
"step": 59
|
| 955 |
+
},
|
| 956 |
+
{
|
| 957 |
+
"clip_ratio": 0.0,
|
| 958 |
+
"completion_length": 2043.9375,
|
| 959 |
+
"epoch": 0.002560436981244799,
|
| 960 |
+
"grad_norm": 0.354441246767313,
|
| 961 |
+
"kl": 0.000789642333984375,
|
| 962 |
+
"learning_rate": 8.9471999940354e-07,
|
| 963 |
+
"loss": 0.0013,
|
| 964 |
+
"num_tokens": 3508637.0,
|
| 965 |
+
"reward": 1.8495957851409912,
|
| 966 |
+
"reward_std": 1.155137300491333,
|
| 967 |
+
"rewards/accuracy_reward": 0.03125,
|
| 968 |
+
"rewards/format_reward": 0.84375,
|
| 969 |
+
"rewards/tag_count_reward": 0.9433457851409912,
|
| 970 |
+
"step": 60
|
| 971 |
+
},
|
| 972 |
+
{
|
| 973 |
+
"clip_ratio": 0.0,
|
| 974 |
+
"completion_length": 2048.0,
|
| 975 |
+
"epoch": 0.0026031109309322125,
|
| 976 |
+
"grad_norm": 0.3690724626179575,
|
| 977 |
+
"kl": 0.001007080078125,
|
| 978 |
+
"learning_rate": 8.896193111002475e-07,
|
| 979 |
+
"loss": 0.0,
|
| 980 |
+
"num_tokens": 3579461.0,
|
| 981 |
+
"reward": 2.2597023248672485,
|
| 982 |
+
"reward_std": 0.684966504573822,
|
| 983 |
+
"rewards/accuracy_reward": 0.0,
|
| 984 |
+
"rewards/format_reward": 0.96875,
|
| 985 |
+
"rewards/tag_count_reward": 1.2909523844718933,
|
| 986 |
+
"step": 61
|
| 987 |
+
},
|
| 988 |
+
{
|
| 989 |
+
"clip_ratio": 0.0,
|
| 990 |
+
"completion_length": 2048.0,
|
| 991 |
+
"epoch": 0.002645784880619626,
|
| 992 |
+
"grad_norm": 0.319211799092635,
|
| 993 |
+
"kl": 0.000873565673828125,
|
| 994 |
+
"learning_rate": 8.844151714648274e-07,
|
| 995 |
+
"loss": 0.0,
|
| 996 |
+
"num_tokens": 3653189.0,
|
| 997 |
+
"reward": 1.7180063724517822,
|
| 998 |
+
"reward_std": 0.7376870512962341,
|
| 999 |
+
"rewards/accuracy_reward": 0.0,
|
| 1000 |
+
"rewards/format_reward": 0.90625,
|
| 1001 |
+
"rewards/tag_count_reward": 0.8117563724517822,
|
| 1002 |
+
"step": 62
|
| 1003 |
+
},
|
| 1004 |
+
{
|
| 1005 |
+
"clip_ratio": 0.0,
|
| 1006 |
+
"completion_length": 2042.90625,
|
| 1007 |
+
"epoch": 0.002688458830307039,
|
| 1008 |
+
"grad_norm": 0.3136056310472147,
|
| 1009 |
+
"kl": 0.0007343292236328125,
|
| 1010 |
+
"learning_rate": 8.791091657286267e-07,
|
| 1011 |
+
"loss": 0.0019,
|
| 1012 |
+
"num_tokens": 3722690.0,
|
| 1013 |
+
"reward": 2.304230809211731,
|
| 1014 |
+
"reward_std": 1.4096248149871826,
|
| 1015 |
+
"rewards/accuracy_reward": 0.09375,
|
| 1016 |
+
"rewards/format_reward": 0.9375,
|
| 1017 |
+
"rewards/tag_count_reward": 1.1792307496070862,
|
| 1018 |
+
"step": 63
|
| 1019 |
+
},
|
| 1020 |
+
{
|
| 1021 |
+
"clip_ratio": 0.0,
|
| 1022 |
+
"completion_length": 2003.8125,
|
| 1023 |
+
"epoch": 0.0027311327799944525,
|
| 1024 |
+
"grad_norm": 0.3567969961663852,
|
| 1025 |
+
"kl": 0.0008449554443359375,
|
| 1026 |
+
"learning_rate": 8.737029101523929e-07,
|
| 1027 |
+
"loss": 0.0032,
|
| 1028 |
+
"num_tokens": 3791668.0,
|
| 1029 |
+
"reward": 2.601585865020752,
|
| 1030 |
+
"reward_std": 1.063989520072937,
|
| 1031 |
+
"rewards/accuracy_reward": 0.15625,
|
| 1032 |
+
"rewards/format_reward": 0.9375,
|
| 1033 |
+
"rewards/tag_count_reward": 1.3515858054161072,
|
| 1034 |
+
"step": 64
|
| 1035 |
+
},
|
| 1036 |
+
{
|
| 1037 |
+
"clip_ratio": 0.0,
|
| 1038 |
+
"completion_length": 2037.15625,
|
| 1039 |
+
"epoch": 0.002773806729681866,
|
| 1040 |
+
"grad_norm": 0.3179225934177875,
|
| 1041 |
+
"kl": 0.0008296966552734375,
|
| 1042 |
+
"learning_rate": 8.681980515339463e-07,
|
| 1043 |
+
"loss": 0.0,
|
| 1044 |
+
"num_tokens": 3862681.0,
|
| 1045 |
+
"reward": 1.5619869232177734,
|
| 1046 |
+
"reward_std": 1.4845786094665527,
|
| 1047 |
+
"rewards/accuracy_reward": 0.0,
|
| 1048 |
+
"rewards/format_reward": 0.8125,
|
| 1049 |
+
"rewards/tag_count_reward": 0.7494869530200958,
|
| 1050 |
+
"step": 65
|
| 1051 |
+
},
|
| 1052 |
+
{
|
| 1053 |
+
"clip_ratio": 0.0,
|
| 1054 |
+
"completion_length": 2048.0,
|
| 1055 |
+
"epoch": 0.002816480679369279,
|
| 1056 |
+
"grad_norm": 0.3446894269788866,
|
| 1057 |
+
"kl": 0.000942230224609375,
|
| 1058 |
+
"learning_rate": 8.625962667065487e-07,
|
| 1059 |
+
"loss": 0.0,
|
| 1060 |
+
"num_tokens": 3933537.0,
|
| 1061 |
+
"reward": 2.122675061225891,
|
| 1062 |
+
"reward_std": 0.49530309438705444,
|
| 1063 |
+
"rewards/accuracy_reward": 0.0,
|
| 1064 |
+
"rewards/format_reward": 1.0,
|
| 1065 |
+
"rewards/tag_count_reward": 1.1226750910282135,
|
| 1066 |
+
"step": 66
|
| 1067 |
+
},
|
| 1068 |
+
{
|
| 1069 |
+
"clip_ratio": 0.0,
|
| 1070 |
+
"completion_length": 1995.75,
|
| 1071 |
+
"epoch": 0.002859154629056692,
|
| 1072 |
+
"grad_norm": 0.3401240847454865,
|
| 1073 |
+
"kl": 0.0007991790771484375,
|
| 1074 |
+
"learning_rate": 8.568992620281243e-07,
|
| 1075 |
+
"loss": 0.0127,
|
| 1076 |
+
"num_tokens": 4003729.0,
|
| 1077 |
+
"reward": 2.3472211360931396,
|
| 1078 |
+
"reward_std": 1.199817717075348,
|
| 1079 |
+
"rewards/accuracy_reward": 0.03125,
|
| 1080 |
+
"rewards/format_reward": 0.90625,
|
| 1081 |
+
"rewards/tag_count_reward": 1.3784711956977844,
|
| 1082 |
+
"step": 67
|
| 1083 |
+
},
|
| 1084 |
+
{
|
| 1085 |
+
"clip_ratio": 0.0,
|
| 1086 |
+
"completion_length": 1873.875,
|
| 1087 |
+
"epoch": 0.0029018285787441056,
|
| 1088 |
+
"grad_norm": 0.3346647288884505,
|
| 1089 |
+
"kl": 0.0007038116455078125,
|
| 1090 |
+
"learning_rate": 8.511087728614862e-07,
|
| 1091 |
+
"loss": 0.0185,
|
| 1092 |
+
"num_tokens": 4068509.0,
|
| 1093 |
+
"reward": 3.047283172607422,
|
| 1094 |
+
"reward_std": 1.1735801696777344,
|
| 1095 |
+
"rewards/accuracy_reward": 0.09375,
|
| 1096 |
+
"rewards/format_reward": 0.9375,
|
| 1097 |
+
"rewards/tag_count_reward": 1.9222831726074219,
|
| 1098 |
+
"step": 68
|
| 1099 |
+
},
|
| 1100 |
+
{
|
| 1101 |
+
"clip_ratio": 0.0,
|
| 1102 |
+
"completion_length": 2048.0,
|
| 1103 |
+
"epoch": 0.002944502528431519,
|
| 1104 |
+
"grad_norm": 0.3311836838358215,
|
| 1105 |
+
"kl": 0.000804901123046875,
|
| 1106 |
+
"learning_rate": 8.452265630457282e-07,
|
| 1107 |
+
"loss": 0.0,
|
| 1108 |
+
"num_tokens": 4139301.0,
|
| 1109 |
+
"reward": 2.0747729539871216,
|
| 1110 |
+
"reward_std": 1.001867413520813,
|
| 1111 |
+
"rewards/accuracy_reward": 0.0,
|
| 1112 |
+
"rewards/format_reward": 0.84375,
|
| 1113 |
+
"rewards/tag_count_reward": 1.2310230731964111,
|
| 1114 |
+
"step": 69
|
| 1115 |
+
},
|
| 1116 |
+
{
|
| 1117 |
+
"clip_ratio": 0.0,
|
| 1118 |
+
"completion_length": 1844.28125,
|
| 1119 |
+
"epoch": 0.0029871764781189323,
|
| 1120 |
+
"grad_norm": 0.38403397458473654,
|
| 1121 |
+
"kl": 0.000957489013671875,
|
| 1122 |
+
"learning_rate": 8.392544243589427e-07,
|
| 1123 |
+
"loss": 0.0278,
|
| 1124 |
+
"num_tokens": 4204470.0,
|
| 1125 |
+
"reward": 2.535085082054138,
|
| 1126 |
+
"reward_std": 0.9629121422767639,
|
| 1127 |
+
"rewards/accuracy_reward": 0.28125,
|
| 1128 |
+
"rewards/format_reward": 0.9375,
|
| 1129 |
+
"rewards/tag_count_reward": 1.0350851118564606,
|
| 1130 |
+
"step": 70
|
| 1131 |
+
},
|
| 1132 |
+
{
|
| 1133 |
+
"clip_ratio": 0.0,
|
| 1134 |
+
"completion_length": 1806.25,
|
| 1135 |
+
"epoch": 0.0030298504278063456,
|
| 1136 |
+
"grad_norm": 0.39733773475712775,
|
| 1137 |
+
"kl": 0.00092315673828125,
|
| 1138 |
+
"learning_rate": 8.331941759724268e-07,
|
| 1139 |
+
"loss": 0.0488,
|
| 1140 |
+
"num_tokens": 4268510.0,
|
| 1141 |
+
"reward": 2.49297297000885,
|
| 1142 |
+
"reward_std": 0.6394779086112976,
|
| 1143 |
+
"rewards/accuracy_reward": 0.0,
|
| 1144 |
+
"rewards/format_reward": 1.0,
|
| 1145 |
+
"rewards/tag_count_reward": 1.49297297000885,
|
| 1146 |
+
"step": 71
|
| 1147 |
+
},
|
| 1148 |
+
{
|
| 1149 |
+
"clip_ratio": 0.0,
|
| 1150 |
+
"completion_length": 1907.09375,
|
| 1151 |
+
"epoch": 0.003072524377493759,
|
| 1152 |
+
"grad_norm": 0.34816532127406485,
|
| 1153 |
+
"kl": 0.000858306884765625,
|
| 1154 |
+
"learning_rate": 8.270476638965461e-07,
|
| 1155 |
+
"loss": -0.0105,
|
| 1156 |
+
"num_tokens": 4334433.0,
|
| 1157 |
+
"reward": 2.245678961277008,
|
| 1158 |
+
"reward_std": 1.0141153633594513,
|
| 1159 |
+
"rewards/accuracy_reward": 0.03125,
|
| 1160 |
+
"rewards/format_reward": 0.90625,
|
| 1161 |
+
"rewards/tag_count_reward": 1.276928961277008,
|
| 1162 |
+
"step": 72
|
| 1163 |
+
},
|
| 1164 |
+
{
|
| 1165 |
+
"clip_ratio": 0.0,
|
| 1166 |
+
"completion_length": 1901.46875,
|
| 1167 |
+
"epoch": 0.0031151983271811724,
|
| 1168 |
+
"grad_norm": 0.3303035408232146,
|
| 1169 |
+
"kl": 0.0007266998291015625,
|
| 1170 |
+
"learning_rate": 8.208167604184217e-07,
|
| 1171 |
+
"loss": -0.0037,
|
| 1172 |
+
"num_tokens": 4400344.0,
|
| 1173 |
+
"reward": 2.3381131887435913,
|
| 1174 |
+
"reward_std": 0.9651293754577637,
|
| 1175 |
+
"rewards/accuracy_reward": 0.03125,
|
| 1176 |
+
"rewards/format_reward": 0.90625,
|
| 1177 |
+
"rewards/tag_count_reward": 1.3693632185459137,
|
| 1178 |
+
"step": 73
|
| 1179 |
+
},
|
| 1180 |
+
{
|
| 1181 |
+
"clip_ratio": 0.0,
|
| 1182 |
+
"completion_length": 2048.0,
|
| 1183 |
+
"epoch": 0.0031578722768685857,
|
| 1184 |
+
"grad_norm": 0.33886215689095045,
|
| 1185 |
+
"kl": 0.0007953643798828125,
|
| 1186 |
+
"learning_rate": 8.145033635316128e-07,
|
| 1187 |
+
"loss": 0.0,
|
| 1188 |
+
"num_tokens": 4471248.0,
|
| 1189 |
+
"reward": 1.802172064781189,
|
| 1190 |
+
"reward_std": 0.9692512154579163,
|
| 1191 |
+
"rewards/accuracy_reward": 0.0,
|
| 1192 |
+
"rewards/format_reward": 0.875,
|
| 1193 |
+
"rewards/tag_count_reward": 0.927172064781189,
|
| 1194 |
+
"step": 74
|
| 1195 |
+
},
|
| 1196 |
+
{
|
| 1197 |
+
"clip_ratio": 0.0,
|
| 1198 |
+
"completion_length": 1888.8125,
|
| 1199 |
+
"epoch": 0.003200546226555999,
|
| 1200 |
+
"grad_norm": 0.3914547821892524,
|
| 1201 |
+
"kl": 0.00093841552734375,
|
| 1202 |
+
"learning_rate": 8.081093963579707e-07,
|
| 1203 |
+
"loss": -0.0462,
|
| 1204 |
+
"num_tokens": 4537266.0,
|
| 1205 |
+
"reward": 2.3257221579551697,
|
| 1206 |
+
"reward_std": 0.9169666767120361,
|
| 1207 |
+
"rewards/accuracy_reward": 0.0,
|
| 1208 |
+
"rewards/format_reward": 0.84375,
|
| 1209 |
+
"rewards/tag_count_reward": 1.4819721579551697,
|
| 1210 |
+
"step": 75
|
| 1211 |
+
},
|
| 1212 |
+
{
|
| 1213 |
+
"clip_ratio": 0.0,
|
| 1214 |
+
"completion_length": 2048.0,
|
| 1215 |
+
"epoch": 0.003243220176243412,
|
| 1216 |
+
"grad_norm": 0.381731961030959,
|
| 1217 |
+
"kl": 0.0010223388671875,
|
| 1218 |
+
"learning_rate": 8.01636806561836e-07,
|
| 1219 |
+
"loss": 0.0,
|
| 1220 |
+
"num_tokens": 4609082.0,
|
| 1221 |
+
"reward": 2.0226646065711975,
|
| 1222 |
+
"reward_std": 1.046322524547577,
|
| 1223 |
+
"rewards/accuracy_reward": 0.0,
|
| 1224 |
+
"rewards/format_reward": 0.875,
|
| 1225 |
+
"rewards/tag_count_reward": 1.1476646065711975,
|
| 1226 |
+
"step": 76
|
| 1227 |
+
},
|
| 1228 |
+
{
|
| 1229 |
+
"clip_ratio": 0.0,
|
| 1230 |
+
"completion_length": 2027.8125,
|
| 1231 |
+
"epoch": 0.0032858941259308254,
|
| 1232 |
+
"grad_norm": 0.3271819003849751,
|
| 1233 |
+
"kl": 0.000823974609375,
|
| 1234 |
+
"learning_rate": 7.950875657567621e-07,
|
| 1235 |
+
"loss": 0.0081,
|
| 1236 |
+
"num_tokens": 4680260.0,
|
| 1237 |
+
"reward": 2.094830870628357,
|
| 1238 |
+
"reward_std": 1.0302011966705322,
|
| 1239 |
+
"rewards/accuracy_reward": 0.0,
|
| 1240 |
+
"rewards/format_reward": 0.84375,
|
| 1241 |
+
"rewards/tag_count_reward": 1.2510809302330017,
|
| 1242 |
+
"step": 77
|
| 1243 |
+
},
|
| 1244 |
+
{
|
| 1245 |
+
"clip_ratio": 0.0,
|
| 1246 |
+
"completion_length": 1988.9375,
|
| 1247 |
+
"epoch": 0.0033285680756182387,
|
| 1248 |
+
"grad_norm": 0.340508206278311,
|
| 1249 |
+
"kl": 0.0008754730224609375,
|
| 1250 |
+
"learning_rate": 7.884636689049422e-07,
|
| 1251 |
+
"loss": 0.0158,
|
| 1252 |
+
"num_tokens": 4748922.0,
|
| 1253 |
+
"reward": 1.8495043516159058,
|
| 1254 |
+
"reward_std": 1.125253975391388,
|
| 1255 |
+
"rewards/accuracy_reward": 0.0,
|
| 1256 |
+
"rewards/format_reward": 0.96875,
|
| 1257 |
+
"rewards/tag_count_reward": 0.8807543814182281,
|
| 1258 |
+
"step": 78
|
| 1259 |
+
},
|
| 1260 |
+
{
|
| 1261 |
+
"clip_ratio": 0.0,
|
| 1262 |
+
"completion_length": 1895.25,
|
| 1263 |
+
"epoch": 0.003371242025305652,
|
| 1264 |
+
"grad_norm": 0.3278835248117013,
|
| 1265 |
+
"kl": 0.0008296966552734375,
|
| 1266 |
+
"learning_rate": 7.817671337095244e-07,
|
| 1267 |
+
"loss": 0.0081,
|
| 1268 |
+
"num_tokens": 4814522.0,
|
| 1269 |
+
"reward": 2.285441756248474,
|
| 1270 |
+
"reward_std": 0.9086178541183472,
|
| 1271 |
+
"rewards/accuracy_reward": 0.03125,
|
| 1272 |
+
"rewards/format_reward": 0.875,
|
| 1273 |
+
"rewards/tag_count_reward": 1.3479417562484741,
|
| 1274 |
+
"step": 79
|
| 1275 |
+
},
|
| 1276 |
+
{
|
| 1277 |
+
"clip_ratio": 0.0,
|
| 1278 |
+
"completion_length": 2000.09375,
|
| 1279 |
+
"epoch": 0.0034139159749930655,
|
| 1280 |
+
"grad_norm": 0.3136954958746447,
|
| 1281 |
+
"kl": 0.0007648468017578125,
|
| 1282 |
+
"learning_rate": 7.75e-07,
|
| 1283 |
+
"loss": 0.0023,
|
| 1284 |
+
"num_tokens": 4883885.0,
|
| 1285 |
+
"reward": 2.512591600418091,
|
| 1286 |
+
"reward_std": 0.8996666073799133,
|
| 1287 |
+
"rewards/accuracy_reward": 0.03125,
|
| 1288 |
+
"rewards/format_reward": 0.84375,
|
| 1289 |
+
"rewards/tag_count_reward": 1.6063416004180908,
|
| 1290 |
+
"step": 80
|
| 1291 |
+
},
|
| 1292 |
+
{
|
| 1293 |
+
"clip_ratio": 0.0,
|
| 1294 |
+
"completion_length": 1940.8125,
|
| 1295 |
+
"epoch": 0.003456589924680479,
|
| 1296 |
+
"grad_norm": 0.33155015251771963,
|
| 1297 |
+
"kl": 0.0007305145263671875,
|
| 1298 |
+
"learning_rate": 7.681643291108517e-07,
|
| 1299 |
+
"loss": -0.021,
|
| 1300 |
+
"num_tokens": 4952439.0,
|
| 1301 |
+
"reward": 2.311286687850952,
|
| 1302 |
+
"reward_std": 1.0780829191207886,
|
| 1303 |
+
"rewards/accuracy_reward": 0.0,
|
| 1304 |
+
"rewards/format_reward": 0.9375,
|
| 1305 |
+
"rewards/tag_count_reward": 1.3737868070602417,
|
| 1306 |
+
"step": 81
|
| 1307 |
+
},
|
| 1308 |
+
{
|
| 1309 |
+
"clip_ratio": 0.0,
|
| 1310 |
+
"completion_length": 2000.03125,
|
| 1311 |
+
"epoch": 0.003499263874367892,
|
| 1312 |
+
"grad_norm": 0.3279242059366365,
|
| 1313 |
+
"kl": 0.000835418701171875,
|
| 1314 |
+
"learning_rate": 7.612622032536507e-07,
|
| 1315 |
+
"loss": 0.0366,
|
| 1316 |
+
"num_tokens": 5021952.0,
|
| 1317 |
+
"reward": 2.0096482038497925,
|
| 1318 |
+
"reward_std": 0.8550244867801666,
|
| 1319 |
+
"rewards/accuracy_reward": 0.0,
|
| 1320 |
+
"rewards/format_reward": 0.90625,
|
| 1321 |
+
"rewards/tag_count_reward": 1.1033982038497925,
|
| 1322 |
+
"step": 82
|
| 1323 |
+
},
|
| 1324 |
+
{
|
| 1325 |
+
"clip_ratio": 0.0,
|
| 1326 |
+
"completion_length": 1984.09375,
|
| 1327 |
+
"epoch": 0.0035419378240553055,
|
| 1328 |
+
"grad_norm": 0.3203554063868121,
|
| 1329 |
+
"kl": 0.000823974609375,
|
| 1330 |
+
"learning_rate": 7.54295724882796e-07,
|
| 1331 |
+
"loss": 0.0214,
|
| 1332 |
+
"num_tokens": 5090267.0,
|
| 1333 |
+
"reward": 2.522105097770691,
|
| 1334 |
+
"reward_std": 1.547502875328064,
|
| 1335 |
+
"rewards/accuracy_reward": 0.15625,
|
| 1336 |
+
"rewards/format_reward": 0.84375,
|
| 1337 |
+
"rewards/tag_count_reward": 1.3658550381660461,
|
| 1338 |
+
"step": 83
|
| 1339 |
+
},
|
| 1340 |
+
{
|
| 1341 |
+
"clip_ratio": 0.0,
|
| 1342 |
+
"completion_length": 2045.0,
|
| 1343 |
+
"epoch": 0.003584611773742719,
|
| 1344 |
+
"grad_norm": 0.3036283525899871,
|
| 1345 |
+
"kl": 0.0007343292236328125,
|
| 1346 |
+
"learning_rate": 7.472670160550848e-07,
|
| 1347 |
+
"loss": 0.0006,
|
| 1348 |
+
"num_tokens": 5161203.0,
|
| 1349 |
+
"reward": 2.195971965789795,
|
| 1350 |
+
"reward_std": 1.3376309871673584,
|
| 1351 |
+
"rewards/accuracy_reward": 0.03125,
|
| 1352 |
+
"rewards/format_reward": 0.875,
|
| 1353 |
+
"rewards/tag_count_reward": 1.258471965789795,
|
| 1354 |
+
"step": 84
|
| 1355 |
+
},
|
| 1356 |
+
{
|
| 1357 |
+
"clip_ratio": 0.0,
|
| 1358 |
+
"completion_length": 2048.0,
|
| 1359 |
+
"epoch": 0.0036272857234301323,
|
| 1360 |
+
"grad_norm": 0.37283288943704124,
|
| 1361 |
+
"kl": 0.00099945068359375,
|
| 1362 |
+
"learning_rate": 7.401782177833147e-07,
|
| 1363 |
+
"loss": 0.0,
|
| 1364 |
+
"num_tokens": 5233043.0,
|
| 1365 |
+
"reward": 1.9839001893997192,
|
| 1366 |
+
"reward_std": 0.9118372201919556,
|
| 1367 |
+
"rewards/accuracy_reward": 0.0,
|
| 1368 |
+
"rewards/format_reward": 0.9375,
|
| 1369 |
+
"rewards/tag_count_reward": 1.0464001595973969,
|
| 1370 |
+
"step": 85
|
| 1371 |
+
},
|
| 1372 |
+
{
|
| 1373 |
+
"clip_ratio": 0.0,
|
| 1374 |
+
"completion_length": 2048.0,
|
| 1375 |
+
"epoch": 0.003669959673117545,
|
| 1376 |
+
"grad_norm": 0.3456534169257094,
|
| 1377 |
+
"kl": 0.0008678436279296875,
|
| 1378 |
+
"learning_rate": 7.330314893841101e-07,
|
| 1379 |
+
"loss": 0.0,
|
| 1380 |
+
"num_tokens": 5305531.0,
|
| 1381 |
+
"reward": 1.626398503780365,
|
| 1382 |
+
"reward_std": 1.0243891477584839,
|
| 1383 |
+
"rewards/accuracy_reward": 0.0,
|
| 1384 |
+
"rewards/format_reward": 0.8125,
|
| 1385 |
+
"rewards/tag_count_reward": 0.8138984739780426,
|
| 1386 |
+
"step": 86
|
| 1387 |
+
},
|
| 1388 |
+
{
|
| 1389 |
+
"clip_ratio": 0.0,
|
| 1390 |
+
"completion_length": 1673.4375,
|
| 1391 |
+
"epoch": 0.0037126336228049586,
|
| 1392 |
+
"grad_norm": 0.4669593168506545,
|
| 1393 |
+
"kl": 0.0009441375732421875,
|
| 1394 |
+
"learning_rate": 7.258290078201731e-07,
|
| 1395 |
+
"loss": -0.0258,
|
| 1396 |
+
"num_tokens": 5364001.0,
|
| 1397 |
+
"reward": 2.214863419532776,
|
| 1398 |
+
"reward_std": 1.1067255139350891,
|
| 1399 |
+
"rewards/accuracy_reward": 0.0,
|
| 1400 |
+
"rewards/format_reward": 0.9375,
|
| 1401 |
+
"rewards/tag_count_reward": 1.2773634493350983,
|
| 1402 |
+
"step": 87
|
| 1403 |
+
},
|
| 1404 |
+
{
|
| 1405 |
+
"clip_ratio": 0.0,
|
| 1406 |
+
"completion_length": 2048.0,
|
| 1407 |
+
"epoch": 0.003755307572492372,
|
| 1408 |
+
"grad_norm": 0.31210218958098757,
|
| 1409 |
+
"kl": 0.000820159912109375,
|
| 1410 |
+
"learning_rate": 7.185729670371604e-07,
|
| 1411 |
+
"loss": 0.0,
|
| 1412 |
+
"num_tokens": 5434809.0,
|
| 1413 |
+
"reward": 2.4927505254745483,
|
| 1414 |
+
"reward_std": 1.3483699560165405,
|
| 1415 |
+
"rewards/accuracy_reward": 0.0,
|
| 1416 |
+
"rewards/format_reward": 0.875,
|
| 1417 |
+
"rewards/tag_count_reward": 1.6177505254745483,
|
| 1418 |
+
"step": 88
|
| 1419 |
+
},
|
| 1420 |
+
{
|
| 1421 |
+
"clip_ratio": 0.0,
|
| 1422 |
+
"completion_length": 2047.96875,
|
| 1423 |
+
"epoch": 0.0037979815221797853,
|
| 1424 |
+
"grad_norm": 0.3380404436952231,
|
| 1425 |
+
"kl": 0.0008831024169921875,
|
| 1426 |
+
"learning_rate": 7.11265577295385e-07,
|
| 1427 |
+
"loss": 0.0001,
|
| 1428 |
+
"num_tokens": 5504720.0,
|
| 1429 |
+
"reward": 2.245384097099304,
|
| 1430 |
+
"reward_std": 0.7598803043365479,
|
| 1431 |
+
"rewards/accuracy_reward": 0.0625,
|
| 1432 |
+
"rewards/format_reward": 0.96875,
|
| 1433 |
+
"rewards/tag_count_reward": 1.1516340374946594,
|
| 1434 |
+
"step": 89
|
| 1435 |
+
},
|
| 1436 |
+
{
|
| 1437 |
+
"clip_ratio": 0.0,
|
| 1438 |
+
"completion_length": 2048.0,
|
| 1439 |
+
"epoch": 0.0038406554718671986,
|
| 1440 |
+
"grad_norm": 0.35711870616336955,
|
| 1441 |
+
"kl": 0.0009098052978515625,
|
| 1442 |
+
"learning_rate": 7.039090644965509e-07,
|
| 1443 |
+
"loss": 0.0,
|
| 1444 |
+
"num_tokens": 5576592.0,
|
| 1445 |
+
"reward": 1.8503382205963135,
|
| 1446 |
+
"reward_std": 1.4265711307525635,
|
| 1447 |
+
"rewards/accuracy_reward": 0.0,
|
| 1448 |
+
"rewards/format_reward": 0.8125,
|
| 1449 |
+
"rewards/tag_count_reward": 1.037838190793991,
|
| 1450 |
+
"step": 90
|
| 1451 |
+
},
|
| 1452 |
+
{
|
| 1453 |
+
"clip_ratio": 0.0,
|
| 1454 |
+
"completion_length": 1829.4375,
|
| 1455 |
+
"epoch": 0.003883329421554612,
|
| 1456 |
+
"grad_norm": 0.3528272586128919,
|
| 1457 |
+
"kl": 0.00090789794921875,
|
| 1458 |
+
"learning_rate": 6.965056695057204e-07,
|
| 1459 |
+
"loss": 0.0401,
|
| 1460 |
+
"num_tokens": 5640054.0,
|
| 1461 |
+
"reward": 3.108765721321106,
|
| 1462 |
+
"reward_std": 0.8710709810256958,
|
| 1463 |
+
"rewards/accuracy_reward": 0.25,
|
| 1464 |
+
"rewards/format_reward": 0.96875,
|
| 1465 |
+
"rewards/tag_count_reward": 1.6400156915187836,
|
| 1466 |
+
"step": 91
|
| 1467 |
+
},
|
| 1468 |
+
{
|
| 1469 |
+
"clip_ratio": 0.0,
|
| 1470 |
+
"completion_length": 2048.0,
|
| 1471 |
+
"epoch": 0.003926003371242025,
|
| 1472 |
+
"grad_norm": 0.28270593413166484,
|
| 1473 |
+
"kl": 0.000820159912109375,
|
| 1474 |
+
"learning_rate": 6.890576474687263e-07,
|
| 1475 |
+
"loss": 0.0,
|
| 1476 |
+
"num_tokens": 5711894.0,
|
| 1477 |
+
"reward": 1.9772083759307861,
|
| 1478 |
+
"reward_std": 1.192688226699829,
|
| 1479 |
+
"rewards/accuracy_reward": 0.0,
|
| 1480 |
+
"rewards/format_reward": 0.84375,
|
| 1481 |
+
"rewards/tag_count_reward": 1.1334583163261414,
|
| 1482 |
+
"step": 92
|
| 1483 |
+
},
|
| 1484 |
+
{
|
| 1485 |
+
"clip_ratio": 0.0,
|
| 1486 |
+
"completion_length": 1996.6875,
|
| 1487 |
+
"epoch": 0.003968677320929439,
|
| 1488 |
+
"grad_norm": 0.32204718393920595,
|
| 1489 |
+
"kl": 0.0007877349853515625,
|
| 1490 |
+
"learning_rate": 6.815672671252315e-07,
|
| 1491 |
+
"loss": 0.0169,
|
| 1492 |
+
"num_tokens": 5781164.0,
|
| 1493 |
+
"reward": 2.3092033863067627,
|
| 1494 |
+
"reward_std": 1.3056998252868652,
|
| 1495 |
+
"rewards/accuracy_reward": 0.0,
|
| 1496 |
+
"rewards/format_reward": 0.8125,
|
| 1497 |
+
"rewards/tag_count_reward": 1.496703326702118,
|
| 1498 |
+
"step": 93
|
| 1499 |
+
},
|
| 1500 |
+
{
|
| 1501 |
+
"clip_ratio": 0.0,
|
| 1502 |
+
"completion_length": 2039.28125,
|
| 1503 |
+
"epoch": 0.004011351270616852,
|
| 1504 |
+
"grad_norm": 0.37035967757319893,
|
| 1505 |
+
"kl": 0.001003265380859375,
|
| 1506 |
+
"learning_rate": 6.740368101176495e-07,
|
| 1507 |
+
"loss": 0.0072,
|
| 1508 |
+
"num_tokens": 5853045.0,
|
| 1509 |
+
"reward": 1.7449817061424255,
|
| 1510 |
+
"reward_std": 1.071380853652954,
|
| 1511 |
+
"rewards/accuracy_reward": 0.0625,
|
| 1512 |
+
"rewards/format_reward": 0.875,
|
| 1513 |
+
"rewards/tag_count_reward": 0.7449816763401031,
|
| 1514 |
+
"step": 94
|
| 1515 |
+
},
|
| 1516 |
+
{
|
| 1517 |
+
"clip_ratio": 0.0,
|
| 1518 |
+
"completion_length": 2048.0,
|
| 1519 |
+
"epoch": 0.0040540252203042654,
|
| 1520 |
+
"grad_norm": 0.34550850811835476,
|
| 1521 |
+
"kl": 0.00098419189453125,
|
| 1522 |
+
"learning_rate": 6.664685702961344e-07,
|
| 1523 |
+
"loss": 0.0,
|
| 1524 |
+
"num_tokens": 5922517.0,
|
| 1525 |
+
"reward": 2.044236481189728,
|
| 1526 |
+
"reward_std": 1.2233920097351074,
|
| 1527 |
+
"rewards/accuracy_reward": 0.0,
|
| 1528 |
+
"rewards/format_reward": 0.875,
|
| 1529 |
+
"rewards/tag_count_reward": 1.1692364513874054,
|
| 1530 |
+
"step": 95
|
| 1531 |
+
},
|
| 1532 |
+
{
|
| 1533 |
+
"clip_ratio": 0.0,
|
| 1534 |
+
"completion_length": 2048.0,
|
| 1535 |
+
"epoch": 0.004096699169991678,
|
| 1536 |
+
"grad_norm": 0.36594816251267936,
|
| 1537 |
+
"kl": 0.0009975433349609375,
|
| 1538 |
+
"learning_rate": 6.588648530198504e-07,
|
| 1539 |
+
"loss": 0.0,
|
| 1540 |
+
"num_tokens": 5993805.0,
|
| 1541 |
+
"reward": 1.9244810938835144,
|
| 1542 |
+
"reward_std": 0.8103159219026566,
|
| 1543 |
+
"rewards/accuracy_reward": 0.0,
|
| 1544 |
+
"rewards/format_reward": 0.9375,
|
| 1545 |
+
"rewards/tag_count_reward": 0.986981064081192,
|
| 1546 |
+
"step": 96
|
| 1547 |
+
},
|
| 1548 |
+
{
|
| 1549 |
+
"clip_ratio": 0.0,
|
| 1550 |
+
"completion_length": 2037.78125,
|
| 1551 |
+
"epoch": 0.004139373119679092,
|
| 1552 |
+
"grad_norm": 0.3406609265579752,
|
| 1553 |
+
"kl": 0.000988006591796875,
|
| 1554 |
+
"learning_rate": 6.512279744547392e-07,
|
| 1555 |
+
"loss": 0.0061,
|
| 1556 |
+
"num_tokens": 6065622.0,
|
| 1557 |
+
"reward": 2.0064436197280884,
|
| 1558 |
+
"reward_std": 1.1249323785305023,
|
| 1559 |
+
"rewards/accuracy_reward": 0.03125,
|
| 1560 |
+
"rewards/format_reward": 0.90625,
|
| 1561 |
+
"rewards/tag_count_reward": 1.0376937091350555,
|
| 1562 |
+
"step": 97
|
| 1563 |
+
},
|
| 1564 |
+
{
|
| 1565 |
+
"clip_ratio": 0.0,
|
| 1566 |
+
"completion_length": 2048.0,
|
| 1567 |
+
"epoch": 0.004182047069366505,
|
| 1568 |
+
"grad_norm": 0.34408636972411044,
|
| 1569 |
+
"kl": 0.0008792877197265625,
|
| 1570 |
+
"learning_rate": 6.435602608679916e-07,
|
| 1571 |
+
"loss": 0.0,
|
| 1572 |
+
"num_tokens": 6136350.0,
|
| 1573 |
+
"reward": 2.4079575538635254,
|
| 1574 |
+
"reward_std": 1.2895533442497253,
|
| 1575 |
+
"rewards/accuracy_reward": 0.09375,
|
| 1576 |
+
"rewards/format_reward": 0.90625,
|
| 1577 |
+
"rewards/tag_count_reward": 1.3142073154449463,
|
| 1578 |
+
"step": 98
|
| 1579 |
+
},
|
| 1580 |
+
{
|
| 1581 |
+
"clip_ratio": 0.0,
|
| 1582 |
+
"completion_length": 1946.875,
|
| 1583 |
+
"epoch": 0.004224721019053919,
|
| 1584 |
+
"grad_norm": 0.35206904845951326,
|
| 1585 |
+
"kl": 0.00102996826171875,
|
| 1586 |
+
"learning_rate": 6.358640479194451e-07,
|
| 1587 |
+
"loss": 0.0073,
|
| 1588 |
+
"num_tokens": 6203834.0,
|
| 1589 |
+
"reward": 2.830492854118347,
|
| 1590 |
+
"reward_std": 1.020926684141159,
|
| 1591 |
+
"rewards/accuracy_reward": 0.21875,
|
| 1592 |
+
"rewards/format_reward": 0.9375,
|
| 1593 |
+
"rewards/tag_count_reward": 1.4554929733276367,
|
| 1594 |
+
"step": 99
|
| 1595 |
+
},
|
| 1596 |
+
{
|
| 1597 |
+
"clip_ratio": 0.0,
|
| 1598 |
+
"completion_length": 2048.0,
|
| 1599 |
+
"epoch": 0.004267394968741332,
|
| 1600 |
+
"grad_norm": 0.36235059012611426,
|
| 1601 |
+
"kl": 0.001010894775390625,
|
| 1602 |
+
"learning_rate": 6.281416799501187e-07,
|
| 1603 |
+
"loss": 0.0,
|
| 1604 |
+
"num_tokens": 6274330.0,
|
| 1605 |
+
"reward": 1.618079423904419,
|
| 1606 |
+
"reward_std": 1.1985960006713867,
|
| 1607 |
+
"rewards/accuracy_reward": 0.0,
|
| 1608 |
+
"rewards/format_reward": 0.84375,
|
| 1609 |
+
"rewards/tag_count_reward": 0.774329423904419,
|
| 1610 |
+
"step": 100
|
| 1611 |
+
},
|
| 1612 |
+
{
|
| 1613 |
+
"clip_ratio": 0.0,
|
| 1614 |
+
"completion_length": 2020.03125,
|
| 1615 |
+
"epoch": 0.004310068918428745,
|
| 1616 |
+
"grad_norm": 0.3368840721033306,
|
| 1617 |
+
"kl": 0.0009517669677734375,
|
| 1618 |
+
"learning_rate": 6.203955092681039e-07,
|
| 1619 |
+
"loss": -0.005,
|
| 1620 |
+
"num_tokens": 6346171.0,
|
| 1621 |
+
"reward": 2.3748244643211365,
|
| 1622 |
+
"reward_std": 1.0288445353507996,
|
| 1623 |
+
"rewards/accuracy_reward": 0.21875,
|
| 1624 |
+
"rewards/format_reward": 0.90625,
|
| 1625 |
+
"rewards/tag_count_reward": 1.0310744643211365,
|
| 1626 |
+
"step": 101
|
| 1627 |
+
},
|
| 1628 |
+
{
|
| 1629 |
+
"clip_ratio": 0.0,
|
| 1630 |
+
"completion_length": 1889.5,
|
| 1631 |
+
"epoch": 0.0043527428681161585,
|
| 1632 |
+
"grad_norm": 0.337924446778461,
|
| 1633 |
+
"kl": 0.0008182525634765625,
|
| 1634 |
+
"learning_rate": 6.126278954320294e-07,
|
| 1635 |
+
"loss": 0.0146,
|
| 1636 |
+
"num_tokens": 6411795.0,
|
| 1637 |
+
"reward": 2.6838208436965942,
|
| 1638 |
+
"reward_std": 1.266103446483612,
|
| 1639 |
+
"rewards/accuracy_reward": 0.28125,
|
| 1640 |
+
"rewards/format_reward": 0.90625,
|
| 1641 |
+
"rewards/tag_count_reward": 1.2150707840919495,
|
| 1642 |
+
"step": 102
|
| 1643 |
+
},
|
| 1644 |
+
{
|
| 1645 |
+
"clip_ratio": 0.0,
|
| 1646 |
+
"completion_length": 1889.28125,
|
| 1647 |
+
"epoch": 0.0043954168178035715,
|
| 1648 |
+
"grad_norm": 0.4071185934051508,
|
| 1649 |
+
"kl": 0.001194000244140625,
|
| 1650 |
+
"learning_rate": 6.048412045323164e-07,
|
| 1651 |
+
"loss": 0.0216,
|
| 1652 |
+
"num_tokens": 6478500.0,
|
| 1653 |
+
"reward": 2.569925904273987,
|
| 1654 |
+
"reward_std": 1.1854565739631653,
|
| 1655 |
+
"rewards/accuracy_reward": 0.15625,
|
| 1656 |
+
"rewards/format_reward": 0.875,
|
| 1657 |
+
"rewards/tag_count_reward": 1.3824260234832764,
|
| 1658 |
+
"step": 103
|
| 1659 |
+
},
|
| 1660 |
+
{
|
| 1661 |
+
"clip_ratio": 0.0,
|
| 1662 |
+
"completion_length": 1877.90625,
|
| 1663 |
+
"epoch": 0.004438090767490985,
|
| 1664 |
+
"grad_norm": 0.38448600652414766,
|
| 1665 |
+
"kl": 0.000946044921875,
|
| 1666 |
+
"learning_rate": 5.97037808470444e-07,
|
| 1667 |
+
"loss": 0.0405,
|
| 1668 |
+
"num_tokens": 6543881.0,
|
| 1669 |
+
"reward": 2.7356655597686768,
|
| 1670 |
+
"reward_std": 1.2655977606773376,
|
| 1671 |
+
"rewards/accuracy_reward": 0.15625,
|
| 1672 |
+
"rewards/format_reward": 0.875,
|
| 1673 |
+
"rewards/tag_count_reward": 1.5481655597686768,
|
| 1674 |
+
"step": 104
|
| 1675 |
+
},
|
| 1676 |
+
{
|
| 1677 |
+
"clip_ratio": 0.0,
|
| 1678 |
+
"completion_length": 1920.8125,
|
| 1679 |
+
"epoch": 0.004480764717178398,
|
| 1680 |
+
"grad_norm": 0.3426045186198588,
|
| 1681 |
+
"kl": 0.000942230224609375,
|
| 1682 |
+
"learning_rate": 5.892200842364462e-07,
|
| 1683 |
+
"loss": 0.0108,
|
| 1684 |
+
"num_tokens": 6610955.0,
|
| 1685 |
+
"reward": 2.6052074432373047,
|
| 1686 |
+
"reward_std": 1.2626380622386932,
|
| 1687 |
+
"rewards/accuracy_reward": 0.25,
|
| 1688 |
+
"rewards/format_reward": 0.875,
|
| 1689 |
+
"rewards/tag_count_reward": 1.2302075028419495,
|
| 1690 |
+
"step": 105
|
| 1691 |
+
},
|
| 1692 |
+
{
|
| 1693 |
+
"clip_ratio": 0.0,
|
| 1694 |
+
"completion_length": 2048.0,
|
| 1695 |
+
"epoch": 0.004523438666865812,
|
| 1696 |
+
"grad_norm": 0.3379998633677249,
|
| 1697 |
+
"kl": 0.0009174346923828125,
|
| 1698 |
+
"learning_rate": 5.813904131848564e-07,
|
| 1699 |
+
"loss": 0.0,
|
| 1700 |
+
"num_tokens": 6681579.0,
|
| 1701 |
+
"reward": 2.126375198364258,
|
| 1702 |
+
"reward_std": 0.8321560025215149,
|
| 1703 |
+
"rewards/accuracy_reward": 0.0,
|
| 1704 |
+
"rewards/format_reward": 0.90625,
|
| 1705 |
+
"rewards/tag_count_reward": 1.2201251089572906,
|
| 1706 |
+
"step": 106
|
| 1707 |
+
},
|
| 1708 |
+
{
|
| 1709 |
+
"clip_ratio": 0.0,
|
| 1710 |
+
"completion_length": 2037.9375,
|
| 1711 |
+
"epoch": 0.004566112616553225,
|
| 1712 |
+
"grad_norm": 0.36356032002037114,
|
| 1713 |
+
"kl": 0.00099945068359375,
|
| 1714 |
+
"learning_rate": 5.735511803093248e-07,
|
| 1715 |
+
"loss": 0.0048,
|
| 1716 |
+
"num_tokens": 6753849.0,
|
| 1717 |
+
"reward": 2.113395571708679,
|
| 1718 |
+
"reward_std": 1.3873547911643982,
|
| 1719 |
+
"rewards/accuracy_reward": 0.0,
|
| 1720 |
+
"rewards/format_reward": 0.875,
|
| 1721 |
+
"rewards/tag_count_reward": 1.2383955717086792,
|
| 1722 |
+
"step": 107
|
| 1723 |
+
},
|
| 1724 |
+
{
|
| 1725 |
+
"clip_ratio": 0.0,
|
| 1726 |
+
"completion_length": 2019.75,
|
| 1727 |
+
"epoch": 0.004608786566240639,
|
| 1728 |
+
"grad_norm": 0.36749657124476337,
|
| 1729 |
+
"kl": 0.0009746551513671875,
|
| 1730 |
+
"learning_rate": 5.657047735161255e-07,
|
| 1731 |
+
"loss": 0.0103,
|
| 1732 |
+
"num_tokens": 6824953.0,
|
| 1733 |
+
"reward": 2.068533778190613,
|
| 1734 |
+
"reward_std": 1.2282358407974243,
|
| 1735 |
+
"rewards/accuracy_reward": 0.0,
|
| 1736 |
+
"rewards/format_reward": 0.8125,
|
| 1737 |
+
"rewards/tag_count_reward": 1.2560337781906128,
|
| 1738 |
+
"step": 108
|
| 1739 |
+
},
|
| 1740 |
+
{
|
| 1741 |
+
"clip_ratio": 0.0,
|
| 1742 |
+
"completion_length": 2048.0,
|
| 1743 |
+
"epoch": 0.004651460515928052,
|
| 1744 |
+
"grad_norm": 0.3803488894376047,
|
| 1745 |
+
"kl": 0.001068115234375,
|
| 1746 |
+
"learning_rate": 5.578535828967777e-07,
|
| 1747 |
+
"loss": 0.0,
|
| 1748 |
+
"num_tokens": 6896057.0,
|
| 1749 |
+
"reward": 1.9097256660461426,
|
| 1750 |
+
"reward_std": 1.2217622995376587,
|
| 1751 |
+
"rewards/accuracy_reward": 0.0,
|
| 1752 |
+
"rewards/format_reward": 0.8125,
|
| 1753 |
+
"rewards/tag_count_reward": 1.0972257256507874,
|
| 1754 |
+
"step": 109
|
| 1755 |
+
},
|
| 1756 |
+
{
|
| 1757 |
+
"clip_ratio": 0.0,
|
| 1758 |
+
"completion_length": 1945.09375,
|
| 1759 |
+
"epoch": 0.0046941344656154654,
|
| 1760 |
+
"grad_norm": 0.33576955000963377,
|
| 1761 |
+
"kl": 0.001079559326171875,
|
| 1762 |
+
"learning_rate": 5.5e-07,
|
| 1763 |
+
"loss": -0.0099,
|
| 1764 |
+
"num_tokens": 6964372.0,
|
| 1765 |
+
"reward": 2.556748628616333,
|
| 1766 |
+
"reward_std": 1.163586437702179,
|
| 1767 |
+
"rewards/accuracy_reward": 0.03125,
|
| 1768 |
+
"rewards/format_reward": 0.9375,
|
| 1769 |
+
"rewards/tag_count_reward": 1.5567485094070435,
|
| 1770 |
+
"step": 110
|
| 1771 |
+
},
|
| 1772 |
+
{
|
| 1773 |
+
"clip_ratio": 0.0,
|
| 1774 |
+
"completion_length": 2011.0,
|
| 1775 |
+
"epoch": 0.004736808415302878,
|
| 1776 |
+
"grad_norm": 0.3245985410049708,
|
| 1777 |
+
"kl": 0.00090789794921875,
|
| 1778 |
+
"learning_rate": 5.421464171032224e-07,
|
| 1779 |
+
"loss": 0.0152,
|
| 1780 |
+
"num_tokens": 7033492.0,
|
| 1781 |
+
"reward": 2.198875606060028,
|
| 1782 |
+
"reward_std": 1.1500471830368042,
|
| 1783 |
+
"rewards/accuracy_reward": 0.0,
|
| 1784 |
+
"rewards/format_reward": 0.90625,
|
| 1785 |
+
"rewards/tag_count_reward": 1.2926255464553833,
|
| 1786 |
+
"step": 111
|
| 1787 |
+
},
|
| 1788 |
+
{
|
| 1789 |
+
"clip_ratio": 0.0,
|
| 1790 |
+
"completion_length": 2048.0,
|
| 1791 |
+
"epoch": 0.004779482364990291,
|
| 1792 |
+
"grad_norm": 0.32890781756594784,
|
| 1793 |
+
"kl": 0.001018524169921875,
|
| 1794 |
+
"learning_rate": 5.342952264838747e-07,
|
| 1795 |
+
"loss": 0.0,
|
| 1796 |
+
"num_tokens": 7104964.0,
|
| 1797 |
+
"reward": 2.050785005092621,
|
| 1798 |
+
"reward_std": 1.3915192484855652,
|
| 1799 |
+
"rewards/accuracy_reward": 0.03125,
|
| 1800 |
+
"rewards/format_reward": 0.75,
|
| 1801 |
+
"rewards/tag_count_reward": 1.2382850646972656,
|
| 1802 |
+
"step": 112
|
| 1803 |
+
},
|
| 1804 |
+
{
|
| 1805 |
+
"clip_ratio": 0.0,
|
| 1806 |
+
"completion_length": 2048.0,
|
| 1807 |
+
"epoch": 0.004822156314677705,
|
| 1808 |
+
"grad_norm": 0.34559349472976164,
|
| 1809 |
+
"kl": 0.00093841552734375,
|
| 1810 |
+
"learning_rate": 5.264488196906752e-07,
|
| 1811 |
+
"loss": 0.0,
|
| 1812 |
+
"num_tokens": 7177252.0,
|
| 1813 |
+
"reward": 1.7611371278762817,
|
| 1814 |
+
"reward_std": 1.1777510344982147,
|
| 1815 |
+
"rewards/accuracy_reward": 0.0,
|
| 1816 |
+
"rewards/format_reward": 0.84375,
|
| 1817 |
+
"rewards/tag_count_reward": 0.917387068271637,
|
| 1818 |
+
"step": 113
|
| 1819 |
+
},
|
| 1820 |
+
{
|
| 1821 |
+
"clip_ratio": 0.0,
|
| 1822 |
+
"completion_length": 1952.40625,
|
| 1823 |
+
"epoch": 0.004864830264365118,
|
| 1824 |
+
"grad_norm": 0.30144362610838493,
|
| 1825 |
+
"kl": 0.0009708404541015625,
|
| 1826 |
+
"learning_rate": 5.186095868151436e-07,
|
| 1827 |
+
"loss": 0.0002,
|
| 1828 |
+
"num_tokens": 7245105.0,
|
| 1829 |
+
"reward": 3.2852156162261963,
|
| 1830 |
+
"reward_std": 1.8544931411743164,
|
| 1831 |
+
"rewards/accuracy_reward": 0.28125,
|
| 1832 |
+
"rewards/format_reward": 0.8125,
|
| 1833 |
+
"rewards/tag_count_reward": 1.910215437412262,
|
| 1834 |
+
"step": 114
|
| 1835 |
+
},
|
| 1836 |
+
{
|
| 1837 |
+
"clip_ratio": 0.0,
|
| 1838 |
+
"completion_length": 2040.34375,
|
| 1839 |
+
"epoch": 0.004907504214052532,
|
| 1840 |
+
"grad_norm": 0.31929750752991015,
|
| 1841 |
+
"kl": 0.0009746551513671875,
|
| 1842 |
+
"learning_rate": 5.107799157635538e-07,
|
| 1843 |
+
"loss": 0.0015,
|
| 1844 |
+
"num_tokens": 7317572.0,
|
| 1845 |
+
"reward": 2.291470944881439,
|
| 1846 |
+
"reward_std": 1.326077938079834,
|
| 1847 |
+
"rewards/accuracy_reward": 0.125,
|
| 1848 |
+
"rewards/format_reward": 0.875,
|
| 1849 |
+
"rewards/tag_count_reward": 1.1664709448814392,
|
| 1850 |
+
"step": 115
|
| 1851 |
+
},
|
| 1852 |
+
{
|
| 1853 |
+
"clip_ratio": 0.0,
|
| 1854 |
+
"completion_length": 2048.0,
|
| 1855 |
+
"epoch": 0.004950178163739945,
|
| 1856 |
+
"grad_norm": 0.35025027787200474,
|
| 1857 |
+
"kl": 0.001041412353515625,
|
| 1858 |
+
"learning_rate": 5.02962191529556e-07,
|
| 1859 |
+
"loss": 0.0,
|
| 1860 |
+
"num_tokens": 7388380.0,
|
| 1861 |
+
"reward": 2.158087372779846,
|
| 1862 |
+
"reward_std": 1.1965221166610718,
|
| 1863 |
+
"rewards/accuracy_reward": 0.0,
|
| 1864 |
+
"rewards/format_reward": 0.90625,
|
| 1865 |
+
"rewards/tag_count_reward": 1.2518373727798462,
|
| 1866 |
+
"step": 116
|
| 1867 |
+
},
|
| 1868 |
+
{
|
| 1869 |
+
"clip_ratio": 0.0,
|
| 1870 |
+
"completion_length": 1986.84375,
|
| 1871 |
+
"epoch": 0.0049928521134273585,
|
| 1872 |
+
"grad_norm": 0.2789268550906952,
|
| 1873 |
+
"kl": 0.0009517669677734375,
|
| 1874 |
+
"learning_rate": 4.951587954676837e-07,
|
| 1875 |
+
"loss": 0.0028,
|
| 1876 |
+
"num_tokens": 7458015.0,
|
| 1877 |
+
"reward": 3.049384593963623,
|
| 1878 |
+
"reward_std": 1.4109481573104858,
|
| 1879 |
+
"rewards/accuracy_reward": 0.09375,
|
| 1880 |
+
"rewards/format_reward": 0.90625,
|
| 1881 |
+
"rewards/tag_count_reward": 1.955634593963623,
|
| 1882 |
+
"step": 117
|
| 1883 |
+
},
|
| 1884 |
+
{
|
| 1885 |
+
"clip_ratio": 0.0,
|
| 1886 |
+
"completion_length": 1849.0625,
|
| 1887 |
+
"epoch": 0.0050355260631147715,
|
| 1888 |
+
"grad_norm": 0.3549379731693497,
|
| 1889 |
+
"kl": 0.0010509490966796875,
|
| 1890 |
+
"learning_rate": 4.873721045679706e-07,
|
| 1891 |
+
"loss": 0.0091,
|
| 1892 |
+
"num_tokens": 7522417.0,
|
| 1893 |
+
"reward": 2.7751612663269043,
|
| 1894 |
+
"reward_std": 1.0009263753890991,
|
| 1895 |
+
"rewards/accuracy_reward": 0.03125,
|
| 1896 |
+
"rewards/format_reward": 0.96875,
|
| 1897 |
+
"rewards/tag_count_reward": 1.7439113855361938,
|
| 1898 |
+
"step": 118
|
| 1899 |
+
},
|
| 1900 |
+
{
|
| 1901 |
+
"clip_ratio": 0.0,
|
| 1902 |
+
"completion_length": 2045.1875,
|
| 1903 |
+
"epoch": 0.005078200012802185,
|
| 1904 |
+
"grad_norm": 0.3221098442067067,
|
| 1905 |
+
"kl": 0.0009918212890625,
|
| 1906 |
+
"learning_rate": 4.79604490731896e-07,
|
| 1907 |
+
"loss": 0.0009,
|
| 1908 |
+
"num_tokens": 7593711.0,
|
| 1909 |
+
"reward": 2.1338720321655273,
|
| 1910 |
+
"reward_std": 1.2568315863609314,
|
| 1911 |
+
"rewards/accuracy_reward": 0.0,
|
| 1912 |
+
"rewards/format_reward": 0.90625,
|
| 1913 |
+
"rewards/tag_count_reward": 1.227622002363205,
|
| 1914 |
+
"step": 119
|
| 1915 |
+
},
|
| 1916 |
+
{
|
| 1917 |
+
"clip_ratio": 0.0,
|
| 1918 |
+
"completion_length": 1947.6875,
|
| 1919 |
+
"epoch": 0.005120873962489598,
|
| 1920 |
+
"grad_norm": 0.3456924606201143,
|
| 1921 |
+
"kl": 0.001132965087890625,
|
| 1922 |
+
"learning_rate": 4.7185832004988133e-07,
|
| 1923 |
+
"loss": 0.0276,
|
| 1924 |
+
"num_tokens": 7660837.0,
|
| 1925 |
+
"reward": 2.8727614879608154,
|
| 1926 |
+
"reward_std": 1.0015690922737122,
|
| 1927 |
+
"rewards/accuracy_reward": 0.25,
|
| 1928 |
+
"rewards/format_reward": 0.9375,
|
| 1929 |
+
"rewards/tag_count_reward": 1.435261607170105,
|
| 1930 |
+
"step": 120
|
| 1931 |
+
},
|
| 1932 |
+
{
|
| 1933 |
+
"clip_ratio": 0.0,
|
| 1934 |
+
"completion_length": 2039.5625,
|
| 1935 |
+
"epoch": 0.005163547912177011,
|
| 1936 |
+
"grad_norm": 0.3407712108637507,
|
| 1937 |
+
"kl": 0.0010280609130859375,
|
| 1938 |
+
"learning_rate": 4.641359520805548e-07,
|
| 1939 |
+
"loss": 0.0036,
|
| 1940 |
+
"num_tokens": 7732751.0,
|
| 1941 |
+
"reward": 1.6611133813858032,
|
| 1942 |
+
"reward_std": 1.125486671924591,
|
| 1943 |
+
"rewards/accuracy_reward": 0.03125,
|
| 1944 |
+
"rewards/format_reward": 0.8125,
|
| 1945 |
+
"rewards/tag_count_reward": 0.7861133515834808,
|
| 1946 |
+
"step": 121
|
| 1947 |
+
},
|
| 1948 |
+
{
|
| 1949 |
+
"clip_ratio": 0.0,
|
| 1950 |
+
"completion_length": 2048.0,
|
| 1951 |
+
"epoch": 0.005206221861864425,
|
| 1952 |
+
"grad_norm": 0.32347957247456194,
|
| 1953 |
+
"kl": 0.000988006591796875,
|
| 1954 |
+
"learning_rate": 4.5643973913200837e-07,
|
| 1955 |
+
"loss": 0.0,
|
| 1956 |
+
"num_tokens": 7803295.0,
|
| 1957 |
+
"reward": 2.0753968954086304,
|
| 1958 |
+
"reward_std": 1.0208499729633331,
|
| 1959 |
+
"rewards/accuracy_reward": 0.03125,
|
| 1960 |
+
"rewards/format_reward": 0.90625,
|
| 1961 |
+
"rewards/tag_count_reward": 1.1066468954086304,
|
| 1962 |
+
"step": 122
|
| 1963 |
+
},
|
| 1964 |
+
{
|
| 1965 |
+
"clip_ratio": 0.0,
|
| 1966 |
+
"completion_length": 2048.0,
|
| 1967 |
+
"epoch": 0.005248895811551838,
|
| 1968 |
+
"grad_norm": 0.33468792230403704,
|
| 1969 |
+
"kl": 0.0010223388671875,
|
| 1970 |
+
"learning_rate": 4.4877202554526084e-07,
|
| 1971 |
+
"loss": 0.0,
|
| 1972 |
+
"num_tokens": 7874015.0,
|
| 1973 |
+
"reward": 1.995333194732666,
|
| 1974 |
+
"reward_std": 1.0585032105445862,
|
| 1975 |
+
"rewards/accuracy_reward": 0.0,
|
| 1976 |
+
"rewards/format_reward": 0.8125,
|
| 1977 |
+
"rewards/tag_count_reward": 1.1828331351280212,
|
| 1978 |
+
"step": 123
|
| 1979 |
+
},
|
| 1980 |
+
{
|
| 1981 |
+
"clip_ratio": 0.0,
|
| 1982 |
+
"completion_length": 1744.0625,
|
| 1983 |
+
"epoch": 0.005291569761239252,
|
| 1984 |
+
"grad_norm": 0.3686886780915071,
|
| 1985 |
+
"kl": 0.001056671142578125,
|
| 1986 |
+
"learning_rate": 4.4113514698014953e-07,
|
| 1987 |
+
"loss": -0.0286,
|
| 1988 |
+
"num_tokens": 7935625.0,
|
| 1989 |
+
"reward": 2.379127860069275,
|
| 1990 |
+
"reward_std": 1.1280320882797241,
|
| 1991 |
+
"rewards/accuracy_reward": 0.0,
|
| 1992 |
+
"rewards/format_reward": 0.90625,
|
| 1993 |
+
"rewards/tag_count_reward": 1.4728778004646301,
|
| 1994 |
+
"step": 124
|
| 1995 |
+
},
|
| 1996 |
+
{
|
| 1997 |
+
"clip_ratio": 0.0,
|
| 1998 |
+
"completion_length": 2043.3125,
|
| 1999 |
+
"epoch": 0.0053342437109266646,
|
| 2000 |
+
"grad_norm": 0.34656429321684246,
|
| 2001 |
+
"kl": 0.00107574462890625,
|
| 2002 |
+
"learning_rate": 4.3353142970386557e-07,
|
| 2003 |
+
"loss": 0.0041,
|
| 2004 |
+
"num_tokens": 8007515.0,
|
| 2005 |
+
"reward": 2.0800933837890625,
|
| 2006 |
+
"reward_std": 0.8104768097400665,
|
| 2007 |
+
"rewards/accuracy_reward": 0.0,
|
| 2008 |
+
"rewards/format_reward": 0.90625,
|
| 2009 |
+
"rewards/tag_count_reward": 1.1738433837890625,
|
| 2010 |
+
"step": 125
|
| 2011 |
+
},
|
| 2012 |
+
{
|
| 2013 |
+
"clip_ratio": 0.0,
|
| 2014 |
+
"completion_length": 1996.34375,
|
| 2015 |
+
"epoch": 0.005376917660614078,
|
| 2016 |
+
"grad_norm": 0.41859075180317595,
|
| 2017 |
+
"kl": 0.00128936767578125,
|
| 2018 |
+
"learning_rate": 4.2596318988235037e-07,
|
| 2019 |
+
"loss": 0.0132,
|
| 2020 |
+
"num_tokens": 8077486.0,
|
| 2021 |
+
"reward": 2.490622878074646,
|
| 2022 |
+
"reward_std": 1.4971110820770264,
|
| 2023 |
+
"rewards/accuracy_reward": 0.125,
|
| 2024 |
+
"rewards/format_reward": 0.8125,
|
| 2025 |
+
"rewards/tag_count_reward": 1.428122729063034,
|
| 2026 |
+
"step": 126
|
| 2027 |
+
},
|
| 2028 |
+
{
|
| 2029 |
+
"clip_ratio": 0.0,
|
| 2030 |
+
"completion_length": 1941.0,
|
| 2031 |
+
"epoch": 0.005419591610301491,
|
| 2032 |
+
"grad_norm": 0.3736166895797362,
|
| 2033 |
+
"kl": 0.0011749267578125,
|
| 2034 |
+
"learning_rate": 4.1843273287476854e-07,
|
| 2035 |
+
"loss": 0.0115,
|
| 2036 |
+
"num_tokens": 8144590.0,
|
| 2037 |
+
"reward": 2.725239336490631,
|
| 2038 |
+
"reward_std": 1.1180997490882874,
|
| 2039 |
+
"rewards/accuracy_reward": 0.25,
|
| 2040 |
+
"rewards/format_reward": 0.78125,
|
| 2041 |
+
"rewards/tag_count_reward": 1.4439894556999207,
|
| 2042 |
+
"step": 127
|
| 2043 |
+
},
|
| 2044 |
+
{
|
| 2045 |
+
"clip_ratio": 0.0,
|
| 2046 |
+
"completion_length": 1716.75,
|
| 2047 |
+
"epoch": 0.005462265559988905,
|
| 2048 |
+
"grad_norm": 0.4779542997251708,
|
| 2049 |
+
"kl": 0.001068115234375,
|
| 2050 |
+
"learning_rate": 4.1094235253127374e-07,
|
| 2051 |
+
"loss": 0.0868,
|
| 2052 |
+
"num_tokens": 8204822.0,
|
| 2053 |
+
"reward": 2.6972365379333496,
|
| 2054 |
+
"reward_std": 1.204147458076477,
|
| 2055 |
+
"rewards/accuracy_reward": 0.21875,
|
| 2056 |
+
"rewards/format_reward": 0.875,
|
| 2057 |
+
"rewards/tag_count_reward": 1.3847364783287048,
|
| 2058 |
+
"step": 128
|
| 2059 |
+
},
|
| 2060 |
+
{
|
| 2061 |
+
"clip_ratio": 0.0,
|
| 2062 |
+
"completion_length": 1914.84375,
|
| 2063 |
+
"epoch": 0.005504939509676318,
|
| 2064 |
+
"grad_norm": 0.3042008559461902,
|
| 2065 |
+
"kl": 0.0008449554443359375,
|
| 2066 |
+
"learning_rate": 4.034943304942796e-07,
|
| 2067 |
+
"loss": 0.0111,
|
| 2068 |
+
"num_tokens": 8271769.0,
|
| 2069 |
+
"reward": 1.965047836303711,
|
| 2070 |
+
"reward_std": 0.9716286659240723,
|
| 2071 |
+
"rewards/accuracy_reward": 0.0,
|
| 2072 |
+
"rewards/format_reward": 0.875,
|
| 2073 |
+
"rewards/tag_count_reward": 1.0900478065013885,
|
| 2074 |
+
"step": 129
|
| 2075 |
+
},
|
| 2076 |
+
{
|
| 2077 |
+
"clip_ratio": 0.0,
|
| 2078 |
+
"completion_length": 1889.875,
|
| 2079 |
+
"epoch": 0.005547613459363732,
|
| 2080 |
+
"grad_norm": 0.3947768414740478,
|
| 2081 |
+
"kl": 0.001407623291015625,
|
| 2082 |
+
"learning_rate": 3.9609093550344907e-07,
|
| 2083 |
+
"loss": 0.0054,
|
| 2084 |
+
"num_tokens": 8337773.0,
|
| 2085 |
+
"reward": 3.042949616909027,
|
| 2086 |
+
"reward_std": 1.368441104888916,
|
| 2087 |
+
"rewards/accuracy_reward": 0.21875,
|
| 2088 |
+
"rewards/format_reward": 0.8125,
|
| 2089 |
+
"rewards/tag_count_reward": 1.792949616909027,
|
| 2090 |
+
"step": 130
|
| 2091 |
+
},
|
| 2092 |
+
{
|
| 2093 |
+
"clip_ratio": 0.0,
|
| 2094 |
+
"completion_length": 2041.8125,
|
| 2095 |
+
"epoch": 0.005590287409051145,
|
| 2096 |
+
"grad_norm": 0.31015871057673766,
|
| 2097 |
+
"kl": 0.00103759765625,
|
| 2098 |
+
"learning_rate": 3.8873442270461485e-07,
|
| 2099 |
+
"loss": 0.0019,
|
| 2100 |
+
"num_tokens": 8407695.0,
|
| 2101 |
+
"reward": 2.332514524459839,
|
| 2102 |
+
"reward_std": 1.6139944195747375,
|
| 2103 |
+
"rewards/accuracy_reward": 0.15625,
|
| 2104 |
+
"rewards/format_reward": 0.78125,
|
| 2105 |
+
"rewards/tag_count_reward": 1.2387645244598389,
|
| 2106 |
+
"step": 131
|
| 2107 |
+
},
|
| 2108 |
+
{
|
| 2109 |
+
"clip_ratio": 0.0,
|
| 2110 |
+
"completion_length": 1796.6875,
|
| 2111 |
+
"epoch": 0.005632961358738558,
|
| 2112 |
+
"grad_norm": 0.36693301717015325,
|
| 2113 |
+
"kl": 0.00124359130859375,
|
| 2114 |
+
"learning_rate": 3.8142703296283953e-07,
|
| 2115 |
+
"loss": 0.0496,
|
| 2116 |
+
"num_tokens": 8471805.0,
|
| 2117 |
+
"reward": 3.234046697616577,
|
| 2118 |
+
"reward_std": 1.5076982378959656,
|
| 2119 |
+
"rewards/accuracy_reward": 0.09375,
|
| 2120 |
+
"rewards/format_reward": 0.8125,
|
| 2121 |
+
"rewards/tag_count_reward": 2.2340465784072876,
|
| 2122 |
+
"step": 132
|
| 2123 |
+
},
|
| 2124 |
+
{
|
| 2125 |
+
"clip_ratio": 0.0,
|
| 2126 |
+
"completion_length": 2048.0,
|
| 2127 |
+
"epoch": 0.0056756353084259714,
|
| 2128 |
+
"grad_norm": 0.3691319691314675,
|
| 2129 |
+
"kl": 0.001190185546875,
|
| 2130 |
+
"learning_rate": 3.7417099217982686e-07,
|
| 2131 |
+
"loss": 0.0,
|
| 2132 |
+
"num_tokens": 8544429.0,
|
| 2133 |
+
"reward": 2.220004916191101,
|
| 2134 |
+
"reward_std": 0.8086199164390564,
|
| 2135 |
+
"rewards/accuracy_reward": 0.0,
|
| 2136 |
+
"rewards/format_reward": 0.90625,
|
| 2137 |
+
"rewards/tag_count_reward": 1.3137548565864563,
|
| 2138 |
+
"step": 133
|
| 2139 |
+
},
|
| 2140 |
+
{
|
| 2141 |
+
"clip_ratio": 0.0,
|
| 2142 |
+
"completion_length": 2048.0,
|
| 2143 |
+
"epoch": 0.005718309258113384,
|
| 2144 |
+
"grad_norm": 0.3118553992708358,
|
| 2145 |
+
"kl": 0.001224517822265625,
|
| 2146 |
+
"learning_rate": 3.6696851061588994e-07,
|
| 2147 |
+
"loss": 0.0,
|
| 2148 |
+
"num_tokens": 8614901.0,
|
| 2149 |
+
"reward": 2.4043807983398438,
|
| 2150 |
+
"reward_std": 1.0689508318901062,
|
| 2151 |
+
"rewards/accuracy_reward": 0.0,
|
| 2152 |
+
"rewards/format_reward": 0.875,
|
| 2153 |
+
"rewards/tag_count_reward": 1.529380738735199,
|
| 2154 |
+
"step": 134
|
| 2155 |
+
},
|
| 2156 |
+
{
|
| 2157 |
+
"clip_ratio": 0.0,
|
| 2158 |
+
"completion_length": 2010.0,
|
| 2159 |
+
"epoch": 0.005760983207800798,
|
| 2160 |
+
"grad_norm": 0.31464372391687856,
|
| 2161 |
+
"kl": 0.001071929931640625,
|
| 2162 |
+
"learning_rate": 3.5982178221668533e-07,
|
| 2163 |
+
"loss": 0.02,
|
| 2164 |
+
"num_tokens": 8684421.0,
|
| 2165 |
+
"reward": 2.694029927253723,
|
| 2166 |
+
"reward_std": 1.4979963898658752,
|
| 2167 |
+
"rewards/accuracy_reward": 0.1875,
|
| 2168 |
+
"rewards/format_reward": 0.8125,
|
| 2169 |
+
"rewards/tag_count_reward": 1.506529986858368,
|
| 2170 |
+
"step": 135
|
| 2171 |
+
},
|
| 2172 |
+
{
|
| 2173 |
+
"clip_ratio": 0.0,
|
| 2174 |
+
"completion_length": 2048.0,
|
| 2175 |
+
"epoch": 0.005803657157488211,
|
| 2176 |
+
"grad_norm": 0.38525969132463184,
|
| 2177 |
+
"kl": 0.001331329345703125,
|
| 2178 |
+
"learning_rate": 3.5273298394491515e-07,
|
| 2179 |
+
"loss": 0.0001,
|
| 2180 |
+
"num_tokens": 8756133.0,
|
| 2181 |
+
"reward": 1.5289509296417236,
|
| 2182 |
+
"reward_std": 1.3215290307998657,
|
| 2183 |
+
"rewards/accuracy_reward": 0.0,
|
| 2184 |
+
"rewards/format_reward": 0.75,
|
| 2185 |
+
"rewards/tag_count_reward": 0.7789508700370789,
|
| 2186 |
+
"step": 136
|
| 2187 |
+
},
|
| 2188 |
+
{
|
| 2189 |
+
"clip_ratio": 0.0,
|
| 2190 |
+
"completion_length": 2048.0,
|
| 2191 |
+
"epoch": 0.005846331107175625,
|
| 2192 |
+
"grad_norm": 0.3644044385324543,
|
| 2193 |
+
"kl": 0.00121307373046875,
|
| 2194 |
+
"learning_rate": 3.45704275117204e-07,
|
| 2195 |
+
"loss": 0.0,
|
| 2196 |
+
"num_tokens": 8827509.0,
|
| 2197 |
+
"reward": 2.668272852897644,
|
| 2198 |
+
"reward_std": 1.1597683429718018,
|
| 2199 |
+
"rewards/accuracy_reward": 0.0,
|
| 2200 |
+
"rewards/format_reward": 0.90625,
|
| 2201 |
+
"rewards/tag_count_reward": 1.7620229125022888,
|
| 2202 |
+
"step": 137
|
| 2203 |
+
},
|
| 2204 |
+
{
|
| 2205 |
+
"clip_ratio": 0.0,
|
| 2206 |
+
"completion_length": 1872.21875,
|
| 2207 |
+
"epoch": 0.005889005056863038,
|
| 2208 |
+
"grad_norm": 0.35534057190923696,
|
| 2209 |
+
"kl": 0.001163482666015625,
|
| 2210 |
+
"learning_rate": 3.387377967463493e-07,
|
| 2211 |
+
"loss": 0.0162,
|
| 2212 |
+
"num_tokens": 8894012.0,
|
| 2213 |
+
"reward": 2.496087372303009,
|
| 2214 |
+
"reward_std": 1.2811987400054932,
|
| 2215 |
+
"rewards/accuracy_reward": 0.25,
|
| 2216 |
+
"rewards/format_reward": 0.8125,
|
| 2217 |
+
"rewards/tag_count_reward": 1.1835873275995255,
|
| 2218 |
+
"step": 138
|
| 2219 |
+
},
|
| 2220 |
+
{
|
| 2221 |
+
"clip_ratio": 0.0,
|
| 2222 |
+
"completion_length": 2048.0,
|
| 2223 |
+
"epoch": 0.005931679006550452,
|
| 2224 |
+
"grad_norm": 0.39728711388972204,
|
| 2225 |
+
"kl": 0.001323699951171875,
|
| 2226 |
+
"learning_rate": 3.3183567088914833e-07,
|
| 2227 |
+
"loss": 0.0001,
|
| 2228 |
+
"num_tokens": 8966772.0,
|
| 2229 |
+
"reward": 2.4450310468673706,
|
| 2230 |
+
"reward_std": 0.8909086883068085,
|
| 2231 |
+
"rewards/accuracy_reward": 0.03125,
|
| 2232 |
+
"rewards/format_reward": 0.9375,
|
| 2233 |
+
"rewards/tag_count_reward": 1.445030927658081,
|
| 2234 |
+
"step": 139
|
| 2235 |
+
},
|
| 2236 |
+
{
|
| 2237 |
+
"clip_ratio": 0.0,
|
| 2238 |
+
"completion_length": 1968.28125,
|
| 2239 |
+
"epoch": 0.0059743529562378645,
|
| 2240 |
+
"grad_norm": 0.30500329793094405,
|
| 2241 |
+
"kl": 0.000980377197265625,
|
| 2242 |
+
"learning_rate": 3.250000000000001e-07,
|
| 2243 |
+
"loss": 0.0114,
|
| 2244 |
+
"num_tokens": 9036381.0,
|
| 2245 |
+
"reward": 2.6805665493011475,
|
| 2246 |
+
"reward_std": 1.5157400965690613,
|
| 2247 |
+
"rewards/accuracy_reward": 0.1875,
|
| 2248 |
+
"rewards/format_reward": 0.8125,
|
| 2249 |
+
"rewards/tag_count_reward": 1.4930665493011475,
|
| 2250 |
+
"step": 140
|
| 2251 |
+
},
|
| 2252 |
+
{
|
| 2253 |
+
"clip_ratio": 0.0,
|
| 2254 |
+
"completion_length": 2048.0,
|
| 2255 |
+
"epoch": 0.006017026905925278,
|
| 2256 |
+
"grad_norm": 0.3293865836158587,
|
| 2257 |
+
"kl": 0.001049041748046875,
|
| 2258 |
+
"learning_rate": 3.182328662904756e-07,
|
| 2259 |
+
"loss": 0.0,
|
| 2260 |
+
"num_tokens": 9107821.0,
|
| 2261 |
+
"reward": 2.222343862056732,
|
| 2262 |
+
"reward_std": 0.7479121387004852,
|
| 2263 |
+
"rewards/accuracy_reward": 0.0,
|
| 2264 |
+
"rewards/format_reward": 0.96875,
|
| 2265 |
+
"rewards/tag_count_reward": 1.253593772649765,
|
| 2266 |
+
"step": 141
|
| 2267 |
+
},
|
| 2268 |
+
{
|
| 2269 |
+
"clip_ratio": 0.0,
|
| 2270 |
+
"completion_length": 2048.0,
|
| 2271 |
+
"epoch": 0.006059700855612691,
|
| 2272 |
+
"grad_norm": 0.3405141289074622,
|
| 2273 |
+
"kl": 0.00115966796875,
|
| 2274 |
+
"learning_rate": 3.115363310950578e-07,
|
| 2275 |
+
"loss": 0.0,
|
| 2276 |
+
"num_tokens": 9179373.0,
|
| 2277 |
+
"reward": 2.8447316884994507,
|
| 2278 |
+
"reward_std": 1.4372076988220215,
|
| 2279 |
+
"rewards/accuracy_reward": 0.21875,
|
| 2280 |
+
"rewards/format_reward": 0.8125,
|
| 2281 |
+
"rewards/tag_count_reward": 1.5947316884994507,
|
| 2282 |
+
"step": 142
|
| 2283 |
+
},
|
| 2284 |
+
{
|
| 2285 |
+
"clip_ratio": 0.0,
|
| 2286 |
+
"completion_length": 2048.0,
|
| 2287 |
+
"epoch": 0.006102374805300104,
|
| 2288 |
+
"grad_norm": 0.3646610270005593,
|
| 2289 |
+
"kl": 0.001434326171875,
|
| 2290 |
+
"learning_rate": 3.0491243424323783e-07,
|
| 2291 |
+
"loss": 0.0001,
|
| 2292 |
+
"num_tokens": 9250901.0,
|
| 2293 |
+
"reward": 2.32478129863739,
|
| 2294 |
+
"reward_std": 1.146507978439331,
|
| 2295 |
+
"rewards/accuracy_reward": 0.0,
|
| 2296 |
+
"rewards/format_reward": 0.8125,
|
| 2297 |
+
"rewards/tag_count_reward": 1.5122812390327454,
|
| 2298 |
+
"step": 143
|
| 2299 |
+
},
|
| 2300 |
+
{
|
| 2301 |
+
"clip_ratio": 0.0,
|
| 2302 |
+
"completion_length": 1783.8125,
|
| 2303 |
+
"epoch": 0.006145048754987518,
|
| 2304 |
+
"grad_norm": 0.35963795662975445,
|
| 2305 |
+
"kl": 0.0012359619140625,
|
| 2306 |
+
"learning_rate": 2.9836319343816397e-07,
|
| 2307 |
+
"loss": -0.0165,
|
| 2308 |
+
"num_tokens": 9313207.0,
|
| 2309 |
+
"reward": 2.5646426677703857,
|
| 2310 |
+
"reward_std": 1.2103586196899414,
|
| 2311 |
+
"rewards/accuracy_reward": 0.0,
|
| 2312 |
+
"rewards/format_reward": 0.875,
|
| 2313 |
+
"rewards/tag_count_reward": 1.6896427273750305,
|
| 2314 |
+
"step": 144
|
| 2315 |
+
},
|
| 2316 |
+
{
|
| 2317 |
+
"clip_ratio": 0.0,
|
| 2318 |
+
"completion_length": 2048.0,
|
| 2319 |
+
"epoch": 0.006187722704674931,
|
| 2320 |
+
"grad_norm": 0.32732944937682,
|
| 2321 |
+
"kl": 0.00115966796875,
|
| 2322 |
+
"learning_rate": 2.918906036420294e-07,
|
| 2323 |
+
"loss": 0.0,
|
| 2324 |
+
"num_tokens": 9385103.0,
|
| 2325 |
+
"reward": 2.325900077819824,
|
| 2326 |
+
"reward_std": 0.8893687725067139,
|
| 2327 |
+
"rewards/accuracy_reward": 0.0,
|
| 2328 |
+
"rewards/format_reward": 0.90625,
|
| 2329 |
+
"rewards/tag_count_reward": 1.419650137424469,
|
| 2330 |
+
"step": 145
|
| 2331 |
+
},
|
| 2332 |
+
{
|
| 2333 |
+
"clip_ratio": 0.0,
|
| 2334 |
+
"completion_length": 2023.59375,
|
| 2335 |
+
"epoch": 0.006230396654362345,
|
| 2336 |
+
"grad_norm": 0.3367746029919756,
|
| 2337 |
+
"kl": 0.001140594482421875,
|
| 2338 |
+
"learning_rate": 2.854966364683872e-07,
|
| 2339 |
+
"loss": -0.0037,
|
| 2340 |
+
"num_tokens": 9455522.0,
|
| 2341 |
+
"reward": 2.186673879623413,
|
| 2342 |
+
"reward_std": 1.0466557741165161,
|
| 2343 |
+
"rewards/accuracy_reward": 0.0,
|
| 2344 |
+
"rewards/format_reward": 0.9375,
|
| 2345 |
+
"rewards/tag_count_reward": 1.249173879623413,
|
| 2346 |
+
"step": 146
|
| 2347 |
+
},
|
| 2348 |
+
{
|
| 2349 |
+
"clip_ratio": 0.0,
|
| 2350 |
+
"completion_length": 2048.0,
|
| 2351 |
+
"epoch": 0.006273070604049758,
|
| 2352 |
+
"grad_norm": 0.3425981108832875,
|
| 2353 |
+
"kl": 0.001125335693359375,
|
| 2354 |
+
"learning_rate": 2.791832395815782e-07,
|
| 2355 |
+
"loss": 0.0,
|
| 2356 |
+
"num_tokens": 9527178.0,
|
| 2357 |
+
"reward": 2.1425520181655884,
|
| 2358 |
+
"reward_std": 1.1318203508853912,
|
| 2359 |
+
"rewards/accuracy_reward": 0.03125,
|
| 2360 |
+
"rewards/format_reward": 0.90625,
|
| 2361 |
+
"rewards/tag_count_reward": 1.1738020181655884,
|
| 2362 |
+
"step": 147
|
| 2363 |
+
},
|
| 2364 |
+
{
|
| 2365 |
+
"clip_ratio": 0.0,
|
| 2366 |
+
"completion_length": 2048.0,
|
| 2367 |
+
"epoch": 0.0063157445537371714,
|
| 2368 |
+
"grad_norm": 0.34117424918888717,
|
| 2369 |
+
"kl": 0.001178741455078125,
|
| 2370 |
+
"learning_rate": 2.729523361034538e-07,
|
| 2371 |
+
"loss": 0.0,
|
| 2372 |
+
"num_tokens": 9597994.0,
|
| 2373 |
+
"reward": 2.022543430328369,
|
| 2374 |
+
"reward_std": 1.2339077591896057,
|
| 2375 |
+
"rewards/accuracy_reward": 0.0,
|
| 2376 |
+
"rewards/format_reward": 0.875,
|
| 2377 |
+
"rewards/tag_count_reward": 1.1475434303283691,
|
| 2378 |
+
"step": 148
|
| 2379 |
+
},
|
| 2380 |
+
{
|
| 2381 |
+
"clip_ratio": 0.0,
|
| 2382 |
+
"completion_length": 2013.46875,
|
| 2383 |
+
"epoch": 0.006358418503424584,
|
| 2384 |
+
"grad_norm": 0.3171263272765558,
|
| 2385 |
+
"kl": 0.001087188720703125,
|
| 2386 |
+
"learning_rate": 2.6680582402757324e-07,
|
| 2387 |
+
"loss": 0.0068,
|
| 2388 |
+
"num_tokens": 9667689.0,
|
| 2389 |
+
"reward": 2.9725751876831055,
|
| 2390 |
+
"reward_std": 0.8952206969261169,
|
| 2391 |
+
"rewards/accuracy_reward": 0.0625,
|
| 2392 |
+
"rewards/format_reward": 0.96875,
|
| 2393 |
+
"rewards/tag_count_reward": 1.8788251280784607,
|
| 2394 |
+
"step": 149
|
| 2395 |
+
},
|
| 2396 |
+
{
|
| 2397 |
+
"clip_ratio": 0.0,
|
| 2398 |
+
"completion_length": 2048.0,
|
| 2399 |
+
"epoch": 0.006401092453111998,
|
| 2400 |
+
"grad_norm": 0.3399321484253172,
|
| 2401 |
+
"kl": 0.001422882080078125,
|
| 2402 |
+
"learning_rate": 2.6074557564105724e-07,
|
| 2403 |
+
"loss": 0.0001,
|
| 2404 |
+
"num_tokens": 9739009.0,
|
| 2405 |
+
"reward": 2.5678025484085083,
|
| 2406 |
+
"reward_std": 1.3405039310455322,
|
| 2407 |
+
"rewards/accuracy_reward": 0.0,
|
| 2408 |
+
"rewards/format_reward": 0.8125,
|
| 2409 |
+
"rewards/tag_count_reward": 1.7553024888038635,
|
| 2410 |
+
"step": 150
|
| 2411 |
+
},
|
| 2412 |
+
{
|
| 2413 |
+
"clip_ratio": 0.0,
|
| 2414 |
+
"completion_length": 2048.0,
|
| 2415 |
+
"epoch": 0.006443766402799411,
|
| 2416 |
+
"grad_norm": 0.35099517362770644,
|
| 2417 |
+
"kl": 0.001190185546875,
|
| 2418 |
+
"learning_rate": 2.547734369542718e-07,
|
| 2419 |
+
"loss": 0.0,
|
| 2420 |
+
"num_tokens": 9810873.0,
|
| 2421 |
+
"reward": 2.2431211471557617,
|
| 2422 |
+
"reward_std": 1.3722872734069824,
|
| 2423 |
+
"rewards/accuracy_reward": 0.0,
|
| 2424 |
+
"rewards/format_reward": 0.875,
|
| 2425 |
+
"rewards/tag_count_reward": 1.3681212067604065,
|
| 2426 |
+
"step": 151
|
| 2427 |
+
},
|
| 2428 |
+
{
|
| 2429 |
+
"clip_ratio": 0.0,
|
| 2430 |
+
"completion_length": 2048.0,
|
| 2431 |
+
"epoch": 0.006486440352486824,
|
| 2432 |
+
"grad_norm": 0.33488383607136546,
|
| 2433 |
+
"kl": 0.00116729736328125,
|
| 2434 |
+
"learning_rate": 2.488912271385139e-07,
|
| 2435 |
+
"loss": 0.0,
|
| 2436 |
+
"num_tokens": 9882217.0,
|
| 2437 |
+
"reward": 1.8149373531341553,
|
| 2438 |
+
"reward_std": 1.1559866666793823,
|
| 2439 |
+
"rewards/accuracy_reward": 0.0,
|
| 2440 |
+
"rewards/format_reward": 0.8125,
|
| 2441 |
+
"rewards/tag_count_reward": 1.0024373531341553,
|
| 2442 |
+
"step": 152
|
| 2443 |
+
},
|
| 2444 |
+
{
|
| 2445 |
+
"clip_ratio": 0.0,
|
| 2446 |
+
"completion_length": 1994.5,
|
| 2447 |
+
"epoch": 0.006529114302174238,
|
| 2448 |
+
"grad_norm": 0.33504272985481237,
|
| 2449 |
+
"kl": 0.00128936767578125,
|
| 2450 |
+
"learning_rate": 2.4310073797187573e-07,
|
| 2451 |
+
"loss": 0.0045,
|
| 2452 |
+
"num_tokens": 9952529.0,
|
| 2453 |
+
"reward": 2.2245644330978394,
|
| 2454 |
+
"reward_std": 1.139819324016571,
|
| 2455 |
+
"rewards/accuracy_reward": 0.0,
|
| 2456 |
+
"rewards/format_reward": 0.875,
|
| 2457 |
+
"rewards/tag_count_reward": 1.349564552307129,
|
| 2458 |
+
"step": 153
|
| 2459 |
+
},
|
| 2460 |
+
{
|
| 2461 |
+
"clip_ratio": 0.0,
|
| 2462 |
+
"completion_length": 1797.4375,
|
| 2463 |
+
"epoch": 0.006571788251861651,
|
| 2464 |
+
"grad_norm": 0.35894851237733527,
|
| 2465 |
+
"kl": 0.00115203857421875,
|
| 2466 |
+
"learning_rate": 2.374037332934512e-07,
|
| 2467 |
+
"loss": -0.0123,
|
| 2468 |
+
"num_tokens": 10016175.0,
|
| 2469 |
+
"reward": 3.7275938987731934,
|
| 2470 |
+
"reward_std": 1.2432032227516174,
|
| 2471 |
+
"rewards/accuracy_reward": 0.4375,
|
| 2472 |
+
"rewards/format_reward": 0.9375,
|
| 2473 |
+
"rewards/tag_count_reward": 1.9150938987731934,
|
| 2474 |
+
"step": 154
|
| 2475 |
+
},
|
| 2476 |
+
{
|
| 2477 |
+
"clip_ratio": 0.0,
|
| 2478 |
+
"completion_length": 1954.71875,
|
| 2479 |
+
"epoch": 0.0066144622015490645,
|
| 2480 |
+
"grad_norm": 0.3785393538532397,
|
| 2481 |
+
"kl": 0.00127410888671875,
|
| 2482 |
+
"learning_rate": 2.3180194846605364e-07,
|
| 2483 |
+
"loss": 0.0306,
|
| 2484 |
+
"num_tokens": 10083918.0,
|
| 2485 |
+
"reward": 3.143017292022705,
|
| 2486 |
+
"reward_std": 1.1609839797019958,
|
| 2487 |
+
"rewards/accuracy_reward": 0.03125,
|
| 2488 |
+
"rewards/format_reward": 0.96875,
|
| 2489 |
+
"rewards/tag_count_reward": 2.111767292022705,
|
| 2490 |
+
"step": 155
|
| 2491 |
+
},
|
| 2492 |
+
{
|
| 2493 |
+
"clip_ratio": 0.0,
|
| 2494 |
+
"completion_length": 2048.0,
|
| 2495 |
+
"epoch": 0.0066571361512364775,
|
| 2496 |
+
"grad_norm": 0.3538179953006686,
|
| 2497 |
+
"kl": 0.001293182373046875,
|
| 2498 |
+
"learning_rate": 2.2629708984760706e-07,
|
| 2499 |
+
"loss": 0.0001,
|
| 2500 |
+
"num_tokens": 10154846.0,
|
| 2501 |
+
"reward": 2.255380153656006,
|
| 2502 |
+
"reward_std": 1.1156858205795288,
|
| 2503 |
+
"rewards/accuracy_reward": 0.0,
|
| 2504 |
+
"rewards/format_reward": 0.90625,
|
| 2505 |
+
"rewards/tag_count_reward": 1.3491302728652954,
|
| 2506 |
+
"step": 156
|
| 2507 |
+
},
|
| 2508 |
+
{
|
| 2509 |
+
"clip_ratio": 0.0,
|
| 2510 |
+
"completion_length": 2048.0,
|
| 2511 |
+
"epoch": 0.006699810100923891,
|
| 2512 |
+
"grad_norm": 0.3630369815797508,
|
| 2513 |
+
"kl": 0.0011653900146484375,
|
| 2514 |
+
"learning_rate": 2.2089083427137329e-07,
|
| 2515 |
+
"loss": 0.0,
|
| 2516 |
+
"num_tokens": 10226038.0,
|
| 2517 |
+
"reward": 1.8460949659347534,
|
| 2518 |
+
"reward_std": 1.3213353753089905,
|
| 2519 |
+
"rewards/accuracy_reward": 0.0,
|
| 2520 |
+
"rewards/format_reward": 0.875,
|
| 2521 |
+
"rewards/tag_count_reward": 0.9710949063301086,
|
| 2522 |
+
"step": 157
|
| 2523 |
+
},
|
| 2524 |
+
{
|
| 2525 |
+
"clip_ratio": 0.0,
|
| 2526 |
+
"completion_length": 2045.21875,
|
| 2527 |
+
"epoch": 0.006742484050611304,
|
| 2528 |
+
"grad_norm": 0.32015027980782584,
|
| 2529 |
+
"kl": 0.001232147216796875,
|
| 2530 |
+
"learning_rate": 2.1558482853517253e-07,
|
| 2531 |
+
"loss": 0.0005,
|
| 2532 |
+
"num_tokens": 10298725.0,
|
| 2533 |
+
"reward": 2.849201202392578,
|
| 2534 |
+
"reward_std": 1.5960667133331299,
|
| 2535 |
+
"rewards/accuracy_reward": 0.09375,
|
| 2536 |
+
"rewards/format_reward": 0.6875,
|
| 2537 |
+
"rewards/tag_count_reward": 1.9742012023925781,
|
| 2538 |
+
"step": 158
|
| 2539 |
+
},
|
| 2540 |
+
{
|
| 2541 |
+
"clip_ratio": 0.0,
|
| 2542 |
+
"completion_length": 2048.0,
|
| 2543 |
+
"epoch": 0.006785158000298718,
|
| 2544 |
+
"grad_norm": 0.3557371248900319,
|
| 2545 |
+
"kl": 0.001220703125,
|
| 2546 |
+
"learning_rate": 2.1038068889975259e-07,
|
| 2547 |
+
"loss": 0.0,
|
| 2548 |
+
"num_tokens": 10369877.0,
|
| 2549 |
+
"reward": 1.9885050058364868,
|
| 2550 |
+
"reward_std": 0.937650203704834,
|
| 2551 |
+
"rewards/accuracy_reward": 0.0,
|
| 2552 |
+
"rewards/format_reward": 0.8125,
|
| 2553 |
+
"rewards/tag_count_reward": 1.1760048866271973,
|
| 2554 |
+
"step": 159
|
| 2555 |
+
},
|
| 2556 |
+
{
|
| 2557 |
+
"clip_ratio": 0.0,
|
| 2558 |
+
"completion_length": 2026.96875,
|
| 2559 |
+
"epoch": 0.006827831949986131,
|
| 2560 |
+
"grad_norm": 0.3409616985369697,
|
| 2561 |
+
"kl": 0.001262664794921875,
|
| 2562 |
+
"learning_rate": 2.0528000059645995e-07,
|
| 2563 |
+
"loss": 0.0057,
|
| 2564 |
+
"num_tokens": 10439572.0,
|
| 2565 |
+
"reward": 2.9818347692489624,
|
| 2566 |
+
"reward_std": 1.221041351556778,
|
| 2567 |
+
"rewards/accuracy_reward": 0.15625,
|
| 2568 |
+
"rewards/format_reward": 0.90625,
|
| 2569 |
+
"rewards/tag_count_reward": 1.763084590435028,
|
| 2570 |
+
"step": 160
|
| 2571 |
+
},
|
| 2572 |
+
{
|
| 2573 |
+
"clip_ratio": 0.0,
|
| 2574 |
+
"completion_length": 2048.0,
|
| 2575 |
+
"epoch": 0.006870505899673545,
|
| 2576 |
+
"grad_norm": 0.3100710621108934,
|
| 2577 |
+
"kl": 0.00110626220703125,
|
| 2578 |
+
"learning_rate": 2.0028431734436308e-07,
|
| 2579 |
+
"loss": 0.0,
|
| 2580 |
+
"num_tokens": 10509972.0,
|
| 2581 |
+
"reward": 2.510861039161682,
|
| 2582 |
+
"reward_std": 1.3879331350326538,
|
| 2583 |
+
"rewards/accuracy_reward": 0.0,
|
| 2584 |
+
"rewards/format_reward": 0.84375,
|
| 2585 |
+
"rewards/tag_count_reward": 1.6671110391616821,
|
| 2586 |
+
"step": 161
|
| 2587 |
+
},
|
| 2588 |
+
{
|
| 2589 |
+
"clip_ratio": 0.0,
|
| 2590 |
+
"completion_length": 1879.46875,
|
| 2591 |
+
"epoch": 0.006913179849360958,
|
| 2592 |
+
"grad_norm": 0.34042820711135796,
|
| 2593 |
+
"kl": 0.0010318756103515625,
|
| 2594 |
+
"learning_rate": 1.9539516087697517e-07,
|
| 2595 |
+
"loss": -0.0109,
|
| 2596 |
+
"num_tokens": 10577259.0,
|
| 2597 |
+
"reward": 2.861981987953186,
|
| 2598 |
+
"reward_std": 0.8765715956687927,
|
| 2599 |
+
"rewards/accuracy_reward": 0.03125,
|
| 2600 |
+
"rewards/format_reward": 0.9375,
|
| 2601 |
+
"rewards/tag_count_reward": 1.8619821071624756,
|
| 2602 |
+
"step": 162
|
| 2603 |
+
},
|
| 2604 |
+
{
|
| 2605 |
+
"clip_ratio": 0.0,
|
| 2606 |
+
"completion_length": 2048.0,
|
| 2607 |
+
"epoch": 0.0069558537990483706,
|
| 2608 |
+
"grad_norm": 0.3441185180414332,
|
| 2609 |
+
"kl": 0.00124359130859375,
|
| 2610 |
+
"learning_rate": 1.9061402047871833e-07,
|
| 2611 |
+
"loss": 0.0,
|
| 2612 |
+
"num_tokens": 10647747.0,
|
| 2613 |
+
"reward": 2.2926758527755737,
|
| 2614 |
+
"reward_std": 0.9687027335166931,
|
| 2615 |
+
"rewards/accuracy_reward": 0.0,
|
| 2616 |
+
"rewards/format_reward": 0.9375,
|
| 2617 |
+
"rewards/tag_count_reward": 1.355175793170929,
|
| 2618 |
+
"step": 163
|
| 2619 |
+
},
|
| 2620 |
+
{
|
| 2621 |
+
"clip_ratio": 0.0,
|
| 2622 |
+
"completion_length": 2048.0,
|
| 2623 |
+
"epoch": 0.006998527748735784,
|
| 2624 |
+
"grad_norm": 0.2983579777199048,
|
| 2625 |
+
"kl": 0.001068115234375,
|
| 2626 |
+
"learning_rate": 1.8594235253127372e-07,
|
| 2627 |
+
"loss": 0.0,
|
| 2628 |
+
"num_tokens": 10718667.0,
|
| 2629 |
+
"reward": 2.2883304357528687,
|
| 2630 |
+
"reward_std": 1.3085912466049194,
|
| 2631 |
+
"rewards/accuracy_reward": 0.03125,
|
| 2632 |
+
"rewards/format_reward": 0.90625,
|
| 2633 |
+
"rewards/tag_count_reward": 1.3195804357528687,
|
| 2634 |
+
"step": 164
|
| 2635 |
+
},
|
| 2636 |
+
{
|
| 2637 |
+
"clip_ratio": 0.0,
|
| 2638 |
+
"completion_length": 1919.0625,
|
| 2639 |
+
"epoch": 0.007041201698423197,
|
| 2640 |
+
"grad_norm": 0.3385692348992647,
|
| 2641 |
+
"kl": 0.0013275146484375,
|
| 2642 |
+
"learning_rate": 1.8138158006995363e-07,
|
| 2643 |
+
"loss": 0.0365,
|
| 2644 |
+
"num_tokens": 10786173.0,
|
| 2645 |
+
"reward": 3.073681950569153,
|
| 2646 |
+
"reward_std": 1.6071664094924927,
|
| 2647 |
+
"rewards/accuracy_reward": 0.25,
|
| 2648 |
+
"rewards/format_reward": 0.78125,
|
| 2649 |
+
"rewards/tag_count_reward": 1.7924319505691528,
|
| 2650 |
+
"step": 165
|
| 2651 |
+
},
|
| 2652 |
+
{
|
| 2653 |
+
"clip_ratio": 0.0,
|
| 2654 |
+
"completion_length": 1993.21875,
|
| 2655 |
+
"epoch": 0.007083875648110611,
|
| 2656 |
+
"grad_norm": 0.31481423841418293,
|
| 2657 |
+
"kl": 0.001148223876953125,
|
| 2658 |
+
"learning_rate": 1.7693309235023127e-07,
|
| 2659 |
+
"loss": 0.0042,
|
| 2660 |
+
"num_tokens": 10854628.0,
|
| 2661 |
+
"reward": 2.505576729774475,
|
| 2662 |
+
"reward_std": 1.592041254043579,
|
| 2663 |
+
"rewards/accuracy_reward": 0.0,
|
| 2664 |
+
"rewards/format_reward": 0.78125,
|
| 2665 |
+
"rewards/tag_count_reward": 1.724326729774475,
|
| 2666 |
+
"step": 166
|
| 2667 |
+
},
|
| 2668 |
+
{
|
| 2669 |
+
"clip_ratio": 0.0,
|
| 2670 |
+
"completion_length": 1828.40625,
|
| 2671 |
+
"epoch": 0.007126549597798024,
|
| 2672 |
+
"grad_norm": 0.3490637186685043,
|
| 2673 |
+
"kl": 0.001293182373046875,
|
| 2674 |
+
"learning_rate": 1.7259824442455923e-07,
|
| 2675 |
+
"loss": -0.0089,
|
| 2676 |
+
"num_tokens": 10919537.0,
|
| 2677 |
+
"reward": 2.9972105026245117,
|
| 2678 |
+
"reward_std": 1.2622793018817902,
|
| 2679 |
+
"rewards/accuracy_reward": 0.25,
|
| 2680 |
+
"rewards/format_reward": 0.8125,
|
| 2681 |
+
"rewards/tag_count_reward": 1.6847106218338013,
|
| 2682 |
+
"step": 167
|
| 2683 |
+
},
|
| 2684 |
+
{
|
| 2685 |
+
"clip_ratio": 0.0,
|
| 2686 |
+
"completion_length": 1878.90625,
|
| 2687 |
+
"epoch": 0.007169223547485438,
|
| 2688 |
+
"grad_norm": 0.371407986766817,
|
| 2689 |
+
"kl": 0.001331329345703125,
|
| 2690 |
+
"learning_rate": 1.6837835672960831e-07,
|
| 2691 |
+
"loss": 0.0256,
|
| 2692 |
+
"num_tokens": 10984766.0,
|
| 2693 |
+
"reward": 3.2406015396118164,
|
| 2694 |
+
"reward_std": 0.9913251399993896,
|
| 2695 |
+
"rewards/accuracy_reward": 0.21875,
|
| 2696 |
+
"rewards/format_reward": 1.0,
|
| 2697 |
+
"rewards/tag_count_reward": 1.803101360797882,
|
| 2698 |
+
"step": 168
|
| 2699 |
+
},
|
| 2700 |
+
{
|
| 2701 |
+
"clip_ratio": 0.0,
|
| 2702 |
+
"completion_length": 2014.9375,
|
| 2703 |
+
"epoch": 0.007211897497172851,
|
| 2704 |
+
"grad_norm": 0.3426853342397948,
|
| 2705 |
+
"kl": 0.001155853271484375,
|
| 2706 |
+
"learning_rate": 1.6427471468404952e-07,
|
| 2707 |
+
"loss": -0.0043,
|
| 2708 |
+
"num_tokens": 11055836.0,
|
| 2709 |
+
"reward": 2.9239946603775024,
|
| 2710 |
+
"reward_std": 0.9754577875137329,
|
| 2711 |
+
"rewards/accuracy_reward": 0.25,
|
| 2712 |
+
"rewards/format_reward": 0.9375,
|
| 2713 |
+
"rewards/tag_count_reward": 1.4864946603775024,
|
| 2714 |
+
"step": 169
|
| 2715 |
+
},
|
| 2716 |
+
{
|
| 2717 |
+
"clip_ratio": 0.0,
|
| 2718 |
+
"completion_length": 2026.6875,
|
| 2719 |
+
"epoch": 0.0072545714468602645,
|
| 2720 |
+
"grad_norm": 0.3262166299620366,
|
| 2721 |
+
"kl": 0.001125335693359375,
|
| 2722 |
+
"learning_rate": 1.6028856829700258e-07,
|
| 2723 |
+
"loss": 0.0014,
|
| 2724 |
+
"num_tokens": 11126026.0,
|
| 2725 |
+
"reward": 2.656955599784851,
|
| 2726 |
+
"reward_std": 1.0347934365272522,
|
| 2727 |
+
"rewards/accuracy_reward": 0.03125,
|
| 2728 |
+
"rewards/format_reward": 0.9375,
|
| 2729 |
+
"rewards/tag_count_reward": 1.6569556593894958,
|
| 2730 |
+
"step": 170
|
| 2731 |
+
},
|
| 2732 |
+
{
|
| 2733 |
+
"clip_ratio": 0.0,
|
| 2734 |
+
"completion_length": 2048.0,
|
| 2735 |
+
"epoch": 0.0036486616239744807,
|
| 2736 |
+
"grad_norm": 0.37937413535912834,
|
| 2737 |
+
"kl": 0.00127410888671875,
|
| 2738 |
+
"learning_rate": 1.526735832134828e-07,
|
| 2739 |
+
"loss": 0.0001,
|
| 2740 |
+
"num_tokens": 72240.0,
|
| 2741 |
+
"reward": 1.882001280784607,
|
| 2742 |
+
"reward_std": 1.2824007868766785,
|
| 2743 |
+
"rewards/accuracy_reward": 0.0,
|
| 2744 |
+
"rewards/format_reward": 0.75,
|
| 2745 |
+
"rewards/tag_count_reward": 1.1320013403892517,
|
| 2746 |
+
"step": 171
|
| 2747 |
+
},
|
| 2748 |
+
{
|
| 2749 |
+
"epoch": 0.0036486616239744807,
|
| 2750 |
+
"step": 171,
|
| 2751 |
+
"total_flos": 0.0,
|
| 2752 |
+
"train_loss": 2.9861142760828923e-07,
|
| 2753 |
+
"train_runtime": 124.7789,
|
| 2754 |
+
"train_samples_per_second": 38.468,
|
| 2755 |
+
"train_steps_per_second": 1.202
|
| 2756 |
+
}
|
| 2757 |
+
],
|
| 2758 |
+
"logging_steps": 1,
|
| 2759 |
+
"max_steps": 150,
|
| 2760 |
+
"num_input_tokens_seen": 0,
|
| 2761 |
+
"num_train_epochs": 1,
|
| 2762 |
+
"save_steps": 10,
|
| 2763 |
+
"stateful_callbacks": {
|
| 2764 |
+
"TrainerControl": {
|
| 2765 |
+
"args": {
|
| 2766 |
+
"should_epoch_stop": false,
|
| 2767 |
+
"should_evaluate": false,
|
| 2768 |
+
"should_log": false,
|
| 2769 |
+
"should_save": true,
|
| 2770 |
+
"should_training_stop": true
|
| 2771 |
+
},
|
| 2772 |
+
"attributes": {}
|
| 2773 |
+
}
|
| 2774 |
+
},
|
| 2775 |
+
"total_flos": 0.0,
|
| 2776 |
+
"train_batch_size": 4,
|
| 2777 |
+
"trial_name": null,
|
| 2778 |
+
"trial_params": null
|
| 2779 |
+
}
|