Model save
Browse files- README.md +68 -0
- all_results.json +8 -0
- config.json +1 -1
- generation_config.json +14 -0
- train_results.json +8 -0
- trainer_state.json +2967 -0
README.md
ADDED
|
@@ -0,0 +1,68 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
base_model: Qwen/Qwen2.5-3B-Instruct
|
| 3 |
+
library_name: transformers
|
| 4 |
+
model_name: qwen-2.5-3b-r1-countdown
|
| 5 |
+
tags:
|
| 6 |
+
- generated_from_trainer
|
| 7 |
+
- trl
|
| 8 |
+
- grpo
|
| 9 |
+
licence: license
|
| 10 |
+
---
|
| 11 |
+
|
| 12 |
+
# Model Card for qwen-2.5-3b-r1-countdown
|
| 13 |
+
|
| 14 |
+
This model is a fine-tuned version of [Qwen/Qwen2.5-3B-Instruct](https://huggingface.co/Qwen/Qwen2.5-3B-Instruct).
|
| 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="gpandrad/qwen-2.5-3b-r1-countdown", 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 |
+
|
| 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: 0.14.0
|
| 38 |
+
- Transformers: 4.48.2
|
| 39 |
+
- Pytorch: 2.5.1
|
| 40 |
+
- Datasets: 3.2.0
|
| 41 |
+
- Tokenizers: 0.21.0
|
| 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": 2.9823833263829473e-05,
|
| 4 |
+
"train_runtime": 18161.8968,
|
| 5 |
+
"train_samples": 45000,
|
| 6 |
+
"train_samples_per_second": 1.388,
|
| 7 |
+
"train_steps_per_second": 0.025
|
| 8 |
+
}
|
config.json
CHANGED
|
@@ -23,7 +23,7 @@
|
|
| 23 |
"tie_word_embeddings": true,
|
| 24 |
"torch_dtype": "bfloat16",
|
| 25 |
"transformers_version": "4.48.2",
|
| 26 |
-
"use_cache":
|
| 27 |
"use_sliding_window": false,
|
| 28 |
"vocab_size": 151936
|
| 29 |
}
|
|
|
|
| 23 |
"tie_word_embeddings": true,
|
| 24 |
"torch_dtype": "bfloat16",
|
| 25 |
"transformers_version": "4.48.2",
|
| 26 |
+
"use_cache": true,
|
| 27 |
"use_sliding_window": false,
|
| 28 |
"vocab_size": 151936
|
| 29 |
}
|
generation_config.json
ADDED
|
@@ -0,0 +1,14 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bos_token_id": 151643,
|
| 3 |
+
"do_sample": true,
|
| 4 |
+
"eos_token_id": [
|
| 5 |
+
151645,
|
| 6 |
+
151643
|
| 7 |
+
],
|
| 8 |
+
"pad_token_id": 151643,
|
| 9 |
+
"repetition_penalty": 1.05,
|
| 10 |
+
"temperature": 0.7,
|
| 11 |
+
"top_k": 20,
|
| 12 |
+
"top_p": 0.8,
|
| 13 |
+
"transformers_version": "4.48.2"
|
| 14 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"total_flos": 0.0,
|
| 3 |
+
"train_loss": 2.9823833263829473e-05,
|
| 4 |
+
"train_runtime": 18161.8968,
|
| 5 |
+
"train_samples": 45000,
|
| 6 |
+
"train_samples_per_second": 1.388,
|
| 7 |
+
"train_steps_per_second": 0.025
|
| 8 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,2967 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 0.5599626691553896,
|
| 5 |
+
"eval_steps": 500,
|
| 6 |
+
"global_step": 450,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"completion_length": 400.44086837768555,
|
| 13 |
+
"epoch": 0.002488722974023954,
|
| 14 |
+
"grad_norm": 0.10229702522786686,
|
| 15 |
+
"kl": 0.0,
|
| 16 |
+
"learning_rate": 7.142857142857142e-08,
|
| 17 |
+
"loss": 0.0,
|
| 18 |
+
"reward": 0.5412946753203869,
|
| 19 |
+
"reward_std": 0.5055910516530275,
|
| 20 |
+
"rewards/equation_reward_func": 0.0,
|
| 21 |
+
"rewards/format_reward_func": 0.5412946753203869,
|
| 22 |
+
"step": 2
|
| 23 |
+
},
|
| 24 |
+
{
|
| 25 |
+
"completion_length": 385.4564914703369,
|
| 26 |
+
"epoch": 0.004977445948047908,
|
| 27 |
+
"grad_norm": 0.10059111736111713,
|
| 28 |
+
"kl": 0.0004247426986694336,
|
| 29 |
+
"learning_rate": 1.4285714285714285e-07,
|
| 30 |
+
"loss": 0.0,
|
| 31 |
+
"reward": 0.47544645331799984,
|
| 32 |
+
"reward_std": 0.4911654181778431,
|
| 33 |
+
"rewards/equation_reward_func": 0.0,
|
| 34 |
+
"rewards/format_reward_func": 0.47544645331799984,
|
| 35 |
+
"step": 4
|
| 36 |
+
},
|
| 37 |
+
{
|
| 38 |
+
"completion_length": 389.0725612640381,
|
| 39 |
+
"epoch": 0.007466168922071862,
|
| 40 |
+
"grad_norm": 0.11061618345240035,
|
| 41 |
+
"kl": 0.0004165172576904297,
|
| 42 |
+
"learning_rate": 2.1428571428571426e-07,
|
| 43 |
+
"loss": 0.0,
|
| 44 |
+
"reward": 0.5279018115252256,
|
| 45 |
+
"reward_std": 0.49625692516565323,
|
| 46 |
+
"rewards/equation_reward_func": 0.0,
|
| 47 |
+
"rewards/format_reward_func": 0.5279018115252256,
|
| 48 |
+
"step": 6
|
| 49 |
+
},
|
| 50 |
+
{
|
| 51 |
+
"completion_length": 396.0189895629883,
|
| 52 |
+
"epoch": 0.009954891896095816,
|
| 53 |
+
"grad_norm": 0.11927608013833589,
|
| 54 |
+
"kl": 0.00040137767791748047,
|
| 55 |
+
"learning_rate": 2.857142857142857e-07,
|
| 56 |
+
"loss": 0.0,
|
| 57 |
+
"reward": 0.4854910969734192,
|
| 58 |
+
"reward_std": 0.4980995524674654,
|
| 59 |
+
"rewards/equation_reward_func": 0.0,
|
| 60 |
+
"rewards/format_reward_func": 0.4854910969734192,
|
| 61 |
+
"step": 8
|
| 62 |
+
},
|
| 63 |
+
{
|
| 64 |
+
"completion_length": 392.9497928619385,
|
| 65 |
+
"epoch": 0.01244361487011977,
|
| 66 |
+
"grad_norm": 0.10949296556166527,
|
| 67 |
+
"kl": 0.0004659891128540039,
|
| 68 |
+
"learning_rate": 3.5714285714285716e-07,
|
| 69 |
+
"loss": 0.0,
|
| 70 |
+
"reward": 0.5334821622818708,
|
| 71 |
+
"reward_std": 0.483795752748847,
|
| 72 |
+
"rewards/equation_reward_func": 0.0,
|
| 73 |
+
"rewards/format_reward_func": 0.5334821622818708,
|
| 74 |
+
"step": 10
|
| 75 |
+
},
|
| 76 |
+
{
|
| 77 |
+
"completion_length": 361.7143039703369,
|
| 78 |
+
"epoch": 0.014932337844143724,
|
| 79 |
+
"grad_norm": 0.11491431070654654,
|
| 80 |
+
"kl": 0.0004961490631103516,
|
| 81 |
+
"learning_rate": 4.285714285714285e-07,
|
| 82 |
+
"loss": 0.0,
|
| 83 |
+
"reward": 0.5725446715950966,
|
| 84 |
+
"reward_std": 0.49694120697677135,
|
| 85 |
+
"rewards/equation_reward_func": 0.0011160714784637094,
|
| 86 |
+
"rewards/format_reward_func": 0.5714285969734192,
|
| 87 |
+
"step": 12
|
| 88 |
+
},
|
| 89 |
+
{
|
| 90 |
+
"completion_length": 383.80359077453613,
|
| 91 |
+
"epoch": 0.017421060818167678,
|
| 92 |
+
"grad_norm": 0.10217225238845459,
|
| 93 |
+
"kl": 0.0006783008575439453,
|
| 94 |
+
"learning_rate": 5e-07,
|
| 95 |
+
"loss": 0.0,
|
| 96 |
+
"reward": 0.6261161006987095,
|
| 97 |
+
"reward_std": 0.47387135215103626,
|
| 98 |
+
"rewards/equation_reward_func": 0.0,
|
| 99 |
+
"rewards/format_reward_func": 0.6261161006987095,
|
| 100 |
+
"step": 14
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"completion_length": 385.1930961608887,
|
| 104 |
+
"epoch": 0.019909783792191632,
|
| 105 |
+
"grad_norm": 0.09079397444986828,
|
| 106 |
+
"kl": 0.0009381771087646484,
|
| 107 |
+
"learning_rate": 4.999740409224932e-07,
|
| 108 |
+
"loss": 0.0,
|
| 109 |
+
"reward": 0.6529018133878708,
|
| 110 |
+
"reward_std": 0.46884808875620365,
|
| 111 |
+
"rewards/equation_reward_func": 0.0,
|
| 112 |
+
"rewards/format_reward_func": 0.6529018133878708,
|
| 113 |
+
"step": 16
|
| 114 |
+
},
|
| 115 |
+
{
|
| 116 |
+
"completion_length": 387.5089473724365,
|
| 117 |
+
"epoch": 0.022398506766215586,
|
| 118 |
+
"grad_norm": 0.07789924771682649,
|
| 119 |
+
"kl": 0.0022106170654296875,
|
| 120 |
+
"learning_rate": 4.998961690809627e-07,
|
| 121 |
+
"loss": 0.0,
|
| 122 |
+
"reward": 0.8169643208384514,
|
| 123 |
+
"reward_std": 0.35768389888107777,
|
| 124 |
+
"rewards/equation_reward_func": 0.0,
|
| 125 |
+
"rewards/format_reward_func": 0.8169643208384514,
|
| 126 |
+
"step": 18
|
| 127 |
+
},
|
| 128 |
+
{
|
| 129 |
+
"completion_length": 380.96653938293457,
|
| 130 |
+
"epoch": 0.02488722974023954,
|
| 131 |
+
"grad_norm": 0.06315055485645503,
|
| 132 |
+
"kl": 0.004170417785644531,
|
| 133 |
+
"learning_rate": 4.997664006472578e-07,
|
| 134 |
+
"loss": 0.0,
|
| 135 |
+
"reward": 0.887276828289032,
|
| 136 |
+
"reward_std": 0.23898791754618287,
|
| 137 |
+
"rewards/equation_reward_func": 0.0,
|
| 138 |
+
"rewards/format_reward_func": 0.887276828289032,
|
| 139 |
+
"step": 20
|
| 140 |
+
},
|
| 141 |
+
{
|
| 142 |
+
"completion_length": 378.5915355682373,
|
| 143 |
+
"epoch": 0.027375952714263494,
|
| 144 |
+
"grad_norm": 0.0566780627533256,
|
| 145 |
+
"kl": 0.005069732666015625,
|
| 146 |
+
"learning_rate": 4.995847625707292e-07,
|
| 147 |
+
"loss": 0.0,
|
| 148 |
+
"reward": 0.9129464663565159,
|
| 149 |
+
"reward_std": 0.19532618205994368,
|
| 150 |
+
"rewards/equation_reward_func": 0.0,
|
| 151 |
+
"rewards/format_reward_func": 0.9129464663565159,
|
| 152 |
+
"step": 22
|
| 153 |
+
},
|
| 154 |
+
{
|
| 155 |
+
"completion_length": 385.043550491333,
|
| 156 |
+
"epoch": 0.029864675688287448,
|
| 157 |
+
"grad_norm": 0.0507404366996494,
|
| 158 |
+
"kl": 0.005603790283203125,
|
| 159 |
+
"learning_rate": 4.993512925726318e-07,
|
| 160 |
+
"loss": 0.0,
|
| 161 |
+
"reward": 0.9107143208384514,
|
| 162 |
+
"reward_std": 0.20043640863150358,
|
| 163 |
+
"rewards/equation_reward_func": 0.0,
|
| 164 |
+
"rewards/format_reward_func": 0.9107143208384514,
|
| 165 |
+
"step": 24
|
| 166 |
+
},
|
| 167 |
+
{
|
| 168 |
+
"completion_length": 374.24667167663574,
|
| 169 |
+
"epoch": 0.0323533986623114,
|
| 170 |
+
"grad_norm": 0.03460311969195311,
|
| 171 |
+
"kl": 0.0071620941162109375,
|
| 172 |
+
"learning_rate": 4.990660391382923e-07,
|
| 173 |
+
"loss": 0.0,
|
| 174 |
+
"reward": 0.9419643171131611,
|
| 175 |
+
"reward_std": 0.1277988995425403,
|
| 176 |
+
"rewards/equation_reward_func": 0.0,
|
| 177 |
+
"rewards/format_reward_func": 0.9419643171131611,
|
| 178 |
+
"step": 26
|
| 179 |
+
},
|
| 180 |
+
{
|
| 181 |
+
"completion_length": 356.6495723724365,
|
| 182 |
+
"epoch": 0.034842121636335356,
|
| 183 |
+
"grad_norm": 0.014238488177721704,
|
| 184 |
+
"kl": 0.010402679443359375,
|
| 185 |
+
"learning_rate": 4.987290615070384e-07,
|
| 186 |
+
"loss": 0.0,
|
| 187 |
+
"reward": 0.9732143245637417,
|
| 188 |
+
"reward_std": 0.07358111487701535,
|
| 189 |
+
"rewards/equation_reward_func": 0.0,
|
| 190 |
+
"rewards/format_reward_func": 0.9732143245637417,
|
| 191 |
+
"step": 28
|
| 192 |
+
},
|
| 193 |
+
{
|
| 194 |
+
"completion_length": 356.62501525878906,
|
| 195 |
+
"epoch": 0.03733084461035931,
|
| 196 |
+
"grad_norm": 0.016695682547454543,
|
| 197 |
+
"kl": 0.013233184814453125,
|
| 198 |
+
"learning_rate": 4.983404296598978e-07,
|
| 199 |
+
"loss": 0.0,
|
| 200 |
+
"reward": 0.9799107350409031,
|
| 201 |
+
"reward_std": 0.0502800983376801,
|
| 202 |
+
"rewards/equation_reward_func": 0.0,
|
| 203 |
+
"rewards/format_reward_func": 0.9799107350409031,
|
| 204 |
+
"step": 30
|
| 205 |
+
},
|
| 206 |
+
{
|
| 207 |
+
"completion_length": 337.0390739440918,
|
| 208 |
+
"epoch": 0.039819567584383264,
|
| 209 |
+
"grad_norm": 0.0178363646878267,
|
| 210 |
+
"kl": 0.019924163818359375,
|
| 211 |
+
"learning_rate": 4.979002243050646e-07,
|
| 212 |
+
"loss": 0.0,
|
| 213 |
+
"reward": 0.9821428917348385,
|
| 214 |
+
"reward_std": 0.04832730023190379,
|
| 215 |
+
"rewards/equation_reward_func": 0.0,
|
| 216 |
+
"rewards/format_reward_func": 0.9821428917348385,
|
| 217 |
+
"step": 32
|
| 218 |
+
},
|
| 219 |
+
{
|
| 220 |
+
"completion_length": 329.28014945983887,
|
| 221 |
+
"epoch": 0.042308290558407215,
|
| 222 |
+
"grad_norm": 0.014714626628216123,
|
| 223 |
+
"kl": 0.016998291015625,
|
| 224 |
+
"learning_rate": 4.974085368611381e-07,
|
| 225 |
+
"loss": 0.0,
|
| 226 |
+
"reward": 0.9787946827709675,
|
| 227 |
+
"reward_std": 0.05561715178191662,
|
| 228 |
+
"rewards/equation_reward_func": 0.0,
|
| 229 |
+
"rewards/format_reward_func": 0.9787946827709675,
|
| 230 |
+
"step": 34
|
| 231 |
+
},
|
| 232 |
+
{
|
| 233 |
+
"completion_length": 322.30246925354004,
|
| 234 |
+
"epoch": 0.04479701353243117,
|
| 235 |
+
"grad_norm": 0.021040890462160396,
|
| 236 |
+
"kl": 0.01895904541015625,
|
| 237 |
+
"learning_rate": 4.968654694381379e-07,
|
| 238 |
+
"loss": 0.0,
|
| 239 |
+
"reward": 0.9743303917348385,
|
| 240 |
+
"reward_std": 0.06606373097747564,
|
| 241 |
+
"rewards/equation_reward_func": 0.0,
|
| 242 |
+
"rewards/format_reward_func": 0.9743303917348385,
|
| 243 |
+
"step": 36
|
| 244 |
+
},
|
| 245 |
+
{
|
| 246 |
+
"completion_length": 305.00335693359375,
|
| 247 |
+
"epoch": 0.04728573650645512,
|
| 248 |
+
"grad_norm": 0.01863566038382118,
|
| 249 |
+
"kl": 0.0214385986328125,
|
| 250 |
+
"learning_rate": 4.962711348162987e-07,
|
| 251 |
+
"loss": 0.0,
|
| 252 |
+
"reward": 0.9910714514553547,
|
| 253 |
+
"reward_std": 0.0252538132481277,
|
| 254 |
+
"rewards/equation_reward_func": 0.0,
|
| 255 |
+
"rewards/format_reward_func": 0.9910714514553547,
|
| 256 |
+
"step": 38
|
| 257 |
+
},
|
| 258 |
+
{
|
| 259 |
+
"completion_length": 310.13282775878906,
|
| 260 |
+
"epoch": 0.04977445948047908,
|
| 261 |
+
"grad_norm": 0.011919224174973252,
|
| 262 |
+
"kl": 0.0224151611328125,
|
| 263 |
+
"learning_rate": 4.956256564226487e-07,
|
| 264 |
+
"loss": 0.0,
|
| 265 |
+
"reward": 0.9854910969734192,
|
| 266 |
+
"reward_std": 0.0410374472849071,
|
| 267 |
+
"rewards/equation_reward_func": 0.0,
|
| 268 |
+
"rewards/format_reward_func": 0.9854910969734192,
|
| 269 |
+
"step": 40
|
| 270 |
+
},
|
| 271 |
+
{
|
| 272 |
+
"completion_length": 308.84375953674316,
|
| 273 |
+
"epoch": 0.05226318245450303,
|
| 274 |
+
"grad_norm": 0.00934945126695278,
|
| 275 |
+
"kl": 0.023162841796875,
|
| 276 |
+
"learning_rate": 4.949291683053768e-07,
|
| 277 |
+
"loss": 0.0,
|
| 278 |
+
"reward": 0.9899553768336773,
|
| 279 |
+
"reward_std": 0.028410539962351322,
|
| 280 |
+
"rewards/equation_reward_func": 0.0,
|
| 281 |
+
"rewards/format_reward_func": 0.9899553768336773,
|
| 282 |
+
"step": 42
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"completion_length": 285.6082706451416,
|
| 286 |
+
"epoch": 0.05475190542852699,
|
| 287 |
+
"grad_norm": 0.006489996788191869,
|
| 288 |
+
"kl": 0.0259552001953125,
|
| 289 |
+
"learning_rate": 4.941818151059955e-07,
|
| 290 |
+
"loss": 0.0,
|
| 291 |
+
"reward": 0.9921875149011612,
|
| 292 |
+
"reward_std": 0.019916759338229895,
|
| 293 |
+
"rewards/equation_reward_func": 0.0,
|
| 294 |
+
"rewards/format_reward_func": 0.9921875149011612,
|
| 295 |
+
"step": 44
|
| 296 |
+
},
|
| 297 |
+
{
|
| 298 |
+
"completion_length": 279.28461265563965,
|
| 299 |
+
"epoch": 0.05724062840255094,
|
| 300 |
+
"grad_norm": 0.011058147576593262,
|
| 301 |
+
"kl": 0.027862548828125,
|
| 302 |
+
"learning_rate": 4.933837520293017e-07,
|
| 303 |
+
"loss": 0.0,
|
| 304 |
+
"reward": 0.993303582072258,
|
| 305 |
+
"reward_std": 0.01676003262400627,
|
| 306 |
+
"rewards/equation_reward_func": 0.0,
|
| 307 |
+
"rewards/format_reward_func": 0.993303582072258,
|
| 308 |
+
"step": 46
|
| 309 |
+
},
|
| 310 |
+
{
|
| 311 |
+
"completion_length": 281.8884048461914,
|
| 312 |
+
"epoch": 0.059729351376574896,
|
| 313 |
+
"grad_norm": 0.0055576859057008985,
|
| 314 |
+
"kl": 0.02691650390625,
|
| 315 |
+
"learning_rate": 4.925351448111454e-07,
|
| 316 |
+
"loss": 0.0,
|
| 317 |
+
"reward": 0.9944196566939354,
|
| 318 |
+
"reward_std": 0.015783633571118116,
|
| 319 |
+
"rewards/equation_reward_func": 0.0,
|
| 320 |
+
"rewards/format_reward_func": 0.9944196566939354,
|
| 321 |
+
"step": 48
|
| 322 |
+
},
|
| 323 |
+
{
|
| 324 |
+
"completion_length": 288.05023765563965,
|
| 325 |
+
"epoch": 0.06221807435059885,
|
| 326 |
+
"grad_norm": 0.012749047747490404,
|
| 327 |
+
"kl": 0.027008056640625,
|
| 328 |
+
"learning_rate": 4.91636169684011e-07,
|
| 329 |
+
"loss": 0.0,
|
| 330 |
+
"reward": 0.994419664144516,
|
| 331 |
+
"reward_std": 0.015783633571118116,
|
| 332 |
+
"rewards/equation_reward_func": 0.0,
|
| 333 |
+
"rewards/format_reward_func": 0.994419664144516,
|
| 334 |
+
"step": 50
|
| 335 |
+
},
|
| 336 |
+
{
|
| 337 |
+
"completion_length": 286.1239013671875,
|
| 338 |
+
"epoch": 0.0647067973246228,
|
| 339 |
+
"grad_norm": 0.018013792950835963,
|
| 340 |
+
"kl": 0.03295135498046875,
|
| 341 |
+
"learning_rate": 4.906870133404186e-07,
|
| 342 |
+
"loss": 0.0,
|
| 343 |
+
"reward": 0.9866071566939354,
|
| 344 |
+
"reward_std": 0.03788072057068348,
|
| 345 |
+
"rewards/equation_reward_func": 0.0,
|
| 346 |
+
"rewards/format_reward_func": 0.9866071566939354,
|
| 347 |
+
"step": 52
|
| 348 |
+
},
|
| 349 |
+
{
|
| 350 |
+
"completion_length": 267.4062604904175,
|
| 351 |
+
"epoch": 0.06719552029864675,
|
| 352 |
+
"grad_norm": 0.022259449850372266,
|
| 353 |
+
"kl": 0.03186798095703125,
|
| 354 |
+
"learning_rate": 4.896878728941531e-07,
|
| 355 |
+
"loss": 0.0,
|
| 356 |
+
"reward": 0.991071455180645,
|
| 357 |
+
"reward_std": 0.025253813713788986,
|
| 358 |
+
"rewards/equation_reward_func": 0.0,
|
| 359 |
+
"rewards/format_reward_func": 0.991071455180645,
|
| 360 |
+
"step": 54
|
| 361 |
+
},
|
| 362 |
+
{
|
| 363 |
+
"completion_length": 270.542423248291,
|
| 364 |
+
"epoch": 0.06968424327267071,
|
| 365 |
+
"grad_norm": 0.01088790708734008,
|
| 366 |
+
"kl": 0.03081512451171875,
|
| 367 |
+
"learning_rate": 4.886389558393284e-07,
|
| 368 |
+
"loss": 0.0,
|
| 369 |
+
"reward": 0.9955357313156128,
|
| 370 |
+
"reward_std": 0.012626906856894493,
|
| 371 |
+
"rewards/equation_reward_func": 0.0,
|
| 372 |
+
"rewards/format_reward_func": 0.9955357313156128,
|
| 373 |
+
"step": 56
|
| 374 |
+
},
|
| 375 |
+
{
|
| 376 |
+
"completion_length": 277.4542579650879,
|
| 377 |
+
"epoch": 0.07217296624669467,
|
| 378 |
+
"grad_norm": 0.005367868177232952,
|
| 379 |
+
"kl": 0.033447265625,
|
| 380 |
+
"learning_rate": 4.875404800072976e-07,
|
| 381 |
+
"loss": 0.0,
|
| 382 |
+
"reward": 0.9955357238650322,
|
| 383 |
+
"reward_std": 0.012626906856894493,
|
| 384 |
+
"rewards/equation_reward_func": 0.0,
|
| 385 |
+
"rewards/format_reward_func": 0.9955357238650322,
|
| 386 |
+
"step": 58
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"completion_length": 266.13170623779297,
|
| 390 |
+
"epoch": 0.07466168922071861,
|
| 391 |
+
"grad_norm": 0.018763444577450522,
|
| 392 |
+
"kl": 0.0355224609375,
|
| 393 |
+
"learning_rate": 4.86392673521415e-07,
|
| 394 |
+
"loss": 0.0,
|
| 395 |
+
"reward": 0.9921875186264515,
|
| 396 |
+
"reward_std": 0.022097086533904076,
|
| 397 |
+
"rewards/equation_reward_func": 0.0,
|
| 398 |
+
"rewards/format_reward_func": 0.9921875186264515,
|
| 399 |
+
"step": 60
|
| 400 |
+
},
|
| 401 |
+
{
|
| 402 |
+
"completion_length": 264.5323791503906,
|
| 403 |
+
"epoch": 0.07715041219474257,
|
| 404 |
+
"grad_norm": 0.006685649273344385,
|
| 405 |
+
"kl": 0.032928466796875,
|
| 406 |
+
"learning_rate": 4.851957747496606e-07,
|
| 407 |
+
"loss": 0.0,
|
| 408 |
+
"reward": 0.9977678656578064,
|
| 409 |
+
"reward_std": 0.0063134534284472466,
|
| 410 |
+
"rewards/equation_reward_func": 0.0,
|
| 411 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 412 |
+
"step": 62
|
| 413 |
+
},
|
| 414 |
+
{
|
| 415 |
+
"completion_length": 269.7288055419922,
|
| 416 |
+
"epoch": 0.07963913516876653,
|
| 417 |
+
"grad_norm": 0.005778436612347605,
|
| 418 |
+
"kl": 0.0353546142578125,
|
| 419 |
+
"learning_rate": 4.839500322551386e-07,
|
| 420 |
+
"loss": 0.0,
|
| 421 |
+
"reward": 0.994419664144516,
|
| 422 |
+
"reward_std": 0.015783633571118116,
|
| 423 |
+
"rewards/equation_reward_func": 0.0,
|
| 424 |
+
"rewards/format_reward_func": 0.994419664144516,
|
| 425 |
+
"step": 64
|
| 426 |
+
},
|
| 427 |
+
{
|
| 428 |
+
"completion_length": 259.173002243042,
|
| 429 |
+
"epoch": 0.08212785814279049,
|
| 430 |
+
"grad_norm": 0.005211884618410732,
|
| 431 |
+
"kl": 0.03321075439453125,
|
| 432 |
+
"learning_rate": 4.826557047444563e-07,
|
| 433 |
+
"loss": 0.0,
|
| 434 |
+
"reward": 0.9988839328289032,
|
| 435 |
+
"reward_std": 0.0031567267142236233,
|
| 436 |
+
"rewards/equation_reward_func": 0.0,
|
| 437 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 438 |
+
"step": 66
|
| 439 |
+
},
|
| 440 |
+
{
|
| 441 |
+
"completion_length": 266.08595275878906,
|
| 442 |
+
"epoch": 0.08461658111681443,
|
| 443 |
+
"grad_norm": 0.008193740275267598,
|
| 444 |
+
"kl": 0.03182220458984375,
|
| 445 |
+
"learning_rate": 4.813130610139993e-07,
|
| 446 |
+
"loss": 0.0,
|
| 447 |
+
"reward": 0.9955357313156128,
|
| 448 |
+
"reward_std": 0.012626906856894493,
|
| 449 |
+
"rewards/equation_reward_func": 0.0,
|
| 450 |
+
"rewards/format_reward_func": 0.9955357313156128,
|
| 451 |
+
"step": 68
|
| 452 |
+
},
|
| 453 |
+
{
|
| 454 |
+
"completion_length": 264.5357275009155,
|
| 455 |
+
"epoch": 0.08710530409083839,
|
| 456 |
+
"grad_norm": 0.005236192803718056,
|
| 457 |
+
"kl": 0.03408050537109375,
|
| 458 |
+
"learning_rate": 4.799223798941089e-07,
|
| 459 |
+
"loss": 0.0,
|
| 460 |
+
"reward": 0.9944196492433548,
|
| 461 |
+
"reward_std": 0.015783633571118116,
|
| 462 |
+
"rewards/equation_reward_func": 0.0,
|
| 463 |
+
"rewards/format_reward_func": 0.9944196492433548,
|
| 464 |
+
"step": 70
|
| 465 |
+
},
|
| 466 |
+
{
|
| 467 |
+
"completion_length": 254.89398574829102,
|
| 468 |
+
"epoch": 0.08959402706486234,
|
| 469 |
+
"grad_norm": 0.012593840996977298,
|
| 470 |
+
"kl": 0.03564453125,
|
| 471 |
+
"learning_rate": 4.78483950191177e-07,
|
| 472 |
+
"loss": 0.0,
|
| 473 |
+
"reward": 0.9955357313156128,
|
| 474 |
+
"reward_std": 0.012626906856894493,
|
| 475 |
+
"rewards/equation_reward_func": 0.0,
|
| 476 |
+
"rewards/format_reward_func": 0.9955357313156128,
|
| 477 |
+
"step": 72
|
| 478 |
+
},
|
| 479 |
+
{
|
| 480 |
+
"completion_length": 258.49219703674316,
|
| 481 |
+
"epoch": 0.0920827500388863,
|
| 482 |
+
"grad_norm": 0.008244180466299824,
|
| 483 |
+
"kl": 0.03539276123046875,
|
| 484 |
+
"learning_rate": 4.769980706276687e-07,
|
| 485 |
+
"loss": 0.0,
|
| 486 |
+
"reward": 0.9955357238650322,
|
| 487 |
+
"reward_std": 0.010446579195559025,
|
| 488 |
+
"rewards/equation_reward_func": 0.0,
|
| 489 |
+
"rewards/format_reward_func": 0.9955357238650322,
|
| 490 |
+
"step": 74
|
| 491 |
+
},
|
| 492 |
+
{
|
| 493 |
+
"completion_length": 254.81027793884277,
|
| 494 |
+
"epoch": 0.09457147301291025,
|
| 495 |
+
"grad_norm": 0.015490694406396487,
|
| 496 |
+
"kl": 0.0354461669921875,
|
| 497 |
+
"learning_rate": 4.7546504978008595e-07,
|
| 498 |
+
"loss": 0.0,
|
| 499 |
+
"reward": 0.9910714477300644,
|
| 500 |
+
"reward_std": 0.025253813713788986,
|
| 501 |
+
"rewards/equation_reward_func": 0.0,
|
| 502 |
+
"rewards/format_reward_func": 0.9910714477300644,
|
| 503 |
+
"step": 76
|
| 504 |
+
},
|
| 505 |
+
{
|
| 506 |
+
"completion_length": 249.98550033569336,
|
| 507 |
+
"epoch": 0.0970601959869342,
|
| 508 |
+
"grad_norm": 0.01147964209306704,
|
| 509 |
+
"kl": 0.036407470703125,
|
| 510 |
+
"learning_rate": 4.738852060148848e-07,
|
| 511 |
+
"loss": 0.0,
|
| 512 |
+
"reward": 0.9977678656578064,
|
| 513 |
+
"reward_std": 0.0063134534284472466,
|
| 514 |
+
"rewards/equation_reward_func": 0.0,
|
| 515 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 516 |
+
"step": 78
|
| 517 |
+
},
|
| 518 |
+
{
|
| 519 |
+
"completion_length": 253.0022439956665,
|
| 520 |
+
"epoch": 0.09954891896095816,
|
| 521 |
+
"grad_norm": 0.007444099420668922,
|
| 522 |
+
"kl": 0.0356597900390625,
|
| 523 |
+
"learning_rate": 4.722588674223593e-07,
|
| 524 |
+
"loss": 0.0,
|
| 525 |
+
"reward": 0.9944196566939354,
|
| 526 |
+
"reward_std": 0.015783633571118116,
|
| 527 |
+
"rewards/equation_reward_func": 0.0,
|
| 528 |
+
"rewards/format_reward_func": 0.9944196566939354,
|
| 529 |
+
"step": 80
|
| 530 |
+
},
|
| 531 |
+
{
|
| 532 |
+
"completion_length": 257.6272449493408,
|
| 533 |
+
"epoch": 0.10203764193498212,
|
| 534 |
+
"grad_norm": 0.010780487057438663,
|
| 535 |
+
"kl": 0.03473663330078125,
|
| 536 |
+
"learning_rate": 4.70586371748506e-07,
|
| 537 |
+
"loss": 0.0,
|
| 538 |
+
"reward": 0.9966517984867096,
|
| 539 |
+
"reward_std": 0.00947018014267087,
|
| 540 |
+
"rewards/equation_reward_func": 0.0,
|
| 541 |
+
"rewards/format_reward_func": 0.9966517984867096,
|
| 542 |
+
"step": 82
|
| 543 |
+
},
|
| 544 |
+
{
|
| 545 |
+
"completion_length": 260.4788064956665,
|
| 546 |
+
"epoch": 0.10452636490900606,
|
| 547 |
+
"grad_norm": 0.011815984815212375,
|
| 548 |
+
"kl": 0.0367889404296875,
|
| 549 |
+
"learning_rate": 4.6886806632488363e-07,
|
| 550 |
+
"loss": 0.0,
|
| 551 |
+
"reward": 0.9955357313156128,
|
| 552 |
+
"reward_std": 0.012626906856894493,
|
| 553 |
+
"rewards/equation_reward_func": 0.0,
|
| 554 |
+
"rewards/format_reward_func": 0.9955357313156128,
|
| 555 |
+
"step": 84
|
| 556 |
+
},
|
| 557 |
+
{
|
| 558 |
+
"completion_length": 251.9944305419922,
|
| 559 |
+
"epoch": 0.10701508788303002,
|
| 560 |
+
"grad_norm": 0.008165684742827731,
|
| 561 |
+
"kl": 0.0371551513671875,
|
| 562 |
+
"learning_rate": 4.6710430799648143e-07,
|
| 563 |
+
"loss": 0.0,
|
| 564 |
+
"reward": 0.9955357313156128,
|
| 565 |
+
"reward_std": 0.012626906856894493,
|
| 566 |
+
"rewards/equation_reward_func": 0.0,
|
| 567 |
+
"rewards/format_reward_func": 0.9955357313156128,
|
| 568 |
+
"step": 86
|
| 569 |
+
},
|
| 570 |
+
{
|
| 571 |
+
"completion_length": 250.82032203674316,
|
| 572 |
+
"epoch": 0.10950381085705398,
|
| 573 |
+
"grad_norm": 0.01157958760624767,
|
| 574 |
+
"kl": 0.036041259765625,
|
| 575 |
+
"learning_rate": 4.652954630476127e-07,
|
| 576 |
+
"loss": 0.0,
|
| 577 |
+
"reward": 0.9944196566939354,
|
| 578 |
+
"reward_std": 0.015783633571118116,
|
| 579 |
+
"rewards/equation_reward_func": 0.0,
|
| 580 |
+
"rewards/format_reward_func": 0.9944196566939354,
|
| 581 |
+
"step": 88
|
| 582 |
+
},
|
| 583 |
+
{
|
| 584 |
+
"completion_length": 255.1462163925171,
|
| 585 |
+
"epoch": 0.11199253383107793,
|
| 586 |
+
"grad_norm": 0.0014016046187182924,
|
| 587 |
+
"kl": 0.0375823974609375,
|
| 588 |
+
"learning_rate": 4.6344190712584713e-07,
|
| 589 |
+
"loss": 0.0,
|
| 590 |
+
"reward": 1.0,
|
| 591 |
+
"reward_std": 0.0,
|
| 592 |
+
"rewards/equation_reward_func": 0.0,
|
| 593 |
+
"rewards/format_reward_func": 1.0,
|
| 594 |
+
"step": 90
|
| 595 |
+
},
|
| 596 |
+
{
|
| 597 |
+
"completion_length": 249.1640748977661,
|
| 598 |
+
"epoch": 0.11448125680510188,
|
| 599 |
+
"grad_norm": 0.0014472142931128767,
|
| 600 |
+
"kl": 0.0380706787109375,
|
| 601 |
+
"learning_rate": 4.615440251639995e-07,
|
| 602 |
+
"loss": 0.0,
|
| 603 |
+
"reward": 0.9988839328289032,
|
| 604 |
+
"reward_std": 0.0031567267142236233,
|
| 605 |
+
"rewards/equation_reward_func": 0.0,
|
| 606 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 607 |
+
"step": 92
|
| 608 |
+
},
|
| 609 |
+
{
|
| 610 |
+
"completion_length": 256.1071557998657,
|
| 611 |
+
"epoch": 0.11696997977912583,
|
| 612 |
+
"grad_norm": 0.005854394163673583,
|
| 613 |
+
"kl": 0.0372772216796875,
|
| 614 |
+
"learning_rate": 4.596022113001894e-07,
|
| 615 |
+
"loss": 0.0,
|
| 616 |
+
"reward": 0.9977678656578064,
|
| 617 |
+
"reward_std": 0.0063134534284472466,
|
| 618 |
+
"rewards/equation_reward_func": 0.0,
|
| 619 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 620 |
+
"step": 94
|
| 621 |
+
},
|
| 622 |
+
{
|
| 623 |
+
"completion_length": 239.82813549041748,
|
| 624 |
+
"epoch": 0.11945870275314979,
|
| 625 |
+
"grad_norm": 0.0014814527335160008,
|
| 626 |
+
"kl": 0.03824615478515625,
|
| 627 |
+
"learning_rate": 4.576168687959895e-07,
|
| 628 |
+
"loss": 0.0,
|
| 629 |
+
"reward": 0.9977678656578064,
|
| 630 |
+
"reward_std": 0.0063134534284472466,
|
| 631 |
+
"rewards/equation_reward_func": 0.0,
|
| 632 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 633 |
+
"step": 96
|
| 634 |
+
},
|
| 635 |
+
{
|
| 636 |
+
"completion_length": 245.9185380935669,
|
| 637 |
+
"epoch": 0.12194742572717375,
|
| 638 |
+
"grad_norm": 0.009993182099046437,
|
| 639 |
+
"kl": 0.03774261474609375,
|
| 640 |
+
"learning_rate": 4.555884099526793e-07,
|
| 641 |
+
"loss": 0.0,
|
| 642 |
+
"reward": 0.994419664144516,
|
| 643 |
+
"reward_std": 0.015783633571118116,
|
| 644 |
+
"rewards/equation_reward_func": 0.0,
|
| 645 |
+
"rewards/format_reward_func": 0.994419664144516,
|
| 646 |
+
"step": 98
|
| 647 |
+
},
|
| 648 |
+
{
|
| 649 |
+
"completion_length": 243.829252243042,
|
| 650 |
+
"epoch": 0.1244361487011977,
|
| 651 |
+
"grad_norm": 0.0012629113621896403,
|
| 652 |
+
"kl": 0.037689208984375,
|
| 653 |
+
"learning_rate": 4.5351725602562174e-07,
|
| 654 |
+
"loss": 0.0,
|
| 655 |
+
"reward": 0.9988839328289032,
|
| 656 |
+
"reward_std": 0.0031567267142236233,
|
| 657 |
+
"rewards/equation_reward_func": 0.0,
|
| 658 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 659 |
+
"step": 100
|
| 660 |
+
},
|
| 661 |
+
{
|
| 662 |
+
"completion_length": 240.93304634094238,
|
| 663 |
+
"epoch": 0.12692487167522165,
|
| 664 |
+
"grad_norm": 0.008888933685989486,
|
| 665 |
+
"kl": 0.04000091552734375,
|
| 666 |
+
"learning_rate": 4.514038371367791e-07,
|
| 667 |
+
"loss": 0.0,
|
| 668 |
+
"reward": 0.9977678656578064,
|
| 669 |
+
"reward_std": 0.0063134534284472466,
|
| 670 |
+
"rewards/equation_reward_func": 0.0,
|
| 671 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 672 |
+
"step": 102
|
| 673 |
+
},
|
| 674 |
+
{
|
| 675 |
+
"completion_length": 250.9430913925171,
|
| 676 |
+
"epoch": 0.1294135946492456,
|
| 677 |
+
"grad_norm": 0.007935174731567693,
|
| 678 |
+
"kl": 0.03668212890625,
|
| 679 |
+
"learning_rate": 4.4924859218538936e-07,
|
| 680 |
+
"loss": 0.0,
|
| 681 |
+
"reward": 0.996651791036129,
|
| 682 |
+
"reward_std": 0.0072898524813354015,
|
| 683 |
+
"rewards/equation_reward_func": 0.0,
|
| 684 |
+
"rewards/format_reward_func": 0.996651791036129,
|
| 685 |
+
"step": 104
|
| 686 |
+
},
|
| 687 |
+
{
|
| 688 |
+
"completion_length": 253.6439847946167,
|
| 689 |
+
"epoch": 0.13190231762326957,
|
| 690 |
+
"grad_norm": 0.007207650723410012,
|
| 691 |
+
"kl": 0.059051513671875,
|
| 692 |
+
"learning_rate": 4.470519687568185e-07,
|
| 693 |
+
"loss": 0.0001,
|
| 694 |
+
"reward": 0.993303582072258,
|
| 695 |
+
"reward_std": 0.01894036028534174,
|
| 696 |
+
"rewards/equation_reward_func": 0.0,
|
| 697 |
+
"rewards/format_reward_func": 0.993303582072258,
|
| 698 |
+
"step": 106
|
| 699 |
+
},
|
| 700 |
+
{
|
| 701 |
+
"completion_length": 246.56920719146729,
|
| 702 |
+
"epoch": 0.1343910405972935,
|
| 703 |
+
"grad_norm": 0.006148168480495231,
|
| 704 |
+
"kl": 0.036407470703125,
|
| 705 |
+
"learning_rate": 4.4481442302960923e-07,
|
| 706 |
+
"loss": 0.0,
|
| 707 |
+
"reward": 0.9966517984867096,
|
| 708 |
+
"reward_std": 0.00947018014267087,
|
| 709 |
+
"rewards/equation_reward_func": 0.0,
|
| 710 |
+
"rewards/format_reward_func": 0.9966517984867096,
|
| 711 |
+
"step": 108
|
| 712 |
+
},
|
| 713 |
+
{
|
| 714 |
+
"completion_length": 238.95313549041748,
|
| 715 |
+
"epoch": 0.13687976357131748,
|
| 716 |
+
"grad_norm": 0.009431097188708973,
|
| 717 |
+
"kl": 0.035797119140625,
|
| 718 |
+
"learning_rate": 4.4253641968074505e-07,
|
| 719 |
+
"loss": 0.0,
|
| 720 |
+
"reward": 0.9977678656578064,
|
| 721 |
+
"reward_std": 0.0063134534284472466,
|
| 722 |
+
"rewards/equation_reward_func": 0.0,
|
| 723 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 724 |
+
"step": 110
|
| 725 |
+
},
|
| 726 |
+
{
|
| 727 |
+
"completion_length": 246.18081378936768,
|
| 728 |
+
"epoch": 0.13936848654534142,
|
| 729 |
+
"grad_norm": 0.005876057759970698,
|
| 730 |
+
"kl": 0.037750244140625,
|
| 731 |
+
"learning_rate": 4.402184317891501e-07,
|
| 732 |
+
"loss": 0.0,
|
| 733 |
+
"reward": 0.9977678656578064,
|
| 734 |
+
"reward_std": 0.0063134534284472466,
|
| 735 |
+
"rewards/equation_reward_func": 0.0,
|
| 736 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 737 |
+
"step": 112
|
| 738 |
+
},
|
| 739 |
+
{
|
| 740 |
+
"completion_length": 248.73215198516846,
|
| 741 |
+
"epoch": 0.14185720951936537,
|
| 742 |
+
"grad_norm": 0.0077199360865173455,
|
| 743 |
+
"kl": 0.0368499755859375,
|
| 744 |
+
"learning_rate": 4.37860940737443e-07,
|
| 745 |
+
"loss": 0.0,
|
| 746 |
+
"reward": 0.9955357313156128,
|
| 747 |
+
"reward_std": 0.012626906856894493,
|
| 748 |
+
"rewards/equation_reward_func": 0.0,
|
| 749 |
+
"rewards/format_reward_func": 0.9955357313156128,
|
| 750 |
+
"step": 114
|
| 751 |
+
},
|
| 752 |
+
{
|
| 753 |
+
"completion_length": 243.86273765563965,
|
| 754 |
+
"epoch": 0.14434593249338934,
|
| 755 |
+
"grad_norm": 0.0054189642531993635,
|
| 756 |
+
"kl": 0.03558349609375,
|
| 757 |
+
"learning_rate": 4.354644361119671e-07,
|
| 758 |
+
"loss": 0.0,
|
| 759 |
+
"reward": 0.9977678656578064,
|
| 760 |
+
"reward_std": 0.0063134534284472466,
|
| 761 |
+
"rewards/equation_reward_func": 0.0,
|
| 762 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 763 |
+
"step": 116
|
| 764 |
+
},
|
| 765 |
+
{
|
| 766 |
+
"completion_length": 251.38617134094238,
|
| 767 |
+
"epoch": 0.14683465546741328,
|
| 768 |
+
"grad_norm": 0.006117742951139463,
|
| 769 |
+
"kl": 0.0378570556640625,
|
| 770 |
+
"learning_rate": 4.3302941560111716e-07,
|
| 771 |
+
"loss": 0.0,
|
| 772 |
+
"reward": 0.9966517984867096,
|
| 773 |
+
"reward_std": 0.00947018014267087,
|
| 774 |
+
"rewards/equation_reward_func": 0.0,
|
| 775 |
+
"rewards/format_reward_func": 0.9966517984867096,
|
| 776 |
+
"step": 118
|
| 777 |
+
},
|
| 778 |
+
{
|
| 779 |
+
"completion_length": 250.118314743042,
|
| 780 |
+
"epoch": 0.14932337844143723,
|
| 781 |
+
"grad_norm": 0.008141566580477025,
|
| 782 |
+
"kl": 0.0386962890625,
|
| 783 |
+
"learning_rate": 4.3055638489198236e-07,
|
| 784 |
+
"loss": 0.0,
|
| 785 |
+
"reward": 0.9977678582072258,
|
| 786 |
+
"reward_std": 0.0063134534284472466,
|
| 787 |
+
"rewards/equation_reward_func": 0.0,
|
| 788 |
+
"rewards/format_reward_func": 0.9977678582072258,
|
| 789 |
+
"step": 120
|
| 790 |
+
},
|
| 791 |
+
{
|
| 792 |
+
"completion_length": 248.11273288726807,
|
| 793 |
+
"epoch": 0.1518121014154612,
|
| 794 |
+
"grad_norm": 0.0036890211568659974,
|
| 795 |
+
"kl": 0.04131317138671875,
|
| 796 |
+
"learning_rate": 4.280458575653296e-07,
|
| 797 |
+
"loss": 0.0,
|
| 798 |
+
"reward": 0.9977678656578064,
|
| 799 |
+
"reward_std": 0.0063134534284472466,
|
| 800 |
+
"rewards/equation_reward_func": 0.0,
|
| 801 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 802 |
+
"step": 122
|
| 803 |
+
},
|
| 804 |
+
{
|
| 805 |
+
"completion_length": 248.6015739440918,
|
| 806 |
+
"epoch": 0.15430082438948514,
|
| 807 |
+
"grad_norm": 0.009172809538520958,
|
| 808 |
+
"kl": 0.0360565185546875,
|
| 809 |
+
"learning_rate": 4.2549835498894665e-07,
|
| 810 |
+
"loss": 0.0,
|
| 811 |
+
"reward": 0.994419664144516,
|
| 812 |
+
"reward_std": 0.015783633571118116,
|
| 813 |
+
"rewards/equation_reward_func": 0.0,
|
| 814 |
+
"rewards/format_reward_func": 0.994419664144516,
|
| 815 |
+
"step": 124
|
| 816 |
+
},
|
| 817 |
+
{
|
| 818 |
+
"completion_length": 242.13393878936768,
|
| 819 |
+
"epoch": 0.1567895473635091,
|
| 820 |
+
"grad_norm": 0.021937162250364135,
|
| 821 |
+
"kl": 0.03790283203125,
|
| 822 |
+
"learning_rate": 4.229144062093679e-07,
|
| 823 |
+
"loss": 0.0,
|
| 824 |
+
"reward": 0.9977678656578064,
|
| 825 |
+
"reward_std": 0.0063134534284472466,
|
| 826 |
+
"rewards/equation_reward_func": 0.0,
|
| 827 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 828 |
+
"step": 126
|
| 829 |
+
},
|
| 830 |
+
{
|
| 831 |
+
"completion_length": 244.00000953674316,
|
| 832 |
+
"epoch": 0.15927827033753306,
|
| 833 |
+
"grad_norm": 0.013265008355772755,
|
| 834 |
+
"kl": 0.037567138671875,
|
| 835 |
+
"learning_rate": 4.2029454784200675e-07,
|
| 836 |
+
"loss": 0.0,
|
| 837 |
+
"reward": 0.9966517984867096,
|
| 838 |
+
"reward_std": 0.00947018014267087,
|
| 839 |
+
"rewards/equation_reward_func": 0.0,
|
| 840 |
+
"rewards/format_reward_func": 0.9966517984867096,
|
| 841 |
+
"step": 128
|
| 842 |
+
},
|
| 843 |
+
{
|
| 844 |
+
"completion_length": 241.54353713989258,
|
| 845 |
+
"epoch": 0.161766993311557,
|
| 846 |
+
"grad_norm": 0.013941850356785124,
|
| 847 |
+
"kl": 0.038330078125,
|
| 848 |
+
"learning_rate": 4.1763932395971433e-07,
|
| 849 |
+
"loss": 0.0,
|
| 850 |
+
"reward": 0.9966517984867096,
|
| 851 |
+
"reward_std": 0.00947018014267087,
|
| 852 |
+
"rewards/equation_reward_func": 0.0,
|
| 853 |
+
"rewards/format_reward_func": 0.9966517984867096,
|
| 854 |
+
"step": 130
|
| 855 |
+
},
|
| 856 |
+
{
|
| 857 |
+
"completion_length": 237.6663064956665,
|
| 858 |
+
"epoch": 0.16425571628558097,
|
| 859 |
+
"grad_norm": 0.009267818117271523,
|
| 860 |
+
"kl": 0.0396575927734375,
|
| 861 |
+
"learning_rate": 4.1494928597979117e-07,
|
| 862 |
+
"loss": 0.0,
|
| 863 |
+
"reward": 0.994419664144516,
|
| 864 |
+
"reward_std": 0.015783633571118116,
|
| 865 |
+
"rewards/equation_reward_func": 0.0,
|
| 866 |
+
"rewards/format_reward_func": 0.994419664144516,
|
| 867 |
+
"step": 132
|
| 868 |
+
},
|
| 869 |
+
{
|
| 870 |
+
"completion_length": 240.70090103149414,
|
| 871 |
+
"epoch": 0.16674443925960492,
|
| 872 |
+
"grad_norm": 0.012119765907156485,
|
| 873 |
+
"kl": 0.0418548583984375,
|
| 874 |
+
"learning_rate": 4.122249925494726e-07,
|
| 875 |
+
"loss": 0.0,
|
| 876 |
+
"reward": 0.9966517984867096,
|
| 877 |
+
"reward_std": 0.00947018014267087,
|
| 878 |
+
"rewards/equation_reward_func": 0.0,
|
| 879 |
+
"rewards/format_reward_func": 0.9966517984867096,
|
| 880 |
+
"step": 134
|
| 881 |
+
},
|
| 882 |
+
{
|
| 883 |
+
"completion_length": 235.82367038726807,
|
| 884 |
+
"epoch": 0.16923316223362886,
|
| 885 |
+
"grad_norm": 0.0013331935396181439,
|
| 886 |
+
"kl": 0.0379486083984375,
|
| 887 |
+
"learning_rate": 4.094670094299131e-07,
|
| 888 |
+
"loss": 0.0,
|
| 889 |
+
"reward": 1.0,
|
| 890 |
+
"reward_std": 0.0,
|
| 891 |
+
"rewards/equation_reward_func": 0.0,
|
| 892 |
+
"rewards/format_reward_func": 1.0,
|
| 893 |
+
"step": 136
|
| 894 |
+
},
|
| 895 |
+
{
|
| 896 |
+
"completion_length": 233.64621543884277,
|
| 897 |
+
"epoch": 0.17172188520765283,
|
| 898 |
+
"grad_norm": 0.005329870298582281,
|
| 899 |
+
"kl": 0.0389251708984375,
|
| 900 |
+
"learning_rate": 4.066759093786931e-07,
|
| 901 |
+
"loss": 0.0,
|
| 902 |
+
"reward": 0.9977678656578064,
|
| 903 |
+
"reward_std": 0.0063134534284472466,
|
| 904 |
+
"rewards/equation_reward_func": 0.0,
|
| 905 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 906 |
+
"step": 138
|
| 907 |
+
},
|
| 908 |
+
{
|
| 909 |
+
"completion_length": 244.22546005249023,
|
| 910 |
+
"epoch": 0.17421060818167677,
|
| 911 |
+
"grad_norm": 0.005619989780456964,
|
| 912 |
+
"kl": 0.0381927490234375,
|
| 913 |
+
"learning_rate": 4.038522720308732e-07,
|
| 914 |
+
"loss": 0.0,
|
| 915 |
+
"reward": 0.9988839328289032,
|
| 916 |
+
"reward_std": 0.0031567267142236233,
|
| 917 |
+
"rewards/equation_reward_func": 0.0,
|
| 918 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 919 |
+
"step": 140
|
| 920 |
+
},
|
| 921 |
+
{
|
| 922 |
+
"completion_length": 242.11719799041748,
|
| 923 |
+
"epoch": 0.17669933115570072,
|
| 924 |
+
"grad_norm": 0.006376477165794642,
|
| 925 |
+
"kl": 0.0377197265625,
|
| 926 |
+
"learning_rate": 4.009966837786194e-07,
|
| 927 |
+
"loss": 0.0,
|
| 928 |
+
"reward": 0.9988839328289032,
|
| 929 |
+
"reward_std": 0.0031567267142236233,
|
| 930 |
+
"rewards/equation_reward_func": 0.0,
|
| 931 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 932 |
+
"step": 142
|
| 933 |
+
},
|
| 934 |
+
{
|
| 935 |
+
"completion_length": 238.0703248977661,
|
| 936 |
+
"epoch": 0.1791880541297247,
|
| 937 |
+
"grad_norm": 0.028370327898306225,
|
| 938 |
+
"kl": 0.03790283203125,
|
| 939 |
+
"learning_rate": 3.981097376494259e-07,
|
| 940 |
+
"loss": 0.0,
|
| 941 |
+
"reward": 0.9988839328289032,
|
| 942 |
+
"reward_std": 0.0031567267142236233,
|
| 943 |
+
"rewards/equation_reward_func": 0.0,
|
| 944 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 945 |
+
"step": 144
|
| 946 |
+
},
|
| 947 |
+
{
|
| 948 |
+
"completion_length": 244.7712163925171,
|
| 949 |
+
"epoch": 0.18167677710374863,
|
| 950 |
+
"grad_norm": 0.0015509057376277774,
|
| 951 |
+
"kl": 0.0378265380859375,
|
| 952 |
+
"learning_rate": 3.951920331829592e-07,
|
| 953 |
+
"loss": 0.0,
|
| 954 |
+
"reward": 0.9988839328289032,
|
| 955 |
+
"reward_std": 0.0031567267142236233,
|
| 956 |
+
"rewards/equation_reward_func": 0.0,
|
| 957 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 958 |
+
"step": 146
|
| 959 |
+
},
|
| 960 |
+
{
|
| 961 |
+
"completion_length": 247.24666213989258,
|
| 962 |
+
"epoch": 0.1841655000777726,
|
| 963 |
+
"grad_norm": 0.0012690449195768786,
|
| 964 |
+
"kl": 0.0364837646484375,
|
| 965 |
+
"learning_rate": 3.922441763065506e-07,
|
| 966 |
+
"loss": 0.0,
|
| 967 |
+
"reward": 1.0,
|
| 968 |
+
"reward_std": 0.0,
|
| 969 |
+
"rewards/equation_reward_func": 0.0,
|
| 970 |
+
"rewards/format_reward_func": 1.0,
|
| 971 |
+
"step": 148
|
| 972 |
+
},
|
| 973 |
+
{
|
| 974 |
+
"completion_length": 240.0212163925171,
|
| 975 |
+
"epoch": 0.18665422305179655,
|
| 976 |
+
"grad_norm": 0.013797136486153857,
|
| 977 |
+
"kl": 0.03704833984375,
|
| 978 |
+
"learning_rate": 3.8926677920936093e-07,
|
| 979 |
+
"loss": 0.0,
|
| 980 |
+
"reward": 0.9977678656578064,
|
| 981 |
+
"reward_std": 0.0063134534284472466,
|
| 982 |
+
"rewards/equation_reward_func": 0.0,
|
| 983 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 984 |
+
"step": 150
|
| 985 |
+
},
|
| 986 |
+
{
|
| 987 |
+
"completion_length": 246.80692958831787,
|
| 988 |
+
"epoch": 0.1891429460258205,
|
| 989 |
+
"grad_norm": 0.008542886544667705,
|
| 990 |
+
"kl": 0.0371246337890625,
|
| 991 |
+
"learning_rate": 3.862604602152464e-07,
|
| 992 |
+
"loss": 0.0,
|
| 993 |
+
"reward": 0.9944196566939354,
|
| 994 |
+
"reward_std": 0.015783633571118116,
|
| 995 |
+
"rewards/equation_reward_func": 0.0,
|
| 996 |
+
"rewards/format_reward_func": 0.9944196566939354,
|
| 997 |
+
"step": 152
|
| 998 |
+
},
|
| 999 |
+
{
|
| 1000 |
+
"completion_length": 237.7500123977661,
|
| 1001 |
+
"epoch": 0.19163166899984446,
|
| 1002 |
+
"grad_norm": 0.00848780488153459,
|
| 1003 |
+
"kl": 0.0372467041015625,
|
| 1004 |
+
"learning_rate": 3.8322584365434934e-07,
|
| 1005 |
+
"loss": 0.0,
|
| 1006 |
+
"reward": 0.9977678656578064,
|
| 1007 |
+
"reward_std": 0.0063134534284472466,
|
| 1008 |
+
"rewards/equation_reward_func": 0.0,
|
| 1009 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 1010 |
+
"step": 154
|
| 1011 |
+
},
|
| 1012 |
+
{
|
| 1013 |
+
"completion_length": 235.2388505935669,
|
| 1014 |
+
"epoch": 0.1941203919738684,
|
| 1015 |
+
"grad_norm": 0.0014224607179590997,
|
| 1016 |
+
"kl": 0.0376434326171875,
|
| 1017 |
+
"learning_rate": 3.8016355973344173e-07,
|
| 1018 |
+
"loss": 0.0,
|
| 1019 |
+
"reward": 0.9977678656578064,
|
| 1020 |
+
"reward_std": 0.0063134534284472466,
|
| 1021 |
+
"rewards/equation_reward_func": 0.0,
|
| 1022 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 1023 |
+
"step": 156
|
| 1024 |
+
},
|
| 1025 |
+
{
|
| 1026 |
+
"completion_length": 242.0346097946167,
|
| 1027 |
+
"epoch": 0.19660911494789235,
|
| 1028 |
+
"grad_norm": 0.007810868977196938,
|
| 1029 |
+
"kl": 0.0355224609375,
|
| 1030 |
+
"learning_rate": 3.7707424440504863e-07,
|
| 1031 |
+
"loss": 0.0,
|
| 1032 |
+
"reward": 0.9977678656578064,
|
| 1033 |
+
"reward_std": 0.0063134534284472466,
|
| 1034 |
+
"rewards/equation_reward_func": 0.0,
|
| 1035 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 1036 |
+
"step": 158
|
| 1037 |
+
},
|
| 1038 |
+
{
|
| 1039 |
+
"completion_length": 244.3404130935669,
|
| 1040 |
+
"epoch": 0.19909783792191632,
|
| 1041 |
+
"grad_norm": 0.006684946104716033,
|
| 1042 |
+
"kl": 0.035614013671875,
|
| 1043 |
+
"learning_rate": 3.739585392353787e-07,
|
| 1044 |
+
"loss": 0.0,
|
| 1045 |
+
"reward": 0.9955357313156128,
|
| 1046 |
+
"reward_std": 0.012626906856894493,
|
| 1047 |
+
"rewards/equation_reward_func": 0.0,
|
| 1048 |
+
"rewards/format_reward_func": 0.9955357313156128,
|
| 1049 |
+
"step": 160
|
| 1050 |
+
},
|
| 1051 |
+
{
|
| 1052 |
+
"completion_length": 246.79577159881592,
|
| 1053 |
+
"epoch": 0.20158656089594026,
|
| 1054 |
+
"grad_norm": 0.005256302746271805,
|
| 1055 |
+
"kl": 0.035919189453125,
|
| 1056 |
+
"learning_rate": 3.7081709127108767e-07,
|
| 1057 |
+
"loss": 0.0,
|
| 1058 |
+
"reward": 0.9988839328289032,
|
| 1059 |
+
"reward_std": 0.0031567267142236233,
|
| 1060 |
+
"rewards/equation_reward_func": 0.0,
|
| 1061 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 1062 |
+
"step": 162
|
| 1063 |
+
},
|
| 1064 |
+
{
|
| 1065 |
+
"completion_length": 240.33483219146729,
|
| 1066 |
+
"epoch": 0.20407528386996424,
|
| 1067 |
+
"grad_norm": 0.005275884352682047,
|
| 1068 |
+
"kl": 0.0363922119140625,
|
| 1069 |
+
"learning_rate": 3.6765055290490513e-07,
|
| 1070 |
+
"loss": 0.0,
|
| 1071 |
+
"reward": 0.994419664144516,
|
| 1072 |
+
"reward_std": 0.015783633571118116,
|
| 1073 |
+
"rewards/equation_reward_func": 0.0,
|
| 1074 |
+
"rewards/format_reward_func": 0.994419664144516,
|
| 1075 |
+
"step": 164
|
| 1076 |
+
},
|
| 1077 |
+
{
|
| 1078 |
+
"completion_length": 244.2767972946167,
|
| 1079 |
+
"epoch": 0.20656400684398818,
|
| 1080 |
+
"grad_norm": 0.00905429835962999,
|
| 1081 |
+
"kl": 0.0367431640625,
|
| 1082 |
+
"learning_rate": 3.644595817401501e-07,
|
| 1083 |
+
"loss": 0.0,
|
| 1084 |
+
"reward": 0.9955357313156128,
|
| 1085 |
+
"reward_std": 0.012626906856894493,
|
| 1086 |
+
"rewards/equation_reward_func": 0.0,
|
| 1087 |
+
"rewards/format_reward_func": 0.9955357313156128,
|
| 1088 |
+
"step": 166
|
| 1089 |
+
},
|
| 1090 |
+
{
|
| 1091 |
+
"completion_length": 239.8694314956665,
|
| 1092 |
+
"epoch": 0.20905272981801212,
|
| 1093 |
+
"grad_norm": 0.001515588287808503,
|
| 1094 |
+
"kl": 0.0360565185546875,
|
| 1095 |
+
"learning_rate": 3.6124484045416483e-07,
|
| 1096 |
+
"loss": 0.0,
|
| 1097 |
+
"reward": 0.9977678656578064,
|
| 1098 |
+
"reward_std": 0.0063134534284472466,
|
| 1099 |
+
"rewards/equation_reward_func": 0.0,
|
| 1100 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 1101 |
+
"step": 168
|
| 1102 |
+
},
|
| 1103 |
+
{
|
| 1104 |
+
"completion_length": 245.84041213989258,
|
| 1105 |
+
"epoch": 0.2115414527920361,
|
| 1106 |
+
"grad_norm": 0.01143986411870428,
|
| 1107 |
+
"kl": 0.035736083984375,
|
| 1108 |
+
"learning_rate": 3.580069966606949e-07,
|
| 1109 |
+
"loss": 0.0,
|
| 1110 |
+
"reward": 0.9966517984867096,
|
| 1111 |
+
"reward_std": 0.00947018014267087,
|
| 1112 |
+
"rewards/equation_reward_func": 0.0,
|
| 1113 |
+
"rewards/format_reward_func": 0.9966517984867096,
|
| 1114 |
+
"step": 170
|
| 1115 |
+
},
|
| 1116 |
+
{
|
| 1117 |
+
"completion_length": 234.6406373977661,
|
| 1118 |
+
"epoch": 0.21403017576606004,
|
| 1119 |
+
"grad_norm": 0.012227399752466542,
|
| 1120 |
+
"kl": 0.0365753173828125,
|
| 1121 |
+
"learning_rate": 3.547467227712444e-07,
|
| 1122 |
+
"loss": 0.0,
|
| 1123 |
+
"reward": 0.9966517984867096,
|
| 1124 |
+
"reward_std": 0.00947018014267087,
|
| 1125 |
+
"rewards/equation_reward_func": 0.0,
|
| 1126 |
+
"rewards/format_reward_func": 0.9966517984867096,
|
| 1127 |
+
"step": 172
|
| 1128 |
+
},
|
| 1129 |
+
{
|
| 1130 |
+
"completion_length": 239.46206378936768,
|
| 1131 |
+
"epoch": 0.21651889874008398,
|
| 1132 |
+
"grad_norm": 0.0013708158434779418,
|
| 1133 |
+
"kl": 0.0361328125,
|
| 1134 |
+
"learning_rate": 3.5146469585543386e-07,
|
| 1135 |
+
"loss": 0.0,
|
| 1136 |
+
"reward": 1.0,
|
| 1137 |
+
"reward_std": 0.0,
|
| 1138 |
+
"rewards/equation_reward_func": 0.0,
|
| 1139 |
+
"rewards/format_reward_func": 1.0,
|
| 1140 |
+
"step": 174
|
| 1141 |
+
},
|
| 1142 |
+
{
|
| 1143 |
+
"completion_length": 242.0636281967163,
|
| 1144 |
+
"epoch": 0.21900762171410795,
|
| 1145 |
+
"grad_norm": 0.005985792387537067,
|
| 1146 |
+
"kl": 0.03522491455078125,
|
| 1147 |
+
"learning_rate": 3.481615975003922e-07,
|
| 1148 |
+
"loss": 0.0,
|
| 1149 |
+
"reward": 0.9955357238650322,
|
| 1150 |
+
"reward_std": 0.012626906856894493,
|
| 1151 |
+
"rewards/equation_reward_func": 0.0,
|
| 1152 |
+
"rewards/format_reward_func": 0.9955357238650322,
|
| 1153 |
+
"step": 176
|
| 1154 |
+
},
|
| 1155 |
+
{
|
| 1156 |
+
"completion_length": 236.69867134094238,
|
| 1157 |
+
"epoch": 0.2214963446881319,
|
| 1158 |
+
"grad_norm": 0.005543063900031294,
|
| 1159 |
+
"kl": 0.0347747802734375,
|
| 1160 |
+
"learning_rate": 3.448381136692089e-07,
|
| 1161 |
+
"loss": 0.0,
|
| 1162 |
+
"reward": 0.9988839328289032,
|
| 1163 |
+
"reward_std": 0.0031567267142236233,
|
| 1164 |
+
"rewards/equation_reward_func": 0.0,
|
| 1165 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 1166 |
+
"step": 178
|
| 1167 |
+
},
|
| 1168 |
+
{
|
| 1169 |
+
"completion_length": 232.74442958831787,
|
| 1170 |
+
"epoch": 0.22398506766215587,
|
| 1171 |
+
"grad_norm": 0.016458488651150318,
|
| 1172 |
+
"kl": 0.0449371337890625,
|
| 1173 |
+
"learning_rate": 3.4149493455847897e-07,
|
| 1174 |
+
"loss": 0.0,
|
| 1175 |
+
"reward": 0.9977678656578064,
|
| 1176 |
+
"reward_std": 0.0063134534284472466,
|
| 1177 |
+
"rewards/equation_reward_func": 0.0,
|
| 1178 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 1179 |
+
"step": 180
|
| 1180 |
+
},
|
| 1181 |
+
{
|
| 1182 |
+
"completion_length": 239.68750953674316,
|
| 1183 |
+
"epoch": 0.2264737906361798,
|
| 1184 |
+
"grad_norm": 0.008616008287002855,
|
| 1185 |
+
"kl": 0.035369873046875,
|
| 1186 |
+
"learning_rate": 3.3813275445496766e-07,
|
| 1187 |
+
"loss": 0.0,
|
| 1188 |
+
"reward": 0.9977678656578064,
|
| 1189 |
+
"reward_std": 0.0063134534284472466,
|
| 1190 |
+
"rewards/equation_reward_func": 0.0,
|
| 1191 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 1192 |
+
"step": 182
|
| 1193 |
+
},
|
| 1194 |
+
{
|
| 1195 |
+
"completion_length": 237.28572463989258,
|
| 1196 |
+
"epoch": 0.22896251361020376,
|
| 1197 |
+
"grad_norm": 0.0012520075655868543,
|
| 1198 |
+
"kl": 0.0345611572265625,
|
| 1199 |
+
"learning_rate": 3.347522715914262e-07,
|
| 1200 |
+
"loss": 0.0,
|
| 1201 |
+
"reward": 1.0,
|
| 1202 |
+
"reward_std": 0.0,
|
| 1203 |
+
"rewards/equation_reward_func": 0.0,
|
| 1204 |
+
"rewards/format_reward_func": 1.0,
|
| 1205 |
+
"step": 184
|
| 1206 |
+
},
|
| 1207 |
+
{
|
| 1208 |
+
"completion_length": 239.86608219146729,
|
| 1209 |
+
"epoch": 0.23145123658422773,
|
| 1210 |
+
"grad_norm": 0.006149863991426027,
|
| 1211 |
+
"kl": 0.035400390625,
|
| 1212 |
+
"learning_rate": 3.313541880015877e-07,
|
| 1213 |
+
"loss": 0.0,
|
| 1214 |
+
"reward": 0.9966517984867096,
|
| 1215 |
+
"reward_std": 0.00947018014267087,
|
| 1216 |
+
"rewards/equation_reward_func": 0.0,
|
| 1217 |
+
"rewards/format_reward_func": 0.9966517984867096,
|
| 1218 |
+
"step": 186
|
| 1219 |
+
},
|
| 1220 |
+
{
|
| 1221 |
+
"completion_length": 235.6138515472412,
|
| 1222 |
+
"epoch": 0.23393995955825167,
|
| 1223 |
+
"grad_norm": 0.006501865958706451,
|
| 1224 |
+
"kl": 0.0355377197265625,
|
| 1225 |
+
"learning_rate": 3.279392093743747e-07,
|
| 1226 |
+
"loss": 0.0,
|
| 1227 |
+
"reward": 0.9977678656578064,
|
| 1228 |
+
"reward_std": 0.0063134534284472466,
|
| 1229 |
+
"rewards/equation_reward_func": 0.0,
|
| 1230 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 1231 |
+
"step": 188
|
| 1232 |
+
},
|
| 1233 |
+
{
|
| 1234 |
+
"completion_length": 244.5089406967163,
|
| 1235 |
+
"epoch": 0.2364286825322756,
|
| 1236 |
+
"grad_norm": 0.008690843209276718,
|
| 1237 |
+
"kl": 0.03710174560546875,
|
| 1238 |
+
"learning_rate": 3.245080449073459e-07,
|
| 1239 |
+
"loss": 0.0,
|
| 1240 |
+
"reward": 0.9966517984867096,
|
| 1241 |
+
"reward_std": 0.00947018014267087,
|
| 1242 |
+
"rewards/equation_reward_func": 0.0,
|
| 1243 |
+
"rewards/format_reward_func": 0.9966517984867096,
|
| 1244 |
+
"step": 190
|
| 1245 |
+
},
|
| 1246 |
+
{
|
| 1247 |
+
"completion_length": 236.89286994934082,
|
| 1248 |
+
"epoch": 0.23891740550629958,
|
| 1249 |
+
"grad_norm": 0.010695035904088867,
|
| 1250 |
+
"kl": 0.0343170166015625,
|
| 1251 |
+
"learning_rate": 3.210614071594162e-07,
|
| 1252 |
+
"loss": 0.0,
|
| 1253 |
+
"reward": 0.9988839328289032,
|
| 1254 |
+
"reward_std": 0.0031567267142236233,
|
| 1255 |
+
"rewards/equation_reward_func": 0.0,
|
| 1256 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 1257 |
+
"step": 192
|
| 1258 |
+
},
|
| 1259 |
+
{
|
| 1260 |
+
"completion_length": 237.45425128936768,
|
| 1261 |
+
"epoch": 0.24140612848032353,
|
| 1262 |
+
"grad_norm": 0.008876854007516248,
|
| 1263 |
+
"kl": 0.0358428955078125,
|
| 1264 |
+
"learning_rate": 3.1760001190287695e-07,
|
| 1265 |
+
"loss": 0.0,
|
| 1266 |
+
"reward": 0.9977678656578064,
|
| 1267 |
+
"reward_std": 0.0063134534284472466,
|
| 1268 |
+
"rewards/equation_reward_func": 0.0,
|
| 1269 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 1270 |
+
"step": 194
|
| 1271 |
+
},
|
| 1272 |
+
{
|
| 1273 |
+
"completion_length": 232.3035831451416,
|
| 1274 |
+
"epoch": 0.2438948514543475,
|
| 1275 |
+
"grad_norm": 0.0012580700152787947,
|
| 1276 |
+
"kl": 0.0338134765625,
|
| 1277 |
+
"learning_rate": 3.141245779747502e-07,
|
| 1278 |
+
"loss": 0.0,
|
| 1279 |
+
"reward": 1.0,
|
| 1280 |
+
"reward_std": 0.0,
|
| 1281 |
+
"rewards/equation_reward_func": 0.0,
|
| 1282 |
+
"rewards/format_reward_func": 1.0,
|
| 1283 |
+
"step": 196
|
| 1284 |
+
},
|
| 1285 |
+
{
|
| 1286 |
+
"completion_length": 236.17523384094238,
|
| 1287 |
+
"epoch": 0.24638357442837144,
|
| 1288 |
+
"grad_norm": 0.0013888134732519966,
|
| 1289 |
+
"kl": 0.03568267822265625,
|
| 1290 |
+
"learning_rate": 3.106358271275056e-07,
|
| 1291 |
+
"loss": 0.0,
|
| 1292 |
+
"reward": 0.9988839328289032,
|
| 1293 |
+
"reward_std": 0.0031567267142236233,
|
| 1294 |
+
"rewards/equation_reward_func": 0.0,
|
| 1295 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 1296 |
+
"step": 198
|
| 1297 |
+
},
|
| 1298 |
+
{
|
| 1299 |
+
"completion_length": 232.61384868621826,
|
| 1300 |
+
"epoch": 0.2488722974023954,
|
| 1301 |
+
"grad_norm": 0.0013082596723838322,
|
| 1302 |
+
"kl": 0.0361175537109375,
|
| 1303 |
+
"learning_rate": 3.0713448387917227e-07,
|
| 1304 |
+
"loss": 0.0,
|
| 1305 |
+
"reward": 1.0,
|
| 1306 |
+
"reward_std": 0.0,
|
| 1307 |
+
"rewards/equation_reward_func": 0.0,
|
| 1308 |
+
"rewards/format_reward_func": 1.0,
|
| 1309 |
+
"step": 200
|
| 1310 |
+
},
|
| 1311 |
+
{
|
| 1312 |
+
"completion_length": 234.1596097946167,
|
| 1313 |
+
"epoch": 0.25136102037641933,
|
| 1314 |
+
"grad_norm": 0.005619002043642409,
|
| 1315 |
+
"kl": 0.035064697265625,
|
| 1316 |
+
"learning_rate": 3.0362127536287636e-07,
|
| 1317 |
+
"loss": 0.0,
|
| 1318 |
+
"reward": 0.9988839328289032,
|
| 1319 |
+
"reward_std": 0.0031567267142236233,
|
| 1320 |
+
"rewards/equation_reward_func": 0.0,
|
| 1321 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 1322 |
+
"step": 202
|
| 1323 |
+
},
|
| 1324 |
+
{
|
| 1325 |
+
"completion_length": 240.12724590301514,
|
| 1326 |
+
"epoch": 0.2538497433504433,
|
| 1327 |
+
"grad_norm": 0.001215418276843526,
|
| 1328 |
+
"kl": 0.032501220703125,
|
| 1329 |
+
"learning_rate": 3.0009693117583523e-07,
|
| 1330 |
+
"loss": 0.0,
|
| 1331 |
+
"reward": 1.0,
|
| 1332 |
+
"reward_std": 0.0,
|
| 1333 |
+
"rewards/equation_reward_func": 0.0,
|
| 1334 |
+
"rewards/format_reward_func": 1.0,
|
| 1335 |
+
"step": 204
|
| 1336 |
+
},
|
| 1337 |
+
{
|
| 1338 |
+
"completion_length": 240.5993413925171,
|
| 1339 |
+
"epoch": 0.2563384663244673,
|
| 1340 |
+
"grad_norm": 0.0012980749880742157,
|
| 1341 |
+
"kl": 0.03325653076171875,
|
| 1342 |
+
"learning_rate": 2.965621832278401e-07,
|
| 1343 |
+
"loss": 0.0,
|
| 1344 |
+
"reward": 0.9977678656578064,
|
| 1345 |
+
"reward_std": 0.0063134534284472466,
|
| 1346 |
+
"rewards/equation_reward_func": 0.0,
|
| 1347 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 1348 |
+
"step": 206
|
| 1349 |
+
},
|
| 1350 |
+
{
|
| 1351 |
+
"completion_length": 238.35157299041748,
|
| 1352 |
+
"epoch": 0.2588271892984912,
|
| 1353 |
+
"grad_norm": 0.012326143240222623,
|
| 1354 |
+
"kl": 0.03427886962890625,
|
| 1355 |
+
"learning_rate": 2.9301776558925875e-07,
|
| 1356 |
+
"loss": 0.0,
|
| 1357 |
+
"reward": 0.9966517984867096,
|
| 1358 |
+
"reward_std": 0.00947018014267087,
|
| 1359 |
+
"rewards/equation_reward_func": 0.0,
|
| 1360 |
+
"rewards/format_reward_func": 0.9966517984867096,
|
| 1361 |
+
"step": 208
|
| 1362 |
+
},
|
| 1363 |
+
{
|
| 1364 |
+
"completion_length": 239.8359498977661,
|
| 1365 |
+
"epoch": 0.26131591227251516,
|
| 1366 |
+
"grad_norm": 0.00486814033751187,
|
| 1367 |
+
"kl": 0.03197479248046875,
|
| 1368 |
+
"learning_rate": 2.894644143385885e-07,
|
| 1369 |
+
"loss": 0.0,
|
| 1370 |
+
"reward": 0.9977678656578064,
|
| 1371 |
+
"reward_std": 0.0063134534284472466,
|
| 1372 |
+
"rewards/equation_reward_func": 0.0,
|
| 1373 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 1374 |
+
"step": 210
|
| 1375 |
+
},
|
| 1376 |
+
{
|
| 1377 |
+
"completion_length": 236.45759963989258,
|
| 1378 |
+
"epoch": 0.26380463524653913,
|
| 1379 |
+
"grad_norm": 0.00850905620683849,
|
| 1380 |
+
"kl": 0.0343170166015625,
|
| 1381 |
+
"learning_rate": 2.859028674095937e-07,
|
| 1382 |
+
"loss": 0.0,
|
| 1383 |
+
"reward": 0.996651791036129,
|
| 1384 |
+
"reward_std": 0.00947018014267087,
|
| 1385 |
+
"rewards/equation_reward_func": 0.0,
|
| 1386 |
+
"rewards/format_reward_func": 0.996651791036129,
|
| 1387 |
+
"step": 212
|
| 1388 |
+
},
|
| 1389 |
+
{
|
| 1390 |
+
"completion_length": 234.11273288726807,
|
| 1391 |
+
"epoch": 0.26629335822056305,
|
| 1392 |
+
"grad_norm": 0.0011776757140757428,
|
| 1393 |
+
"kl": 0.0316925048828125,
|
| 1394 |
+
"learning_rate": 2.823338644380566e-07,
|
| 1395 |
+
"loss": 0.0,
|
| 1396 |
+
"reward": 0.9977678656578064,
|
| 1397 |
+
"reward_std": 0.0063134534284472466,
|
| 1398 |
+
"rewards/equation_reward_func": 0.0,
|
| 1399 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 1400 |
+
"step": 214
|
| 1401 |
+
},
|
| 1402 |
+
{
|
| 1403 |
+
"completion_length": 241.9665298461914,
|
| 1404 |
+
"epoch": 0.268782081194587,
|
| 1405 |
+
"grad_norm": 0.005375269654956656,
|
| 1406 |
+
"kl": 0.03156280517578125,
|
| 1407 |
+
"learning_rate": 2.7875814660817504e-07,
|
| 1408 |
+
"loss": 0.0,
|
| 1409 |
+
"reward": 0.9955357313156128,
|
| 1410 |
+
"reward_std": 0.012626906856894493,
|
| 1411 |
+
"rewards/equation_reward_func": 0.0,
|
| 1412 |
+
"rewards/format_reward_func": 0.9955357313156128,
|
| 1413 |
+
"step": 216
|
| 1414 |
+
},
|
| 1415 |
+
{
|
| 1416 |
+
"completion_length": 237.3906364440918,
|
| 1417 |
+
"epoch": 0.271270804168611,
|
| 1418 |
+
"grad_norm": 0.0045854327854698935,
|
| 1419 |
+
"kl": 0.03202056884765625,
|
| 1420 |
+
"learning_rate": 2.751764564986396e-07,
|
| 1421 |
+
"loss": 0.0,
|
| 1422 |
+
"reward": 0.9977678656578064,
|
| 1423 |
+
"reward_std": 0.0063134534284472466,
|
| 1424 |
+
"rewards/equation_reward_func": 0.0,
|
| 1425 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 1426 |
+
"step": 218
|
| 1427 |
+
},
|
| 1428 |
+
{
|
| 1429 |
+
"completion_length": 230.79688358306885,
|
| 1430 |
+
"epoch": 0.27375952714263496,
|
| 1431 |
+
"grad_norm": 0.008443268042406577,
|
| 1432 |
+
"kl": 0.031982421875,
|
| 1433 |
+
"learning_rate": 2.715895379284194e-07,
|
| 1434 |
+
"loss": 0.0,
|
| 1435 |
+
"reward": 0.9977678656578064,
|
| 1436 |
+
"reward_std": 0.0063134534284472466,
|
| 1437 |
+
"rewards/equation_reward_func": 0.0,
|
| 1438 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 1439 |
+
"step": 220
|
| 1440 |
+
},
|
| 1441 |
+
{
|
| 1442 |
+
"completion_length": 237.45759963989258,
|
| 1443 |
+
"epoch": 0.2762482501166589,
|
| 1444 |
+
"grad_norm": 0.0013591790273522189,
|
| 1445 |
+
"kl": 0.03302001953125,
|
| 1446 |
+
"learning_rate": 2.6799813580229174e-07,
|
| 1447 |
+
"loss": 0.0,
|
| 1448 |
+
"reward": 0.9988839328289032,
|
| 1449 |
+
"reward_std": 0.0031567267142236233,
|
| 1450 |
+
"rewards/equation_reward_func": 0.0,
|
| 1451 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 1452 |
+
"step": 222
|
| 1453 |
+
},
|
| 1454 |
+
{
|
| 1455 |
+
"completion_length": 241.00559043884277,
|
| 1456 |
+
"epoch": 0.27873697309068285,
|
| 1457 |
+
"grad_norm": 0.006856945865958338,
|
| 1458 |
+
"kl": 0.031402587890625,
|
| 1459 |
+
"learning_rate": 2.6440299595614606e-07,
|
| 1460 |
+
"loss": 0.0,
|
| 1461 |
+
"reward": 0.9966517984867096,
|
| 1462 |
+
"reward_std": 0.00947018014267087,
|
| 1463 |
+
"rewards/equation_reward_func": 0.0,
|
| 1464 |
+
"rewards/format_reward_func": 0.9966517984867096,
|
| 1465 |
+
"step": 224
|
| 1466 |
+
},
|
| 1467 |
+
{
|
| 1468 |
+
"completion_length": 242.82032585144043,
|
| 1469 |
+
"epoch": 0.2812256960647068,
|
| 1470 |
+
"grad_norm": 0.011147423978654824,
|
| 1471 |
+
"kl": 0.03250885009765625,
|
| 1472 |
+
"learning_rate": 2.6080486500209347e-07,
|
| 1473 |
+
"loss": 0.0,
|
| 1474 |
+
"reward": 0.9966517984867096,
|
| 1475 |
+
"reward_std": 0.00947018014267087,
|
| 1476 |
+
"rewards/equation_reward_func": 0.0,
|
| 1477 |
+
"rewards/format_reward_func": 0.9966517984867096,
|
| 1478 |
+
"step": 226
|
| 1479 |
+
},
|
| 1480 |
+
{
|
| 1481 |
+
"completion_length": 245.30581665039062,
|
| 1482 |
+
"epoch": 0.28371441903873074,
|
| 1483 |
+
"grad_norm": 0.008917482676403654,
|
| 1484 |
+
"kl": 0.03244781494140625,
|
| 1485 |
+
"learning_rate": 2.572044901734166e-07,
|
| 1486 |
+
"loss": 0.0,
|
| 1487 |
+
"reward": 0.9966517984867096,
|
| 1488 |
+
"reward_std": 0.00947018014267087,
|
| 1489 |
+
"rewards/equation_reward_func": 0.0,
|
| 1490 |
+
"rewards/format_reward_func": 0.9966517984867096,
|
| 1491 |
+
"step": 228
|
| 1492 |
+
},
|
| 1493 |
+
{
|
| 1494 |
+
"completion_length": 239.4486722946167,
|
| 1495 |
+
"epoch": 0.2862031420127547,
|
| 1496 |
+
"grad_norm": 0.0012479917659233364,
|
| 1497 |
+
"kl": 0.0322265625,
|
| 1498 |
+
"learning_rate": 2.536026191693893e-07,
|
| 1499 |
+
"loss": 0.0,
|
| 1500 |
+
"reward": 0.9988839328289032,
|
| 1501 |
+
"reward_std": 0.0031567267142236233,
|
| 1502 |
+
"rewards/equation_reward_func": 0.0,
|
| 1503 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 1504 |
+
"step": 230
|
| 1505 |
+
},
|
| 1506 |
+
{
|
| 1507 |
+
"completion_length": 240.26452255249023,
|
| 1508 |
+
"epoch": 0.2886918649867787,
|
| 1509 |
+
"grad_norm": 0.007817347570993229,
|
| 1510 |
+
"kl": 0.03277587890625,
|
| 1511 |
+
"learning_rate": 2.5e-07,
|
| 1512 |
+
"loss": 0.0,
|
| 1513 |
+
"reward": 0.9966517984867096,
|
| 1514 |
+
"reward_std": 0.00947018014267087,
|
| 1515 |
+
"rewards/equation_reward_func": 0.0,
|
| 1516 |
+
"rewards/format_reward_func": 0.9966517984867096,
|
| 1517 |
+
"step": 232
|
| 1518 |
+
},
|
| 1519 |
+
{
|
| 1520 |
+
"completion_length": 242.38840198516846,
|
| 1521 |
+
"epoch": 0.2911805879608026,
|
| 1522 |
+
"grad_norm": 0.004794984024963255,
|
| 1523 |
+
"kl": 0.03249359130859375,
|
| 1524 |
+
"learning_rate": 2.4639738083061073e-07,
|
| 1525 |
+
"loss": 0.0,
|
| 1526 |
+
"reward": 0.9988839328289032,
|
| 1527 |
+
"reward_std": 0.0031567267142236233,
|
| 1528 |
+
"rewards/equation_reward_func": 0.0,
|
| 1529 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 1530 |
+
"step": 234
|
| 1531 |
+
},
|
| 1532 |
+
{
|
| 1533 |
+
"completion_length": 234.61161708831787,
|
| 1534 |
+
"epoch": 0.29366931093482657,
|
| 1535 |
+
"grad_norm": 0.001402625251204755,
|
| 1536 |
+
"kl": 0.03316497802734375,
|
| 1537 |
+
"learning_rate": 2.4279550982658345e-07,
|
| 1538 |
+
"loss": 0.0,
|
| 1539 |
+
"reward": 1.0,
|
| 1540 |
+
"reward_std": 0.0,
|
| 1541 |
+
"rewards/equation_reward_func": 0.0,
|
| 1542 |
+
"rewards/format_reward_func": 1.0,
|
| 1543 |
+
"step": 236
|
| 1544 |
+
},
|
| 1545 |
+
{
|
| 1546 |
+
"completion_length": 231.462064743042,
|
| 1547 |
+
"epoch": 0.29615803390885054,
|
| 1548 |
+
"grad_norm": 0.01673664168050879,
|
| 1549 |
+
"kl": 0.03163909912109375,
|
| 1550 |
+
"learning_rate": 2.3919513499790646e-07,
|
| 1551 |
+
"loss": 0.0,
|
| 1552 |
+
"reward": 0.9988839328289032,
|
| 1553 |
+
"reward_std": 0.0031567267142236233,
|
| 1554 |
+
"rewards/equation_reward_func": 0.0,
|
| 1555 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 1556 |
+
"step": 238
|
| 1557 |
+
},
|
| 1558 |
+
{
|
| 1559 |
+
"completion_length": 241.8259038925171,
|
| 1560 |
+
"epoch": 0.29864675688287445,
|
| 1561 |
+
"grad_norm": 0.0021634188282300413,
|
| 1562 |
+
"kl": 0.034698486328125,
|
| 1563 |
+
"learning_rate": 2.3559700404385394e-07,
|
| 1564 |
+
"loss": 0.0,
|
| 1565 |
+
"reward": 0.9988839328289032,
|
| 1566 |
+
"reward_std": 0.0031567267142236233,
|
| 1567 |
+
"rewards/equation_reward_func": 0.0,
|
| 1568 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 1569 |
+
"step": 240
|
| 1570 |
+
},
|
| 1571 |
+
{
|
| 1572 |
+
"completion_length": 240.47992134094238,
|
| 1573 |
+
"epoch": 0.3011354798568984,
|
| 1574 |
+
"grad_norm": 0.008658887288561594,
|
| 1575 |
+
"kl": 0.03231048583984375,
|
| 1576 |
+
"learning_rate": 2.3200186419770823e-07,
|
| 1577 |
+
"loss": 0.0,
|
| 1578 |
+
"reward": 0.9977678656578064,
|
| 1579 |
+
"reward_std": 0.0063134534284472466,
|
| 1580 |
+
"rewards/equation_reward_func": 0.0,
|
| 1581 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 1582 |
+
"step": 242
|
| 1583 |
+
},
|
| 1584 |
+
{
|
| 1585 |
+
"completion_length": 239.5848331451416,
|
| 1586 |
+
"epoch": 0.3036242028309224,
|
| 1587 |
+
"grad_norm": 0.0011648905322271979,
|
| 1588 |
+
"kl": 0.03226470947265625,
|
| 1589 |
+
"learning_rate": 2.284104620715807e-07,
|
| 1590 |
+
"loss": 0.0,
|
| 1591 |
+
"reward": 1.0,
|
| 1592 |
+
"reward_std": 0.0,
|
| 1593 |
+
"rewards/equation_reward_func": 0.0,
|
| 1594 |
+
"rewards/format_reward_func": 1.0,
|
| 1595 |
+
"step": 244
|
| 1596 |
+
},
|
| 1597 |
+
{
|
| 1598 |
+
"completion_length": 230.66853523254395,
|
| 1599 |
+
"epoch": 0.3061129258049463,
|
| 1600 |
+
"grad_norm": 0.00123117657873765,
|
| 1601 |
+
"kl": 0.03214263916015625,
|
| 1602 |
+
"learning_rate": 2.2482354350136043e-07,
|
| 1603 |
+
"loss": 0.0,
|
| 1604 |
+
"reward": 0.9977678656578064,
|
| 1605 |
+
"reward_std": 0.0063134534284472466,
|
| 1606 |
+
"rewards/equation_reward_func": 0.0,
|
| 1607 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 1608 |
+
"step": 246
|
| 1609 |
+
},
|
| 1610 |
+
{
|
| 1611 |
+
"completion_length": 231.6562623977661,
|
| 1612 |
+
"epoch": 0.3086016487789703,
|
| 1613 |
+
"grad_norm": 0.0011476508674390896,
|
| 1614 |
+
"kl": 0.0305328369140625,
|
| 1615 |
+
"learning_rate": 2.2124185339182496e-07,
|
| 1616 |
+
"loss": 0.0,
|
| 1617 |
+
"reward": 1.0,
|
| 1618 |
+
"reward_std": 0.0,
|
| 1619 |
+
"rewards/equation_reward_func": 0.0,
|
| 1620 |
+
"rewards/format_reward_func": 1.0,
|
| 1621 |
+
"step": 248
|
| 1622 |
+
},
|
| 1623 |
+
{
|
| 1624 |
+
"completion_length": 239.59152603149414,
|
| 1625 |
+
"epoch": 0.31109037175299425,
|
| 1626 |
+
"grad_norm": 0.001092011224535122,
|
| 1627 |
+
"kl": 0.02968597412109375,
|
| 1628 |
+
"learning_rate": 2.1766613556194344e-07,
|
| 1629 |
+
"loss": 0.0,
|
| 1630 |
+
"reward": 0.9988839328289032,
|
| 1631 |
+
"reward_std": 0.0031567267142236233,
|
| 1632 |
+
"rewards/equation_reward_func": 0.0,
|
| 1633 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 1634 |
+
"step": 250
|
| 1635 |
+
},
|
| 1636 |
+
{
|
| 1637 |
+
"completion_length": 233.17523384094238,
|
| 1638 |
+
"epoch": 0.3135790947270182,
|
| 1639 |
+
"grad_norm": 0.0011917811528517187,
|
| 1640 |
+
"kl": 0.0297393798828125,
|
| 1641 |
+
"learning_rate": 2.1409713259040628e-07,
|
| 1642 |
+
"loss": 0.0,
|
| 1643 |
+
"reward": 0.9988839328289032,
|
| 1644 |
+
"reward_std": 0.0031567267142236233,
|
| 1645 |
+
"rewards/equation_reward_func": 0.0,
|
| 1646 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 1647 |
+
"step": 252
|
| 1648 |
+
},
|
| 1649 |
+
{
|
| 1650 |
+
"completion_length": 240.12166118621826,
|
| 1651 |
+
"epoch": 0.31606781770104214,
|
| 1652 |
+
"grad_norm": 0.005939716498084994,
|
| 1653 |
+
"kl": 0.03424072265625,
|
| 1654 |
+
"learning_rate": 2.105355856614115e-07,
|
| 1655 |
+
"loss": 0.0,
|
| 1656 |
+
"reward": 0.9977678656578064,
|
| 1657 |
+
"reward_std": 0.0063134534284472466,
|
| 1658 |
+
"rewards/equation_reward_func": 0.0,
|
| 1659 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 1660 |
+
"step": 254
|
| 1661 |
+
},
|
| 1662 |
+
{
|
| 1663 |
+
"completion_length": 237.22322368621826,
|
| 1664 |
+
"epoch": 0.3185565406750661,
|
| 1665 |
+
"grad_norm": 0.007985852144002343,
|
| 1666 |
+
"kl": 0.03028106689453125,
|
| 1667 |
+
"learning_rate": 2.069822344107413e-07,
|
| 1668 |
+
"loss": 0.0,
|
| 1669 |
+
"reward": 0.9977678656578064,
|
| 1670 |
+
"reward_std": 0.0063134534284472466,
|
| 1671 |
+
"rewards/equation_reward_func": 0.0,
|
| 1672 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 1673 |
+
"step": 256
|
| 1674 |
+
},
|
| 1675 |
+
{
|
| 1676 |
+
"completion_length": 230.8951005935669,
|
| 1677 |
+
"epoch": 0.3210452636490901,
|
| 1678 |
+
"grad_norm": 0.0012251477176528757,
|
| 1679 |
+
"kl": 0.03383636474609375,
|
| 1680 |
+
"learning_rate": 2.034378167721599e-07,
|
| 1681 |
+
"loss": 0.0,
|
| 1682 |
+
"reward": 0.9988839328289032,
|
| 1683 |
+
"reward_std": 0.0031567267142236233,
|
| 1684 |
+
"rewards/equation_reward_func": 0.0,
|
| 1685 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 1686 |
+
"step": 258
|
| 1687 |
+
},
|
| 1688 |
+
{
|
| 1689 |
+
"completion_length": 241.6506814956665,
|
| 1690 |
+
"epoch": 0.323533986623114,
|
| 1691 |
+
"grad_norm": 0.005073098129706296,
|
| 1692 |
+
"kl": 0.0304107666015625,
|
| 1693 |
+
"learning_rate": 1.9990306882416485e-07,
|
| 1694 |
+
"loss": 0.0,
|
| 1695 |
+
"reward": 0.9988839328289032,
|
| 1696 |
+
"reward_std": 0.0031567267142236233,
|
| 1697 |
+
"rewards/equation_reward_func": 0.0,
|
| 1698 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 1699 |
+
"step": 260
|
| 1700 |
+
},
|
| 1701 |
+
{
|
| 1702 |
+
"completion_length": 237.0290288925171,
|
| 1703 |
+
"epoch": 0.32602270959713797,
|
| 1704 |
+
"grad_norm": 0.0011250188792632052,
|
| 1705 |
+
"kl": 0.0294952392578125,
|
| 1706 |
+
"learning_rate": 1.9637872463712362e-07,
|
| 1707 |
+
"loss": 0.0,
|
| 1708 |
+
"reward": 1.0,
|
| 1709 |
+
"reward_std": 0.0,
|
| 1710 |
+
"rewards/equation_reward_func": 0.0,
|
| 1711 |
+
"rewards/format_reward_func": 1.0,
|
| 1712 |
+
"step": 262
|
| 1713 |
+
},
|
| 1714 |
+
{
|
| 1715 |
+
"completion_length": 241.60603618621826,
|
| 1716 |
+
"epoch": 0.32851143257116194,
|
| 1717 |
+
"grad_norm": 0.001159198174006997,
|
| 1718 |
+
"kl": 0.03064727783203125,
|
| 1719 |
+
"learning_rate": 1.9286551612082773e-07,
|
| 1720 |
+
"loss": 0.0,
|
| 1721 |
+
"reward": 0.9988839328289032,
|
| 1722 |
+
"reward_std": 0.0031567267142236233,
|
| 1723 |
+
"rewards/equation_reward_func": 0.0,
|
| 1724 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 1725 |
+
"step": 264
|
| 1726 |
+
},
|
| 1727 |
+
{
|
| 1728 |
+
"completion_length": 233.5770206451416,
|
| 1729 |
+
"epoch": 0.33100015554518586,
|
| 1730 |
+
"grad_norm": 0.008148989229574561,
|
| 1731 |
+
"kl": 0.02930450439453125,
|
| 1732 |
+
"learning_rate": 1.8936417287249446e-07,
|
| 1733 |
+
"loss": 0.0,
|
| 1734 |
+
"reward": 0.9977678656578064,
|
| 1735 |
+
"reward_std": 0.0063134534284472466,
|
| 1736 |
+
"rewards/equation_reward_func": 0.0,
|
| 1737 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 1738 |
+
"step": 266
|
| 1739 |
+
},
|
| 1740 |
+
{
|
| 1741 |
+
"completion_length": 246.89398384094238,
|
| 1742 |
+
"epoch": 0.33348887851920983,
|
| 1743 |
+
"grad_norm": 0.016173200021972613,
|
| 1744 |
+
"kl": 0.02889251708984375,
|
| 1745 |
+
"learning_rate": 1.8587542202524985e-07,
|
| 1746 |
+
"loss": 0.0,
|
| 1747 |
+
"reward": 0.9944196566939354,
|
| 1748 |
+
"reward_std": 0.015783633571118116,
|
| 1749 |
+
"rewards/equation_reward_func": 0.0,
|
| 1750 |
+
"rewards/format_reward_func": 0.9944196566939354,
|
| 1751 |
+
"step": 268
|
| 1752 |
+
},
|
| 1753 |
+
{
|
| 1754 |
+
"completion_length": 241.57701873779297,
|
| 1755 |
+
"epoch": 0.3359776014932338,
|
| 1756 |
+
"grad_norm": 0.007919979972576274,
|
| 1757 |
+
"kl": 0.0288543701171875,
|
| 1758 |
+
"learning_rate": 1.82399988097123e-07,
|
| 1759 |
+
"loss": 0.0,
|
| 1760 |
+
"reward": 0.9988839328289032,
|
| 1761 |
+
"reward_std": 0.0031567267142236233,
|
| 1762 |
+
"rewards/equation_reward_func": 0.0,
|
| 1763 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 1764 |
+
"step": 270
|
| 1765 |
+
},
|
| 1766 |
+
{
|
| 1767 |
+
"completion_length": 231.98438358306885,
|
| 1768 |
+
"epoch": 0.3384663244672577,
|
| 1769 |
+
"grad_norm": 0.005824865666987302,
|
| 1770 |
+
"kl": 0.0288848876953125,
|
| 1771 |
+
"learning_rate": 1.7893859284058378e-07,
|
| 1772 |
+
"loss": 0.0,
|
| 1773 |
+
"reward": 0.9966517984867096,
|
| 1774 |
+
"reward_std": 0.00947018014267087,
|
| 1775 |
+
"rewards/equation_reward_func": 0.0,
|
| 1776 |
+
"rewards/format_reward_func": 0.9966517984867096,
|
| 1777 |
+
"step": 272
|
| 1778 |
+
},
|
| 1779 |
+
{
|
| 1780 |
+
"completion_length": 240.15403079986572,
|
| 1781 |
+
"epoch": 0.3409550474412817,
|
| 1782 |
+
"grad_norm": 0.005849415216135437,
|
| 1783 |
+
"kl": 0.02838134765625,
|
| 1784 |
+
"learning_rate": 1.7549195509265407e-07,
|
| 1785 |
+
"loss": 0.0,
|
| 1786 |
+
"reward": 0.9988839328289032,
|
| 1787 |
+
"reward_std": 0.0031567267142236233,
|
| 1788 |
+
"rewards/equation_reward_func": 0.0,
|
| 1789 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 1790 |
+
"step": 274
|
| 1791 |
+
},
|
| 1792 |
+
{
|
| 1793 |
+
"completion_length": 236.81586074829102,
|
| 1794 |
+
"epoch": 0.34344377041530566,
|
| 1795 |
+
"grad_norm": 0.014300393698816647,
|
| 1796 |
+
"kl": 0.02855682373046875,
|
| 1797 |
+
"learning_rate": 1.7206079062562536e-07,
|
| 1798 |
+
"loss": 0.0,
|
| 1799 |
+
"reward": 0.9977678656578064,
|
| 1800 |
+
"reward_std": 0.0063134534284472466,
|
| 1801 |
+
"rewards/equation_reward_func": 0.0,
|
| 1802 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 1803 |
+
"step": 276
|
| 1804 |
+
},
|
| 1805 |
+
{
|
| 1806 |
+
"completion_length": 237.93304824829102,
|
| 1807 |
+
"epoch": 0.3459324933893296,
|
| 1808 |
+
"grad_norm": 0.006324763215032033,
|
| 1809 |
+
"kl": 0.02925872802734375,
|
| 1810 |
+
"learning_rate": 1.6864581199841226e-07,
|
| 1811 |
+
"loss": 0.0,
|
| 1812 |
+
"reward": 0.9977678656578064,
|
| 1813 |
+
"reward_std": 0.0063134534284472466,
|
| 1814 |
+
"rewards/equation_reward_func": 0.0,
|
| 1815 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 1816 |
+
"step": 278
|
| 1817 |
+
},
|
| 1818 |
+
{
|
| 1819 |
+
"completion_length": 235.18527793884277,
|
| 1820 |
+
"epoch": 0.34842121636335355,
|
| 1821 |
+
"grad_norm": 0.0011361122835219525,
|
| 1822 |
+
"kl": 0.04534149169921875,
|
| 1823 |
+
"learning_rate": 1.6524772840857388e-07,
|
| 1824 |
+
"loss": 0.0,
|
| 1825 |
+
"reward": 1.0,
|
| 1826 |
+
"reward_std": 0.0,
|
| 1827 |
+
"rewards/equation_reward_func": 0.0,
|
| 1828 |
+
"rewards/format_reward_func": 1.0,
|
| 1829 |
+
"step": 280
|
| 1830 |
+
},
|
| 1831 |
+
{
|
| 1832 |
+
"completion_length": 236.23884868621826,
|
| 1833 |
+
"epoch": 0.3509099393373775,
|
| 1834 |
+
"grad_norm": 0.006624613315554435,
|
| 1835 |
+
"kl": 0.02925872802734375,
|
| 1836 |
+
"learning_rate": 1.6186724554503237e-07,
|
| 1837 |
+
"loss": 0.0,
|
| 1838 |
+
"reward": 0.9977678656578064,
|
| 1839 |
+
"reward_std": 0.0063134534284472466,
|
| 1840 |
+
"rewards/equation_reward_func": 0.0,
|
| 1841 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 1842 |
+
"step": 282
|
| 1843 |
+
},
|
| 1844 |
+
{
|
| 1845 |
+
"completion_length": 233.81363010406494,
|
| 1846 |
+
"epoch": 0.35339866231140143,
|
| 1847 |
+
"grad_norm": 0.0010583585640121704,
|
| 1848 |
+
"kl": 0.02803802490234375,
|
| 1849 |
+
"learning_rate": 1.5850506544152103e-07,
|
| 1850 |
+
"loss": 0.0,
|
| 1851 |
+
"reward": 1.0,
|
| 1852 |
+
"reward_std": 0.0,
|
| 1853 |
+
"rewards/equation_reward_func": 0.0,
|
| 1854 |
+
"rewards/format_reward_func": 1.0,
|
| 1855 |
+
"step": 284
|
| 1856 |
+
},
|
| 1857 |
+
{
|
| 1858 |
+
"completion_length": 235.45648288726807,
|
| 1859 |
+
"epoch": 0.3558873852854254,
|
| 1860 |
+
"grad_norm": 0.006945667793827179,
|
| 1861 |
+
"kl": 0.03301239013671875,
|
| 1862 |
+
"learning_rate": 1.5516188633079107e-07,
|
| 1863 |
+
"loss": 0.0,
|
| 1864 |
+
"reward": 0.9988839328289032,
|
| 1865 |
+
"reward_std": 0.0031567267142236233,
|
| 1866 |
+
"rewards/equation_reward_func": 0.0,
|
| 1867 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 1868 |
+
"step": 286
|
| 1869 |
+
},
|
| 1870 |
+
{
|
| 1871 |
+
"completion_length": 232.47210884094238,
|
| 1872 |
+
"epoch": 0.3583761082594494,
|
| 1873 |
+
"grad_norm": 0.0058822702186793144,
|
| 1874 |
+
"kl": 0.03041839599609375,
|
| 1875 |
+
"learning_rate": 1.5183840249960784e-07,
|
| 1876 |
+
"loss": 0.0,
|
| 1877 |
+
"reward": 0.9977678656578064,
|
| 1878 |
+
"reward_std": 0.0063134534284472466,
|
| 1879 |
+
"rewards/equation_reward_func": 0.0,
|
| 1880 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 1881 |
+
"step": 288
|
| 1882 |
+
},
|
| 1883 |
+
{
|
| 1884 |
+
"completion_length": 233.774564743042,
|
| 1885 |
+
"epoch": 0.36086483123347335,
|
| 1886 |
+
"grad_norm": 0.0010805626564454967,
|
| 1887 |
+
"kl": 0.028167724609375,
|
| 1888 |
+
"learning_rate": 1.4853530414456612e-07,
|
| 1889 |
+
"loss": 0.0,
|
| 1890 |
+
"reward": 1.0,
|
| 1891 |
+
"reward_std": 0.0,
|
| 1892 |
+
"rewards/equation_reward_func": 0.0,
|
| 1893 |
+
"rewards/format_reward_func": 1.0,
|
| 1894 |
+
"step": 290
|
| 1895 |
+
},
|
| 1896 |
+
{
|
| 1897 |
+
"completion_length": 237.87054824829102,
|
| 1898 |
+
"epoch": 0.36335355420749726,
|
| 1899 |
+
"grad_norm": 0.005734105971756099,
|
| 1900 |
+
"kl": 0.02758026123046875,
|
| 1901 |
+
"learning_rate": 1.4525327722875568e-07,
|
| 1902 |
+
"loss": 0.0,
|
| 1903 |
+
"reward": 0.9988839328289032,
|
| 1904 |
+
"reward_std": 0.0031567267142236233,
|
| 1905 |
+
"rewards/equation_reward_func": 0.0,
|
| 1906 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 1907 |
+
"step": 292
|
| 1908 |
+
},
|
| 1909 |
+
{
|
| 1910 |
+
"completion_length": 237.0993413925171,
|
| 1911 |
+
"epoch": 0.36584227718152124,
|
| 1912 |
+
"grad_norm": 0.017115197738888723,
|
| 1913 |
+
"kl": 0.0272064208984375,
|
| 1914 |
+
"learning_rate": 1.4199300333930515e-07,
|
| 1915 |
+
"loss": 0.0,
|
| 1916 |
+
"reward": 0.9955357238650322,
|
| 1917 |
+
"reward_std": 0.012626906856894493,
|
| 1918 |
+
"rewards/equation_reward_func": 0.0,
|
| 1919 |
+
"rewards/format_reward_func": 0.9955357238650322,
|
| 1920 |
+
"step": 294
|
| 1921 |
+
},
|
| 1922 |
+
{
|
| 1923 |
+
"completion_length": 232.11942958831787,
|
| 1924 |
+
"epoch": 0.3683310001555452,
|
| 1925 |
+
"grad_norm": 0.001178682970890626,
|
| 1926 |
+
"kl": 0.02896881103515625,
|
| 1927 |
+
"learning_rate": 1.3875515954583523e-07,
|
| 1928 |
+
"loss": 0.0,
|
| 1929 |
+
"reward": 0.9988839328289032,
|
| 1930 |
+
"reward_std": 0.0031567267142236233,
|
| 1931 |
+
"rewards/equation_reward_func": 0.0,
|
| 1932 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 1933 |
+
"step": 296
|
| 1934 |
+
},
|
| 1935 |
+
{
|
| 1936 |
+
"completion_length": 241.54911708831787,
|
| 1937 |
+
"epoch": 0.3708197231295691,
|
| 1938 |
+
"grad_norm": 0.01055702171846327,
|
| 1939 |
+
"kl": 0.02869415283203125,
|
| 1940 |
+
"learning_rate": 1.3554041825985e-07,
|
| 1941 |
+
"loss": 0.0,
|
| 1942 |
+
"reward": 0.9966517984867096,
|
| 1943 |
+
"reward_std": 0.00947018014267087,
|
| 1944 |
+
"rewards/equation_reward_func": 0.0,
|
| 1945 |
+
"rewards/format_reward_func": 0.9966517984867096,
|
| 1946 |
+
"step": 298
|
| 1947 |
+
},
|
| 1948 |
+
{
|
| 1949 |
+
"completion_length": 235.68304824829102,
|
| 1950 |
+
"epoch": 0.3733084461035931,
|
| 1951 |
+
"grad_norm": 0.006778222288303374,
|
| 1952 |
+
"kl": 0.0286407470703125,
|
| 1953 |
+
"learning_rate": 1.323494470950949e-07,
|
| 1954 |
+
"loss": 0.0,
|
| 1955 |
+
"reward": 0.9977678656578064,
|
| 1956 |
+
"reward_std": 0.0063134534284472466,
|
| 1957 |
+
"rewards/equation_reward_func": 0.0,
|
| 1958 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 1959 |
+
"step": 300
|
| 1960 |
+
},
|
| 1961 |
+
{
|
| 1962 |
+
"completion_length": 238.79130935668945,
|
| 1963 |
+
"epoch": 0.37579716907761707,
|
| 1964 |
+
"grad_norm": 0.008015757124174978,
|
| 1965 |
+
"kl": 0.02869415283203125,
|
| 1966 |
+
"learning_rate": 1.2918290872891236e-07,
|
| 1967 |
+
"loss": 0.0,
|
| 1968 |
+
"reward": 0.9966517984867096,
|
| 1969 |
+
"reward_std": 0.00947018014267087,
|
| 1970 |
+
"rewards/equation_reward_func": 0.0,
|
| 1971 |
+
"rewards/format_reward_func": 0.9966517984867096,
|
| 1972 |
+
"step": 302
|
| 1973 |
+
},
|
| 1974 |
+
{
|
| 1975 |
+
"completion_length": 233.8794755935669,
|
| 1976 |
+
"epoch": 0.378285892051641,
|
| 1977 |
+
"grad_norm": 0.0009938073180528301,
|
| 1978 |
+
"kl": 0.02816009521484375,
|
| 1979 |
+
"learning_rate": 1.260414607646213e-07,
|
| 1980 |
+
"loss": 0.0,
|
| 1981 |
+
"reward": 1.0,
|
| 1982 |
+
"reward_std": 0.0,
|
| 1983 |
+
"rewards/equation_reward_func": 0.0,
|
| 1984 |
+
"rewards/format_reward_func": 1.0,
|
| 1985 |
+
"step": 304
|
| 1986 |
+
},
|
| 1987 |
+
{
|
| 1988 |
+
"completion_length": 242.52456665039062,
|
| 1989 |
+
"epoch": 0.38077461502566495,
|
| 1990 |
+
"grad_norm": 0.007613603270829063,
|
| 1991 |
+
"kl": 0.02838897705078125,
|
| 1992 |
+
"learning_rate": 1.2292575559495143e-07,
|
| 1993 |
+
"loss": 0.0,
|
| 1994 |
+
"reward": 0.9977678656578064,
|
| 1995 |
+
"reward_std": 0.0063134534284472466,
|
| 1996 |
+
"rewards/equation_reward_func": 0.0,
|
| 1997 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 1998 |
+
"step": 306
|
| 1999 |
+
},
|
| 2000 |
+
{
|
| 2001 |
+
"completion_length": 236.0591630935669,
|
| 2002 |
+
"epoch": 0.3832633379996889,
|
| 2003 |
+
"grad_norm": 0.007171730472972239,
|
| 2004 |
+
"kl": 0.028594970703125,
|
| 2005 |
+
"learning_rate": 1.1983644026655835e-07,
|
| 2006 |
+
"loss": 0.0,
|
| 2007 |
+
"reward": 0.9966517984867096,
|
| 2008 |
+
"reward_std": 0.00947018014267087,
|
| 2009 |
+
"rewards/equation_reward_func": 0.0,
|
| 2010 |
+
"rewards/format_reward_func": 0.9966517984867096,
|
| 2011 |
+
"step": 308
|
| 2012 |
+
},
|
| 2013 |
+
{
|
| 2014 |
+
"completion_length": 241.35604000091553,
|
| 2015 |
+
"epoch": 0.38575206097371284,
|
| 2016 |
+
"grad_norm": 0.0011600397120661158,
|
| 2017 |
+
"kl": 0.02840423583984375,
|
| 2018 |
+
"learning_rate": 1.1677415634565066e-07,
|
| 2019 |
+
"loss": 0.0,
|
| 2020 |
+
"reward": 0.9988839328289032,
|
| 2021 |
+
"reward_std": 0.0031567267142236233,
|
| 2022 |
+
"rewards/equation_reward_func": 0.0,
|
| 2023 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 2024 |
+
"step": 310
|
| 2025 |
+
},
|
| 2026 |
+
{
|
| 2027 |
+
"completion_length": 233.03907203674316,
|
| 2028 |
+
"epoch": 0.3882407839477368,
|
| 2029 |
+
"grad_norm": 0.0010814554021977632,
|
| 2030 |
+
"kl": 0.02860260009765625,
|
| 2031 |
+
"learning_rate": 1.1373953978475353e-07,
|
| 2032 |
+
"loss": 0.0,
|
| 2033 |
+
"reward": 0.9988839328289032,
|
| 2034 |
+
"reward_std": 0.0031567267142236233,
|
| 2035 |
+
"rewards/equation_reward_func": 0.0,
|
| 2036 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 2037 |
+
"step": 312
|
| 2038 |
+
},
|
| 2039 |
+
{
|
| 2040 |
+
"completion_length": 237.86831760406494,
|
| 2041 |
+
"epoch": 0.3907295069217608,
|
| 2042 |
+
"grad_norm": 0.0010944470027230678,
|
| 2043 |
+
"kl": 0.02793121337890625,
|
| 2044 |
+
"learning_rate": 1.1073322079063913e-07,
|
| 2045 |
+
"loss": 0.0,
|
| 2046 |
+
"reward": 0.9988839328289032,
|
| 2047 |
+
"reward_std": 0.0031567267142236233,
|
| 2048 |
+
"rewards/equation_reward_func": 0.0,
|
| 2049 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 2050 |
+
"step": 314
|
| 2051 |
+
},
|
| 2052 |
+
{
|
| 2053 |
+
"completion_length": 237.2890748977661,
|
| 2054 |
+
"epoch": 0.3932182298957847,
|
| 2055 |
+
"grad_norm": 0.01903527238338617,
|
| 2056 |
+
"kl": 0.02887725830078125,
|
| 2057 |
+
"learning_rate": 1.0775582369344946e-07,
|
| 2058 |
+
"loss": 0.0,
|
| 2059 |
+
"reward": 0.9955357201397419,
|
| 2060 |
+
"reward_std": 0.012626906391233206,
|
| 2061 |
+
"rewards/equation_reward_func": 0.0,
|
| 2062 |
+
"rewards/format_reward_func": 0.9955357201397419,
|
| 2063 |
+
"step": 316
|
| 2064 |
+
},
|
| 2065 |
+
{
|
| 2066 |
+
"completion_length": 241.7209939956665,
|
| 2067 |
+
"epoch": 0.39570695286980867,
|
| 2068 |
+
"grad_norm": 0.006903717479599212,
|
| 2069 |
+
"kl": 0.02970123291015625,
|
| 2070 |
+
"learning_rate": 1.0480796681704077e-07,
|
| 2071 |
+
"loss": 0.0,
|
| 2072 |
+
"reward": 0.9988839328289032,
|
| 2073 |
+
"reward_std": 0.0031567267142236233,
|
| 2074 |
+
"rewards/equation_reward_func": 0.0,
|
| 2075 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 2076 |
+
"step": 318
|
| 2077 |
+
},
|
| 2078 |
+
{
|
| 2079 |
+
"completion_length": 236.0134038925171,
|
| 2080 |
+
"epoch": 0.39819567584383264,
|
| 2081 |
+
"grad_norm": 0.0011547787615559003,
|
| 2082 |
+
"kl": 0.02854156494140625,
|
| 2083 |
+
"learning_rate": 1.018902623505741e-07,
|
| 2084 |
+
"loss": 0.0,
|
| 2085 |
+
"reward": 1.0,
|
| 2086 |
+
"reward_std": 0.0,
|
| 2087 |
+
"rewards/equation_reward_func": 0.0,
|
| 2088 |
+
"rewards/format_reward_func": 1.0,
|
| 2089 |
+
"step": 320
|
| 2090 |
+
},
|
| 2091 |
+
{
|
| 2092 |
+
"completion_length": 241.60826969146729,
|
| 2093 |
+
"epoch": 0.4006843988178566,
|
| 2094 |
+
"grad_norm": 0.0010017599761343319,
|
| 2095 |
+
"kl": 0.02718353271484375,
|
| 2096 |
+
"learning_rate": 9.900331622138063e-08,
|
| 2097 |
+
"loss": 0.0,
|
| 2098 |
+
"reward": 0.9988839328289032,
|
| 2099 |
+
"reward_std": 0.0031567267142236233,
|
| 2100 |
+
"rewards/equation_reward_func": 0.0,
|
| 2101 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 2102 |
+
"step": 322
|
| 2103 |
+
},
|
| 2104 |
+
{
|
| 2105 |
+
"completion_length": 233.4330472946167,
|
| 2106 |
+
"epoch": 0.40317312179188053,
|
| 2107 |
+
"grad_norm": 0.0013114923906156683,
|
| 2108 |
+
"kl": 0.02910614013671875,
|
| 2109 |
+
"learning_rate": 9.614772796912681e-08,
|
| 2110 |
+
"loss": 0.0,
|
| 2111 |
+
"reward": 0.996651791036129,
|
| 2112 |
+
"reward_std": 0.00947018014267087,
|
| 2113 |
+
"rewards/equation_reward_func": 0.0,
|
| 2114 |
+
"rewards/format_reward_func": 0.996651791036129,
|
| 2115 |
+
"step": 324
|
| 2116 |
+
},
|
| 2117 |
+
{
|
| 2118 |
+
"completion_length": 229.5357255935669,
|
| 2119 |
+
"epoch": 0.4056618447659045,
|
| 2120 |
+
"grad_norm": 0.0011343461889418145,
|
| 2121 |
+
"kl": 0.0290069580078125,
|
| 2122 |
+
"learning_rate": 9.332409062130686e-08,
|
| 2123 |
+
"loss": 0.0,
|
| 2124 |
+
"reward": 0.9988839328289032,
|
| 2125 |
+
"reward_std": 0.0031567267142236233,
|
| 2126 |
+
"rewards/equation_reward_func": 0.0,
|
| 2127 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 2128 |
+
"step": 326
|
| 2129 |
+
},
|
| 2130 |
+
{
|
| 2131 |
+
"completion_length": 241.87501335144043,
|
| 2132 |
+
"epoch": 0.40815056773992847,
|
| 2133 |
+
"grad_norm": 0.0066577720593599425,
|
| 2134 |
+
"kl": 0.0284423828125,
|
| 2135 |
+
"learning_rate": 9.053299057008699e-08,
|
| 2136 |
+
"loss": 0.0,
|
| 2137 |
+
"reward": 0.9977678656578064,
|
| 2138 |
+
"reward_std": 0.0063134534284472466,
|
| 2139 |
+
"rewards/equation_reward_func": 0.0,
|
| 2140 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 2141 |
+
"step": 328
|
| 2142 |
+
},
|
| 2143 |
+
{
|
| 2144 |
+
"completion_length": 229.70648288726807,
|
| 2145 |
+
"epoch": 0.4106392907139524,
|
| 2146 |
+
"grad_norm": 0.0010762623483081323,
|
| 2147 |
+
"kl": 0.02895355224609375,
|
| 2148 |
+
"learning_rate": 8.777500745052743e-08,
|
| 2149 |
+
"loss": 0.0,
|
| 2150 |
+
"reward": 1.0,
|
| 2151 |
+
"reward_std": 0.0,
|
| 2152 |
+
"rewards/equation_reward_func": 0.0,
|
| 2153 |
+
"rewards/format_reward_func": 1.0,
|
| 2154 |
+
"step": 330
|
| 2155 |
+
},
|
| 2156 |
+
{
|
| 2157 |
+
"completion_length": 229.516752243042,
|
| 2158 |
+
"epoch": 0.41312801368797636,
|
| 2159 |
+
"grad_norm": 0.001064089704651314,
|
| 2160 |
+
"kl": 0.02860260009765625,
|
| 2161 |
+
"learning_rate": 8.505071402020892e-08,
|
| 2162 |
+
"loss": 0.0,
|
| 2163 |
+
"reward": 1.0,
|
| 2164 |
+
"reward_std": 0.0,
|
| 2165 |
+
"rewards/equation_reward_func": 0.0,
|
| 2166 |
+
"rewards/format_reward_func": 1.0,
|
| 2167 |
+
"step": 332
|
| 2168 |
+
},
|
| 2169 |
+
{
|
| 2170 |
+
"completion_length": 238.25670719146729,
|
| 2171 |
+
"epoch": 0.41561673666200033,
|
| 2172 |
+
"grad_norm": 0.00645868519135834,
|
| 2173 |
+
"kl": 0.03199005126953125,
|
| 2174 |
+
"learning_rate": 8.236067604028562e-08,
|
| 2175 |
+
"loss": 0.0,
|
| 2176 |
+
"reward": 0.9977678656578064,
|
| 2177 |
+
"reward_std": 0.0063134534284472466,
|
| 2178 |
+
"rewards/equation_reward_func": 0.0,
|
| 2179 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 2180 |
+
"step": 334
|
| 2181 |
+
},
|
| 2182 |
+
{
|
| 2183 |
+
"completion_length": 233.23773574829102,
|
| 2184 |
+
"epoch": 0.41810545963602425,
|
| 2185 |
+
"grad_norm": 0.0015168370588455773,
|
| 2186 |
+
"kl": 0.02906036376953125,
|
| 2187 |
+
"learning_rate": 7.970545215799327e-08,
|
| 2188 |
+
"loss": 0.0,
|
| 2189 |
+
"reward": 1.0,
|
| 2190 |
+
"reward_std": 0.0,
|
| 2191 |
+
"rewards/equation_reward_func": 0.0,
|
| 2192 |
+
"rewards/format_reward_func": 1.0,
|
| 2193 |
+
"step": 336
|
| 2194 |
+
},
|
| 2195 |
+
{
|
| 2196 |
+
"completion_length": 233.602689743042,
|
| 2197 |
+
"epoch": 0.4205941826100482,
|
| 2198 |
+
"grad_norm": 0.0010219844895423763,
|
| 2199 |
+
"kl": 0.02813720703125,
|
| 2200 |
+
"learning_rate": 7.708559379063204e-08,
|
| 2201 |
+
"loss": 0.0,
|
| 2202 |
+
"reward": 0.9988839328289032,
|
| 2203 |
+
"reward_std": 0.0031567267142236233,
|
| 2204 |
+
"rewards/equation_reward_func": 0.0,
|
| 2205 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 2206 |
+
"step": 338
|
| 2207 |
+
},
|
| 2208 |
+
{
|
| 2209 |
+
"completion_length": 231.45759773254395,
|
| 2210 |
+
"epoch": 0.4230829055840722,
|
| 2211 |
+
"grad_norm": 0.0013665815983744528,
|
| 2212 |
+
"kl": 0.030242919921875,
|
| 2213 |
+
"learning_rate": 7.45016450110534e-08,
|
| 2214 |
+
"loss": 0.0,
|
| 2215 |
+
"reward": 1.0,
|
| 2216 |
+
"reward_std": 0.0,
|
| 2217 |
+
"rewards/equation_reward_func": 0.0,
|
| 2218 |
+
"rewards/format_reward_func": 1.0,
|
| 2219 |
+
"step": 340
|
| 2220 |
+
},
|
| 2221 |
+
{
|
| 2222 |
+
"completion_length": 238.58929824829102,
|
| 2223 |
+
"epoch": 0.4255716285580961,
|
| 2224 |
+
"grad_norm": 0.0009993803897500766,
|
| 2225 |
+
"kl": 0.02672576904296875,
|
| 2226 |
+
"learning_rate": 7.195414243467029e-08,
|
| 2227 |
+
"loss": 0.0,
|
| 2228 |
+
"reward": 1.0,
|
| 2229 |
+
"reward_std": 0.0,
|
| 2230 |
+
"rewards/equation_reward_func": 0.0,
|
| 2231 |
+
"rewards/format_reward_func": 1.0,
|
| 2232 |
+
"step": 342
|
| 2233 |
+
},
|
| 2234 |
+
{
|
| 2235 |
+
"completion_length": 227.868314743042,
|
| 2236 |
+
"epoch": 0.4280603515321201,
|
| 2237 |
+
"grad_norm": 0.015796748341056373,
|
| 2238 |
+
"kl": 0.0280303955078125,
|
| 2239 |
+
"learning_rate": 6.944361510801763e-08,
|
| 2240 |
+
"loss": 0.0,
|
| 2241 |
+
"reward": 0.9988839328289032,
|
| 2242 |
+
"reward_std": 0.0031567267142236233,
|
| 2243 |
+
"rewards/equation_reward_func": 0.0,
|
| 2244 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 2245 |
+
"step": 344
|
| 2246 |
+
},
|
| 2247 |
+
{
|
| 2248 |
+
"completion_length": 236.65626049041748,
|
| 2249 |
+
"epoch": 0.43054907450614405,
|
| 2250 |
+
"grad_norm": 0.0010452832449649905,
|
| 2251 |
+
"kl": 0.02677154541015625,
|
| 2252 |
+
"learning_rate": 6.697058439888283e-08,
|
| 2253 |
+
"loss": 0.0,
|
| 2254 |
+
"reward": 1.0,
|
| 2255 |
+
"reward_std": 0.0,
|
| 2256 |
+
"rewards/equation_reward_func": 0.0,
|
| 2257 |
+
"rewards/format_reward_func": 1.0,
|
| 2258 |
+
"step": 346
|
| 2259 |
+
},
|
| 2260 |
+
{
|
| 2261 |
+
"completion_length": 230.65514469146729,
|
| 2262 |
+
"epoch": 0.43303779748016796,
|
| 2263 |
+
"grad_norm": 0.001002506595208053,
|
| 2264 |
+
"kl": 0.02809906005859375,
|
| 2265 |
+
"learning_rate": 6.453556388803288e-08,
|
| 2266 |
+
"loss": 0.0,
|
| 2267 |
+
"reward": 1.0,
|
| 2268 |
+
"reward_std": 0.0,
|
| 2269 |
+
"rewards/equation_reward_func": 0.0,
|
| 2270 |
+
"rewards/format_reward_func": 1.0,
|
| 2271 |
+
"step": 348
|
| 2272 |
+
},
|
| 2273 |
+
{
|
| 2274 |
+
"completion_length": 242.3861722946167,
|
| 2275 |
+
"epoch": 0.43552652045419193,
|
| 2276 |
+
"grad_norm": 0.008545847497911966,
|
| 2277 |
+
"kl": 0.02759552001953125,
|
| 2278 |
+
"learning_rate": 6.213905926255697e-08,
|
| 2279 |
+
"loss": 0.0,
|
| 2280 |
+
"reward": 0.9966517984867096,
|
| 2281 |
+
"reward_std": 0.00947018014267087,
|
| 2282 |
+
"rewards/equation_reward_func": 0.0,
|
| 2283 |
+
"rewards/format_reward_func": 0.9966517984867096,
|
| 2284 |
+
"step": 350
|
| 2285 |
+
},
|
| 2286 |
+
{
|
| 2287 |
+
"completion_length": 232.85268878936768,
|
| 2288 |
+
"epoch": 0.4380152434282159,
|
| 2289 |
+
"grad_norm": 0.0011257932930819766,
|
| 2290 |
+
"kl": 0.029296875,
|
| 2291 |
+
"learning_rate": 5.978156821084987e-08,
|
| 2292 |
+
"loss": 0.0,
|
| 2293 |
+
"reward": 1.0,
|
| 2294 |
+
"reward_std": 0.0,
|
| 2295 |
+
"rewards/equation_reward_func": 0.0,
|
| 2296 |
+
"rewards/format_reward_func": 1.0,
|
| 2297 |
+
"step": 352
|
| 2298 |
+
},
|
| 2299 |
+
{
|
| 2300 |
+
"completion_length": 233.83483219146729,
|
| 2301 |
+
"epoch": 0.4405039664022399,
|
| 2302 |
+
"grad_norm": 0.0012045288926858105,
|
| 2303 |
+
"kl": 0.0295257568359375,
|
| 2304 |
+
"learning_rate": 5.7463580319254853e-08,
|
| 2305 |
+
"loss": 0.0,
|
| 2306 |
+
"reward": 1.0,
|
| 2307 |
+
"reward_std": 0.0,
|
| 2308 |
+
"rewards/equation_reward_func": 0.0,
|
| 2309 |
+
"rewards/format_reward_func": 1.0,
|
| 2310 |
+
"step": 354
|
| 2311 |
+
},
|
| 2312 |
+
{
|
| 2313 |
+
"completion_length": 242.27233219146729,
|
| 2314 |
+
"epoch": 0.4429926893762638,
|
| 2315 |
+
"grad_norm": 0.004569960372352263,
|
| 2316 |
+
"kl": 0.02753448486328125,
|
| 2317 |
+
"learning_rate": 5.518557697039081e-08,
|
| 2318 |
+
"loss": 0.0,
|
| 2319 |
+
"reward": 0.9966517984867096,
|
| 2320 |
+
"reward_std": 0.00947018014267087,
|
| 2321 |
+
"rewards/equation_reward_func": 0.0,
|
| 2322 |
+
"rewards/format_reward_func": 0.9966517984867096,
|
| 2323 |
+
"step": 356
|
| 2324 |
+
},
|
| 2325 |
+
{
|
| 2326 |
+
"completion_length": 234.8716630935669,
|
| 2327 |
+
"epoch": 0.44548141235028776,
|
| 2328 |
+
"grad_norm": 0.0010851206277206634,
|
| 2329 |
+
"kl": 0.027191162109375,
|
| 2330 |
+
"learning_rate": 5.294803124318145e-08,
|
| 2331 |
+
"loss": 0.0,
|
| 2332 |
+
"reward": 1.0,
|
| 2333 |
+
"reward_std": 0.0,
|
| 2334 |
+
"rewards/equation_reward_func": 0.0,
|
| 2335 |
+
"rewards/format_reward_func": 1.0,
|
| 2336 |
+
"step": 358
|
| 2337 |
+
},
|
| 2338 |
+
{
|
| 2339 |
+
"completion_length": 238.5435380935669,
|
| 2340 |
+
"epoch": 0.44797013532431174,
|
| 2341 |
+
"grad_norm": 0.005569266348540911,
|
| 2342 |
+
"kl": 0.0275421142578125,
|
| 2343 |
+
"learning_rate": 5.07514078146106e-08,
|
| 2344 |
+
"loss": 0.0,
|
| 2345 |
+
"reward": 0.9988839328289032,
|
| 2346 |
+
"reward_std": 0.0031567267142236233,
|
| 2347 |
+
"rewards/equation_reward_func": 0.0,
|
| 2348 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 2349 |
+
"step": 360
|
| 2350 |
+
},
|
| 2351 |
+
{
|
| 2352 |
+
"completion_length": 236.9576015472412,
|
| 2353 |
+
"epoch": 0.45045885829833565,
|
| 2354 |
+
"grad_norm": 0.020665916015112376,
|
| 2355 |
+
"kl": 0.02768707275390625,
|
| 2356 |
+
"learning_rate": 4.859616286322094e-08,
|
| 2357 |
+
"loss": 0.0,
|
| 2358 |
+
"reward": 0.9977678656578064,
|
| 2359 |
+
"reward_std": 0.0063134534284472466,
|
| 2360 |
+
"rewards/equation_reward_func": 0.0,
|
| 2361 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 2362 |
+
"step": 362
|
| 2363 |
+
},
|
| 2364 |
+
{
|
| 2365 |
+
"completion_length": 239.62054634094238,
|
| 2366 |
+
"epoch": 0.4529475812723596,
|
| 2367 |
+
"grad_norm": 0.006716636201154639,
|
| 2368 |
+
"kl": 0.0279541015625,
|
| 2369 |
+
"learning_rate": 4.648274397437829e-08,
|
| 2370 |
+
"loss": 0.0,
|
| 2371 |
+
"reward": 0.9988839328289032,
|
| 2372 |
+
"reward_std": 0.0031567267142236233,
|
| 2373 |
+
"rewards/equation_reward_func": 0.0,
|
| 2374 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 2375 |
+
"step": 364
|
| 2376 |
+
},
|
| 2377 |
+
{
|
| 2378 |
+
"completion_length": 243.7142972946167,
|
| 2379 |
+
"epoch": 0.4554363042463836,
|
| 2380 |
+
"grad_norm": 0.008375247072193117,
|
| 2381 |
+
"kl": 0.02677154541015625,
|
| 2382 |
+
"learning_rate": 4.4411590047320617e-08,
|
| 2383 |
+
"loss": 0.0,
|
| 2384 |
+
"reward": 0.9955357313156128,
|
| 2385 |
+
"reward_std": 0.012626906856894493,
|
| 2386 |
+
"rewards/equation_reward_func": 0.0,
|
| 2387 |
+
"rewards/format_reward_func": 0.9955357313156128,
|
| 2388 |
+
"step": 366
|
| 2389 |
+
},
|
| 2390 |
+
{
|
| 2391 |
+
"completion_length": 231.89956378936768,
|
| 2392 |
+
"epoch": 0.4579250272204075,
|
| 2393 |
+
"grad_norm": 0.006050299420520613,
|
| 2394 |
+
"kl": 0.02796173095703125,
|
| 2395 |
+
"learning_rate": 4.2383131204010494e-08,
|
| 2396 |
+
"loss": 0.0,
|
| 2397 |
+
"reward": 0.9988839328289032,
|
| 2398 |
+
"reward_std": 0.0031567267142236233,
|
| 2399 |
+
"rewards/equation_reward_func": 0.0,
|
| 2400 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 2401 |
+
"step": 368
|
| 2402 |
+
},
|
| 2403 |
+
{
|
| 2404 |
+
"completion_length": 233.9709930419922,
|
| 2405 |
+
"epoch": 0.4604137501944315,
|
| 2406 |
+
"grad_norm": 0.005614903622972587,
|
| 2407 |
+
"kl": 0.02776336669921875,
|
| 2408 |
+
"learning_rate": 4.039778869981064e-08,
|
| 2409 |
+
"loss": 0.0,
|
| 2410 |
+
"reward": 0.9977678656578064,
|
| 2411 |
+
"reward_std": 0.0063134534284472466,
|
| 2412 |
+
"rewards/equation_reward_func": 0.0,
|
| 2413 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 2414 |
+
"step": 370
|
| 2415 |
+
},
|
| 2416 |
+
{
|
| 2417 |
+
"completion_length": 237.66853618621826,
|
| 2418 |
+
"epoch": 0.46290247316845545,
|
| 2419 |
+
"grad_norm": 0.005600721897853672,
|
| 2420 |
+
"kl": 0.02752685546875,
|
| 2421 |
+
"learning_rate": 3.845597483600049e-08,
|
| 2422 |
+
"loss": 0.0,
|
| 2423 |
+
"reward": 0.9977678656578064,
|
| 2424 |
+
"reward_std": 0.0063134534284472466,
|
| 2425 |
+
"rewards/equation_reward_func": 0.0,
|
| 2426 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 2427 |
+
"step": 372
|
| 2428 |
+
},
|
| 2429 |
+
{
|
| 2430 |
+
"completion_length": 233.05358123779297,
|
| 2431 |
+
"epoch": 0.46539119614247937,
|
| 2432 |
+
"grad_norm": 0.0050807254989125574,
|
| 2433 |
+
"kl": 0.02738189697265625,
|
| 2434 |
+
"learning_rate": 3.655809287415284e-08,
|
| 2435 |
+
"loss": 0.0,
|
| 2436 |
+
"reward": 0.9977678656578064,
|
| 2437 |
+
"reward_std": 0.0063134534284472466,
|
| 2438 |
+
"rewards/equation_reward_func": 0.0,
|
| 2439 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 2440 |
+
"step": 374
|
| 2441 |
+
},
|
| 2442 |
+
{
|
| 2443 |
+
"completion_length": 231.516752243042,
|
| 2444 |
+
"epoch": 0.46787991911650334,
|
| 2445 |
+
"grad_norm": 0.005810047125219509,
|
| 2446 |
+
"kl": 0.0279083251953125,
|
| 2447 |
+
"learning_rate": 3.4704536952387285e-08,
|
| 2448 |
+
"loss": 0.0,
|
| 2449 |
+
"reward": 0.9988839328289032,
|
| 2450 |
+
"reward_std": 0.0031567267142236233,
|
| 2451 |
+
"rewards/equation_reward_func": 0.0,
|
| 2452 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 2453 |
+
"step": 376
|
| 2454 |
+
},
|
| 2455 |
+
{
|
| 2456 |
+
"completion_length": 238.53126049041748,
|
| 2457 |
+
"epoch": 0.4703686420905273,
|
| 2458 |
+
"grad_norm": 0.0010598227537549207,
|
| 2459 |
+
"kl": 0.028472900390625,
|
| 2460 |
+
"learning_rate": 3.2895692003518575e-08,
|
| 2461 |
+
"loss": 0.0,
|
| 2462 |
+
"reward": 1.0,
|
| 2463 |
+
"reward_std": 0.0,
|
| 2464 |
+
"rewards/equation_reward_func": 0.0,
|
| 2465 |
+
"rewards/format_reward_func": 1.0,
|
| 2466 |
+
"step": 378
|
| 2467 |
+
},
|
| 2468 |
+
{
|
| 2469 |
+
"completion_length": 233.06139469146729,
|
| 2470 |
+
"epoch": 0.4728573650645512,
|
| 2471 |
+
"grad_norm": 0.0009997280997362191,
|
| 2472 |
+
"kl": 0.02701568603515625,
|
| 2473 |
+
"learning_rate": 3.113193367511635e-08,
|
| 2474 |
+
"loss": 0.0,
|
| 2475 |
+
"reward": 0.9988839328289032,
|
| 2476 |
+
"reward_std": 0.0031567267142236233,
|
| 2477 |
+
"rewards/equation_reward_func": 0.0,
|
| 2478 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 2479 |
+
"step": 380
|
| 2480 |
+
},
|
| 2481 |
+
{
|
| 2482 |
+
"completion_length": 238.0580472946167,
|
| 2483 |
+
"epoch": 0.4753460880385752,
|
| 2484 |
+
"grad_norm": 0.004960102294910648,
|
| 2485 |
+
"kl": 0.02744293212890625,
|
| 2486 |
+
"learning_rate": 2.9413628251493934e-08,
|
| 2487 |
+
"loss": 0.0,
|
| 2488 |
+
"reward": 0.9988839328289032,
|
| 2489 |
+
"reward_std": 0.0031567267142236233,
|
| 2490 |
+
"rewards/equation_reward_func": 0.0,
|
| 2491 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 2492 |
+
"step": 382
|
| 2493 |
+
},
|
| 2494 |
+
{
|
| 2495 |
+
"completion_length": 233.74108123779297,
|
| 2496 |
+
"epoch": 0.47783481101259917,
|
| 2497 |
+
"grad_norm": 0.0010633611406857888,
|
| 2498 |
+
"kl": 0.02855682373046875,
|
| 2499 |
+
"learning_rate": 2.774113257764066e-08,
|
| 2500 |
+
"loss": 0.0,
|
| 2501 |
+
"reward": 1.0,
|
| 2502 |
+
"reward_std": 0.0,
|
| 2503 |
+
"rewards/equation_reward_func": 0.0,
|
| 2504 |
+
"rewards/format_reward_func": 1.0,
|
| 2505 |
+
"step": 384
|
| 2506 |
+
},
|
| 2507 |
+
{
|
| 2508 |
+
"completion_length": 240.4765739440918,
|
| 2509 |
+
"epoch": 0.48032353398662314,
|
| 2510 |
+
"grad_norm": 0.005811110894816115,
|
| 2511 |
+
"kl": 0.0283355712890625,
|
| 2512 |
+
"learning_rate": 2.611479398511518e-08,
|
| 2513 |
+
"loss": 0.0,
|
| 2514 |
+
"reward": 0.9977678656578064,
|
| 2515 |
+
"reward_std": 0.0063134534284472466,
|
| 2516 |
+
"rewards/equation_reward_func": 0.0,
|
| 2517 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 2518 |
+
"step": 386
|
| 2519 |
+
},
|
| 2520 |
+
{
|
| 2521 |
+
"completion_length": 233.49889278411865,
|
| 2522 |
+
"epoch": 0.48281225696064706,
|
| 2523 |
+
"grad_norm": 0.0010872044363361894,
|
| 2524 |
+
"kl": 0.02774810791015625,
|
| 2525 |
+
"learning_rate": 2.4534950219914057e-08,
|
| 2526 |
+
"loss": 0.0,
|
| 2527 |
+
"reward": 1.0,
|
| 2528 |
+
"reward_std": 0.0,
|
| 2529 |
+
"rewards/equation_reward_func": 0.0,
|
| 2530 |
+
"rewards/format_reward_func": 1.0,
|
| 2531 |
+
"step": 388
|
| 2532 |
+
},
|
| 2533 |
+
{
|
| 2534 |
+
"completion_length": 234.7712173461914,
|
| 2535 |
+
"epoch": 0.48530097993467103,
|
| 2536 |
+
"grad_norm": 0.00647383989173491,
|
| 2537 |
+
"kl": 0.02785491943359375,
|
| 2538 |
+
"learning_rate": 2.300192937233128e-08,
|
| 2539 |
+
"loss": 0.0,
|
| 2540 |
+
"reward": 0.9977678656578064,
|
| 2541 |
+
"reward_std": 0.0063134534284472466,
|
| 2542 |
+
"rewards/equation_reward_func": 0.0,
|
| 2543 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 2544 |
+
"step": 390
|
| 2545 |
+
},
|
| 2546 |
+
{
|
| 2547 |
+
"completion_length": 237.8906373977661,
|
| 2548 |
+
"epoch": 0.487789702908695,
|
| 2549 |
+
"grad_norm": 0.0060801580997462655,
|
| 2550 |
+
"kl": 0.026580810546875,
|
| 2551 |
+
"learning_rate": 2.1516049808822935e-08,
|
| 2552 |
+
"loss": 0.0,
|
| 2553 |
+
"reward": 0.9977678656578064,
|
| 2554 |
+
"reward_std": 0.0063134534284472466,
|
| 2555 |
+
"rewards/equation_reward_func": 0.0,
|
| 2556 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 2557 |
+
"step": 392
|
| 2558 |
+
},
|
| 2559 |
+
{
|
| 2560 |
+
"completion_length": 238.05916023254395,
|
| 2561 |
+
"epoch": 0.4902784258827189,
|
| 2562 |
+
"grad_norm": 0.0043540174002808405,
|
| 2563 |
+
"kl": 0.02710723876953125,
|
| 2564 |
+
"learning_rate": 2.007762010589098e-08,
|
| 2565 |
+
"loss": 0.0,
|
| 2566 |
+
"reward": 0.9977678656578064,
|
| 2567 |
+
"reward_std": 0.0063134534284472466,
|
| 2568 |
+
"rewards/equation_reward_func": 0.0,
|
| 2569 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 2570 |
+
"step": 394
|
| 2571 |
+
},
|
| 2572 |
+
{
|
| 2573 |
+
"completion_length": 236.493314743042,
|
| 2574 |
+
"epoch": 0.4927671488567429,
|
| 2575 |
+
"grad_norm": 0.0009850409330263736,
|
| 2576 |
+
"kl": 0.0270843505859375,
|
| 2577 |
+
"learning_rate": 1.8686938986000627e-08,
|
| 2578 |
+
"loss": 0.0,
|
| 2579 |
+
"reward": 1.0,
|
| 2580 |
+
"reward_std": 0.0,
|
| 2581 |
+
"rewards/equation_reward_func": 0.0,
|
| 2582 |
+
"rewards/format_reward_func": 1.0,
|
| 2583 |
+
"step": 396
|
| 2584 |
+
},
|
| 2585 |
+
{
|
| 2586 |
+
"completion_length": 238.78795623779297,
|
| 2587 |
+
"epoch": 0.49525587183076686,
|
| 2588 |
+
"grad_norm": 0.0009346164946872228,
|
| 2589 |
+
"kl": 0.0274200439453125,
|
| 2590 |
+
"learning_rate": 1.734429525554365e-08,
|
| 2591 |
+
"loss": 0.0,
|
| 2592 |
+
"reward": 1.0,
|
| 2593 |
+
"reward_std": 0.0,
|
| 2594 |
+
"rewards/equation_reward_func": 0.0,
|
| 2595 |
+
"rewards/format_reward_func": 1.0,
|
| 2596 |
+
"step": 398
|
| 2597 |
+
},
|
| 2598 |
+
{
|
| 2599 |
+
"completion_length": 242.94197273254395,
|
| 2600 |
+
"epoch": 0.4977445948047908,
|
| 2601 |
+
"grad_norm": 0.0010475326616489664,
|
| 2602 |
+
"kl": 0.0272369384765625,
|
| 2603 |
+
"learning_rate": 1.604996774486145e-08,
|
| 2604 |
+
"loss": 0.0,
|
| 2605 |
+
"reward": 0.9988839328289032,
|
| 2606 |
+
"reward_std": 0.0031567267142236233,
|
| 2607 |
+
"rewards/equation_reward_func": 0.0,
|
| 2608 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 2609 |
+
"step": 400
|
| 2610 |
+
},
|
| 2611 |
+
{
|
| 2612 |
+
"completion_length": 232.65179538726807,
|
| 2613 |
+
"epoch": 0.5002333177788147,
|
| 2614 |
+
"grad_norm": 0.0015104224923932832,
|
| 2615 |
+
"kl": 0.03054046630859375,
|
| 2616 |
+
"learning_rate": 1.4804225250339281e-08,
|
| 2617 |
+
"loss": 0.0,
|
| 2618 |
+
"reward": 0.9988839328289032,
|
| 2619 |
+
"reward_std": 0.0031567267142236233,
|
| 2620 |
+
"rewards/equation_reward_func": 0.0,
|
| 2621 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 2622 |
+
"step": 402
|
| 2623 |
+
},
|
| 2624 |
+
{
|
| 2625 |
+
"completion_length": 234.68192958831787,
|
| 2626 |
+
"epoch": 0.5027220407528387,
|
| 2627 |
+
"grad_norm": 0.0010144168231863183,
|
| 2628 |
+
"kl": 0.02809906005859375,
|
| 2629 |
+
"learning_rate": 1.360732647858498e-08,
|
| 2630 |
+
"loss": 0.0,
|
| 2631 |
+
"reward": 1.0,
|
| 2632 |
+
"reward_std": 0.0,
|
| 2633 |
+
"rewards/equation_reward_func": 0.0,
|
| 2634 |
+
"rewards/format_reward_func": 1.0,
|
| 2635 |
+
"step": 404
|
| 2636 |
+
},
|
| 2637 |
+
{
|
| 2638 |
+
"completion_length": 240.7020206451416,
|
| 2639 |
+
"epoch": 0.5052107637268627,
|
| 2640 |
+
"grad_norm": 0.011133122318443774,
|
| 2641 |
+
"kl": 0.02752685546875,
|
| 2642 |
+
"learning_rate": 1.2459519992702311e-08,
|
| 2643 |
+
"loss": 0.0,
|
| 2644 |
+
"reward": 0.9988839328289032,
|
| 2645 |
+
"reward_std": 0.0031567267142236233,
|
| 2646 |
+
"rewards/equation_reward_func": 0.0,
|
| 2647 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 2648 |
+
"step": 406
|
| 2649 |
+
},
|
| 2650 |
+
{
|
| 2651 |
+
"completion_length": 237.8292531967163,
|
| 2652 |
+
"epoch": 0.5076994867008866,
|
| 2653 |
+
"grad_norm": 0.0011848309164625682,
|
| 2654 |
+
"kl": 0.027618408203125,
|
| 2655 |
+
"learning_rate": 1.1361044160671629e-08,
|
| 2656 |
+
"loss": 0.0,
|
| 2657 |
+
"reward": 0.9988839328289032,
|
| 2658 |
+
"reward_std": 0.0031567267142236233,
|
| 2659 |
+
"rewards/equation_reward_func": 0.0,
|
| 2660 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 2661 |
+
"step": 408
|
| 2662 |
+
},
|
| 2663 |
+
{
|
| 2664 |
+
"completion_length": 233.75670528411865,
|
| 2665 |
+
"epoch": 0.5101882096749105,
|
| 2666 |
+
"grad_norm": 0.0010625408361392933,
|
| 2667 |
+
"kl": 0.0267333984375,
|
| 2668 |
+
"learning_rate": 1.0312127105846947e-08,
|
| 2669 |
+
"loss": 0.0,
|
| 2670 |
+
"reward": 0.9988839328289032,
|
| 2671 |
+
"reward_std": 0.0031567267142236233,
|
| 2672 |
+
"rewards/equation_reward_func": 0.0,
|
| 2673 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 2674 |
+
"step": 410
|
| 2675 |
+
},
|
| 2676 |
+
{
|
| 2677 |
+
"completion_length": 237.33817958831787,
|
| 2678 |
+
"epoch": 0.5126769326489345,
|
| 2679 |
+
"grad_norm": 0.007793581686460402,
|
| 2680 |
+
"kl": 0.0284576416015625,
|
| 2681 |
+
"learning_rate": 9.312986659581301e-09,
|
| 2682 |
+
"loss": 0.0,
|
| 2683 |
+
"reward": 0.9966517984867096,
|
| 2684 |
+
"reward_std": 0.00947018014267087,
|
| 2685 |
+
"rewards/equation_reward_func": 0.0,
|
| 2686 |
+
"rewards/format_reward_func": 0.9966517984867096,
|
| 2687 |
+
"step": 412
|
| 2688 |
+
},
|
| 2689 |
+
{
|
| 2690 |
+
"completion_length": 233.3125114440918,
|
| 2691 |
+
"epoch": 0.5151656556229585,
|
| 2692 |
+
"grad_norm": 0.0010496450844472917,
|
| 2693 |
+
"kl": 0.02787017822265625,
|
| 2694 |
+
"learning_rate": 8.363830315988945e-09,
|
| 2695 |
+
"loss": 0.0,
|
| 2696 |
+
"reward": 1.0,
|
| 2697 |
+
"reward_std": 0.0,
|
| 2698 |
+
"rewards/equation_reward_func": 0.0,
|
| 2699 |
+
"rewards/format_reward_func": 1.0,
|
| 2700 |
+
"step": 414
|
| 2701 |
+
},
|
| 2702 |
+
{
|
| 2703 |
+
"completion_length": 230.6584939956665,
|
| 2704 |
+
"epoch": 0.5176543785969824,
|
| 2705 |
+
"grad_norm": 0.0011012715920442739,
|
| 2706 |
+
"kl": 0.02850341796875,
|
| 2707 |
+
"learning_rate": 7.46485518885462e-09,
|
| 2708 |
+
"loss": 0.0,
|
| 2709 |
+
"reward": 1.0,
|
| 2710 |
+
"reward_std": 0.0,
|
| 2711 |
+
"rewards/equation_reward_func": 0.0,
|
| 2712 |
+
"rewards/format_reward_func": 1.0,
|
| 2713 |
+
"step": 416
|
| 2714 |
+
},
|
| 2715 |
+
{
|
| 2716 |
+
"completion_length": 234.76675128936768,
|
| 2717 |
+
"epoch": 0.5201431015710064,
|
| 2718 |
+
"grad_norm": 0.0010189132878744552,
|
| 2719 |
+
"kl": 0.0274200439453125,
|
| 2720 |
+
"learning_rate": 6.616247970698319e-09,
|
| 2721 |
+
"loss": 0.0,
|
| 2722 |
+
"reward": 0.9988839328289032,
|
| 2723 |
+
"reward_std": 0.0031567267142236233,
|
| 2724 |
+
"rewards/equation_reward_func": 0.0,
|
| 2725 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 2726 |
+
"step": 418
|
| 2727 |
+
},
|
| 2728 |
+
{
|
| 2729 |
+
"completion_length": 235.1004581451416,
|
| 2730 |
+
"epoch": 0.5226318245450303,
|
| 2731 |
+
"grad_norm": 0.0012076086244090446,
|
| 2732 |
+
"kl": 0.02881622314453125,
|
| 2733 |
+
"learning_rate": 5.8181848940044855e-09,
|
| 2734 |
+
"loss": 0.0,
|
| 2735 |
+
"reward": 0.996651791036129,
|
| 2736 |
+
"reward_std": 0.00947018014267087,
|
| 2737 |
+
"rewards/equation_reward_func": 0.0,
|
| 2738 |
+
"rewards/format_reward_func": 0.996651791036129,
|
| 2739 |
+
"step": 420
|
| 2740 |
+
},
|
| 2741 |
+
{
|
| 2742 |
+
"completion_length": 242.5714406967163,
|
| 2743 |
+
"epoch": 0.5251205475190542,
|
| 2744 |
+
"grad_norm": 0.0009532448128675256,
|
| 2745 |
+
"kl": 0.026458740234375,
|
| 2746 |
+
"learning_rate": 5.070831694623135e-09,
|
| 2747 |
+
"loss": 0.0,
|
| 2748 |
+
"reward": 1.0,
|
| 2749 |
+
"reward_std": 0.0,
|
| 2750 |
+
"rewards/equation_reward_func": 0.0,
|
| 2751 |
+
"rewards/format_reward_func": 1.0,
|
| 2752 |
+
"step": 422
|
| 2753 |
+
},
|
| 2754 |
+
{
|
| 2755 |
+
"completion_length": 235.7656364440918,
|
| 2756 |
+
"epoch": 0.5276092704930783,
|
| 2757 |
+
"grad_norm": 0.0011084492135268256,
|
| 2758 |
+
"kl": 0.02764129638671875,
|
| 2759 |
+
"learning_rate": 4.374343577351336e-09,
|
| 2760 |
+
"loss": 0.0,
|
| 2761 |
+
"reward": 1.0,
|
| 2762 |
+
"reward_std": 0.0,
|
| 2763 |
+
"rewards/equation_reward_func": 0.0,
|
| 2764 |
+
"rewards/format_reward_func": 1.0,
|
| 2765 |
+
"step": 424
|
| 2766 |
+
},
|
| 2767 |
+
{
|
| 2768 |
+
"completion_length": 236.34710884094238,
|
| 2769 |
+
"epoch": 0.5300979934671022,
|
| 2770 |
+
"grad_norm": 0.0010636945299241864,
|
| 2771 |
+
"kl": 0.02762603759765625,
|
| 2772 |
+
"learning_rate": 3.7288651837012745e-09,
|
| 2773 |
+
"loss": 0.0,
|
| 2774 |
+
"reward": 1.0,
|
| 2775 |
+
"reward_std": 0.0,
|
| 2776 |
+
"rewards/equation_reward_func": 0.0,
|
| 2777 |
+
"rewards/format_reward_func": 1.0,
|
| 2778 |
+
"step": 426
|
| 2779 |
+
},
|
| 2780 |
+
{
|
| 2781 |
+
"completion_length": 232.45090007781982,
|
| 2782 |
+
"epoch": 0.5325867164411261,
|
| 2783 |
+
"grad_norm": 0.0010014166929854047,
|
| 2784 |
+
"kl": 0.02701568603515625,
|
| 2785 |
+
"learning_rate": 3.134530561862081e-09,
|
| 2786 |
+
"loss": 0.0,
|
| 2787 |
+
"reward": 0.9988839328289032,
|
| 2788 |
+
"reward_std": 0.0031567267142236233,
|
| 2789 |
+
"rewards/equation_reward_func": 0.0,
|
| 2790 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 2791 |
+
"step": 428
|
| 2792 |
+
},
|
| 2793 |
+
{
|
| 2794 |
+
"completion_length": 229.3448781967163,
|
| 2795 |
+
"epoch": 0.5350754394151501,
|
| 2796 |
+
"grad_norm": 0.001004048594531098,
|
| 2797 |
+
"kl": 0.02599334716796875,
|
| 2798 |
+
"learning_rate": 2.5914631388619103e-09,
|
| 2799 |
+
"loss": 0.0,
|
| 2800 |
+
"reward": 1.0,
|
| 2801 |
+
"reward_std": 0.0,
|
| 2802 |
+
"rewards/equation_reward_func": 0.0,
|
| 2803 |
+
"rewards/format_reward_func": 1.0,
|
| 2804 |
+
"step": 430
|
| 2805 |
+
},
|
| 2806 |
+
{
|
| 2807 |
+
"completion_length": 234.39733123779297,
|
| 2808 |
+
"epoch": 0.537564162389174,
|
| 2809 |
+
"grad_norm": 0.001044397349569461,
|
| 2810 |
+
"kl": 0.02791595458984375,
|
| 2811 |
+
"learning_rate": 2.0997756949353297e-09,
|
| 2812 |
+
"loss": 0.0,
|
| 2813 |
+
"reward": 0.9988839328289032,
|
| 2814 |
+
"reward_std": 0.0031567267142236233,
|
| 2815 |
+
"rewards/equation_reward_func": 0.0,
|
| 2816 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 2817 |
+
"step": 432
|
| 2818 |
+
},
|
| 2819 |
+
{
|
| 2820 |
+
"completion_length": 230.9609498977661,
|
| 2821 |
+
"epoch": 0.540052885363198,
|
| 2822 |
+
"grad_norm": 0.0010577923618969485,
|
| 2823 |
+
"kl": 0.02771759033203125,
|
| 2824 |
+
"learning_rate": 1.6595703401020844e-09,
|
| 2825 |
+
"loss": 0.0,
|
| 2826 |
+
"reward": 1.0,
|
| 2827 |
+
"reward_std": 0.0,
|
| 2828 |
+
"rewards/equation_reward_func": 0.0,
|
| 2829 |
+
"rewards/format_reward_func": 1.0,
|
| 2830 |
+
"step": 434
|
| 2831 |
+
},
|
| 2832 |
+
{
|
| 2833 |
+
"completion_length": 232.87054538726807,
|
| 2834 |
+
"epoch": 0.542541608337222,
|
| 2835 |
+
"grad_norm": 0.00624001051094032,
|
| 2836 |
+
"kl": 0.03231048583984375,
|
| 2837 |
+
"learning_rate": 1.2709384929615596e-09,
|
| 2838 |
+
"loss": 0.0,
|
| 2839 |
+
"reward": 0.9988839328289032,
|
| 2840 |
+
"reward_std": 0.0031567267142236233,
|
| 2841 |
+
"rewards/equation_reward_func": 0.0,
|
| 2842 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 2843 |
+
"step": 436
|
| 2844 |
+
},
|
| 2845 |
+
{
|
| 2846 |
+
"completion_length": 230.40849113464355,
|
| 2847 |
+
"epoch": 0.5450303313112459,
|
| 2848 |
+
"grad_norm": 0.0058621450206555745,
|
| 2849 |
+
"kl": 0.02861785888671875,
|
| 2850 |
+
"learning_rate": 9.339608617077165e-10,
|
| 2851 |
+
"loss": 0.0,
|
| 2852 |
+
"reward": 0.9988839328289032,
|
| 2853 |
+
"reward_std": 0.0031567267142236233,
|
| 2854 |
+
"rewards/equation_reward_func": 0.0,
|
| 2855 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 2856 |
+
"step": 438
|
| 2857 |
+
},
|
| 2858 |
+
{
|
| 2859 |
+
"completion_length": 240.4866180419922,
|
| 2860 |
+
"epoch": 0.5475190542852699,
|
| 2861 |
+
"grad_norm": 0.0011255089792763164,
|
| 2862 |
+
"kl": 0.027740478515625,
|
| 2863 |
+
"learning_rate": 6.487074273681114e-10,
|
| 2864 |
+
"loss": 0.0,
|
| 2865 |
+
"reward": 1.0,
|
| 2866 |
+
"reward_std": 0.0,
|
| 2867 |
+
"rewards/equation_reward_func": 0.0,
|
| 2868 |
+
"rewards/format_reward_func": 1.0,
|
| 2869 |
+
"step": 440
|
| 2870 |
+
},
|
| 2871 |
+
{
|
| 2872 |
+
"completion_length": 235.4363956451416,
|
| 2873 |
+
"epoch": 0.5500077772592938,
|
| 2874 |
+
"grad_norm": 0.0011407272911082328,
|
| 2875 |
+
"kl": 0.029541015625,
|
| 2876 |
+
"learning_rate": 4.152374292708538e-10,
|
| 2877 |
+
"loss": 0.0,
|
| 2878 |
+
"reward": 0.9988839328289032,
|
| 2879 |
+
"reward_std": 0.0031567267142236233,
|
| 2880 |
+
"rewards/equation_reward_func": 0.0,
|
| 2881 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 2882 |
+
"step": 442
|
| 2883 |
+
},
|
| 2884 |
+
{
|
| 2885 |
+
"completion_length": 236.4843873977661,
|
| 2886 |
+
"epoch": 0.5524965002333178,
|
| 2887 |
+
"grad_norm": 0.006496529221920559,
|
| 2888 |
+
"kl": 0.0265655517578125,
|
| 2889 |
+
"learning_rate": 2.3359935274214204e-10,
|
| 2890 |
+
"loss": 0.0,
|
| 2891 |
+
"reward": 0.9988839328289032,
|
| 2892 |
+
"reward_std": 0.0031567267142236233,
|
| 2893 |
+
"rewards/equation_reward_func": 0.0,
|
| 2894 |
+
"rewards/format_reward_func": 0.9988839328289032,
|
| 2895 |
+
"step": 444
|
| 2896 |
+
},
|
| 2897 |
+
{
|
| 2898 |
+
"completion_length": 238.70090293884277,
|
| 2899 |
+
"epoch": 0.5549852232073418,
|
| 2900 |
+
"grad_norm": 0.005033970606058095,
|
| 2901 |
+
"kl": 0.0269317626953125,
|
| 2902 |
+
"learning_rate": 1.0383091903720665e-10,
|
| 2903 |
+
"loss": 0.0,
|
| 2904 |
+
"reward": 0.9977678656578064,
|
| 2905 |
+
"reward_std": 0.0063134534284472466,
|
| 2906 |
+
"rewards/equation_reward_func": 0.0,
|
| 2907 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 2908 |
+
"step": 446
|
| 2909 |
+
},
|
| 2910 |
+
{
|
| 2911 |
+
"completion_length": 235.43974113464355,
|
| 2912 |
+
"epoch": 0.5574739461813657,
|
| 2913 |
+
"grad_norm": 0.001040213545098924,
|
| 2914 |
+
"kl": 0.0272979736328125,
|
| 2915 |
+
"learning_rate": 2.595907750671533e-11,
|
| 2916 |
+
"loss": 0.0,
|
| 2917 |
+
"reward": 0.9977678656578064,
|
| 2918 |
+
"reward_std": 0.0063134534284472466,
|
| 2919 |
+
"rewards/equation_reward_func": 0.0,
|
| 2920 |
+
"rewards/format_reward_func": 0.9977678656578064,
|
| 2921 |
+
"step": 448
|
| 2922 |
+
},
|
| 2923 |
+
{
|
| 2924 |
+
"completion_length": 235.83706188201904,
|
| 2925 |
+
"epoch": 0.5599626691553896,
|
| 2926 |
+
"grad_norm": 0.006177204331510563,
|
| 2927 |
+
"kl": 0.02693939208984375,
|
| 2928 |
+
"learning_rate": 0.0,
|
| 2929 |
+
"loss": 0.0,
|
| 2930 |
+
"reward": 0.9966517984867096,
|
| 2931 |
+
"reward_std": 0.00947018014267087,
|
| 2932 |
+
"rewards/equation_reward_func": 0.0,
|
| 2933 |
+
"rewards/format_reward_func": 0.9966517984867096,
|
| 2934 |
+
"step": 450
|
| 2935 |
+
},
|
| 2936 |
+
{
|
| 2937 |
+
"epoch": 0.5599626691553896,
|
| 2938 |
+
"step": 450,
|
| 2939 |
+
"total_flos": 0.0,
|
| 2940 |
+
"train_loss": 2.9823833263829473e-05,
|
| 2941 |
+
"train_runtime": 18161.8968,
|
| 2942 |
+
"train_samples_per_second": 1.388,
|
| 2943 |
+
"train_steps_per_second": 0.025
|
| 2944 |
+
}
|
| 2945 |
+
],
|
| 2946 |
+
"logging_steps": 2,
|
| 2947 |
+
"max_steps": 450,
|
| 2948 |
+
"num_input_tokens_seen": 0,
|
| 2949 |
+
"num_train_epochs": 1,
|
| 2950 |
+
"save_steps": 25,
|
| 2951 |
+
"stateful_callbacks": {
|
| 2952 |
+
"TrainerControl": {
|
| 2953 |
+
"args": {
|
| 2954 |
+
"should_epoch_stop": false,
|
| 2955 |
+
"should_evaluate": false,
|
| 2956 |
+
"should_log": false,
|
| 2957 |
+
"should_save": true,
|
| 2958 |
+
"should_training_stop": true
|
| 2959 |
+
},
|
| 2960 |
+
"attributes": {}
|
| 2961 |
+
}
|
| 2962 |
+
},
|
| 2963 |
+
"total_flos": 0.0,
|
| 2964 |
+
"train_batch_size": 1,
|
| 2965 |
+
"trial_name": null,
|
| 2966 |
+
"trial_params": null
|
| 2967 |
+
}
|