Model save
Browse files- README.md +68 -0
- all_results.json +9 -0
- generation_config.json +9 -0
- train_results.json +9 -0
- trainer_state.json +1500 -0
README.md
ADDED
|
@@ -0,0 +1,68 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
base_model: AmberYifan/Llama-3.1-8B-sft-ultrachat-hhrlhf
|
| 3 |
+
library_name: transformers
|
| 4 |
+
model_name: Llama-3.1-8B-sft-hhrlhf-dpo
|
| 5 |
+
tags:
|
| 6 |
+
- generated_from_trainer
|
| 7 |
+
- trl
|
| 8 |
+
- dpo
|
| 9 |
+
licence: license
|
| 10 |
+
---
|
| 11 |
+
|
| 12 |
+
# Model Card for Llama-3.1-8B-sft-hhrlhf-dpo
|
| 13 |
+
|
| 14 |
+
This model is a fine-tuned version of [AmberYifan/Llama-3.1-8B-sft-ultrachat-hhrlhf](https://huggingface.co/AmberYifan/Llama-3.1-8B-sft-ultrachat-hhrlhf).
|
| 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="AmberYifan/Llama-3.1-8B-sft-hhrlhf-dpo", device="cuda")
|
| 24 |
+
output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
|
| 25 |
+
print(output["generated_text"])
|
| 26 |
+
```
|
| 27 |
+
|
| 28 |
+
## Training procedure
|
| 29 |
+
|
| 30 |
+
[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/yifanwang/huggingface/runs/yh37fjmz)
|
| 31 |
+
|
| 32 |
+
This model was trained with DPO, a method introduced in [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://huggingface.co/papers/2305.18290).
|
| 33 |
+
|
| 34 |
+
### Framework versions
|
| 35 |
+
|
| 36 |
+
- TRL: 0.12.2
|
| 37 |
+
- Transformers: 4.46.3
|
| 38 |
+
- Pytorch: 2.5.1+cu118
|
| 39 |
+
- Datasets: 3.2.0
|
| 40 |
+
- Tokenizers: 0.20.3
|
| 41 |
+
|
| 42 |
+
## Citations
|
| 43 |
+
|
| 44 |
+
Cite DPO as:
|
| 45 |
+
|
| 46 |
+
```bibtex
|
| 47 |
+
@inproceedings{rafailov2023direct,
|
| 48 |
+
title = {{Direct Preference Optimization: Your Language Model is Secretly a Reward Model}},
|
| 49 |
+
author = {Rafael Rafailov and Archit Sharma and Eric Mitchell and Christopher D. Manning and Stefano Ermon and Chelsea Finn},
|
| 50 |
+
year = 2023,
|
| 51 |
+
booktitle = {Advances in Neural Information Processing Systems 36: Annual Conference on Neural Information Processing Systems 2023, NeurIPS 2023, New Orleans, LA, USA, December 10 - 16, 2023},
|
| 52 |
+
url = {http://papers.nips.cc/paper_files/paper/2023/hash/a85b405ed65c6477a4fe8302b5e06ce7-Abstract-Conference.html},
|
| 53 |
+
editor = {Alice Oh and Tristan Naumann and Amir Globerson and Kate Saenko and Moritz Hardt and Sergey Levine},
|
| 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 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.1719489719113733,
|
| 5 |
+
"train_runtime": 6022.2267,
|
| 6 |
+
"train_samples": 9975,
|
| 7 |
+
"train_samples_per_second": 4.969,
|
| 8 |
+
"train_steps_per_second": 0.155
|
| 9 |
+
}
|
generation_config.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"bos_token_id": 128000,
|
| 4 |
+
"do_sample": true,
|
| 5 |
+
"eos_token_id": 128001,
|
| 6 |
+
"temperature": 0.6,
|
| 7 |
+
"top_p": 0.9,
|
| 8 |
+
"transformers_version": "4.46.3"
|
| 9 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.1719489719113733,
|
| 5 |
+
"train_runtime": 6022.2267,
|
| 6 |
+
"train_samples": 9975,
|
| 7 |
+
"train_samples_per_second": 4.969,
|
| 8 |
+
"train_steps_per_second": 0.155
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,1500 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 3.0,
|
| 5 |
+
"eval_steps": 500,
|
| 6 |
+
"global_step": 936,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.003205128205128205,
|
| 13 |
+
"grad_norm": 38.020048422719306,
|
| 14 |
+
"learning_rate": 5.3191489361702125e-09,
|
| 15 |
+
"logits/chosen": -1.484375,
|
| 16 |
+
"logits/rejected": -1.4140625,
|
| 17 |
+
"logps/chosen": -172.0,
|
| 18 |
+
"logps/rejected": -160.0,
|
| 19 |
+
"loss": 0.6914,
|
| 20 |
+
"rewards/accuracies": 0.0,
|
| 21 |
+
"rewards/chosen": 0.0,
|
| 22 |
+
"rewards/margins": 0.0,
|
| 23 |
+
"rewards/rejected": 0.0,
|
| 24 |
+
"step": 1
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.03205128205128205,
|
| 28 |
+
"grad_norm": 34.89707580881474,
|
| 29 |
+
"learning_rate": 5.3191489361702123e-08,
|
| 30 |
+
"logits/chosen": -1.484375,
|
| 31 |
+
"logits/rejected": -1.4375,
|
| 32 |
+
"logps/chosen": -140.0,
|
| 33 |
+
"logps/rejected": -181.0,
|
| 34 |
+
"loss": 0.691,
|
| 35 |
+
"rewards/accuracies": 0.25,
|
| 36 |
+
"rewards/chosen": -0.00592041015625,
|
| 37 |
+
"rewards/margins": -0.003997802734375,
|
| 38 |
+
"rewards/rejected": -0.0019073486328125,
|
| 39 |
+
"step": 10
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.0641025641025641,
|
| 43 |
+
"grad_norm": 35.22039582592668,
|
| 44 |
+
"learning_rate": 1.0638297872340425e-07,
|
| 45 |
+
"logits/chosen": -1.4375,
|
| 46 |
+
"logits/rejected": -1.40625,
|
| 47 |
+
"logps/chosen": -116.0,
|
| 48 |
+
"logps/rejected": -171.0,
|
| 49 |
+
"loss": 0.6797,
|
| 50 |
+
"rewards/accuracies": 0.48750001192092896,
|
| 51 |
+
"rewards/chosen": 0.0260009765625,
|
| 52 |
+
"rewards/margins": 0.0264892578125,
|
| 53 |
+
"rewards/rejected": -0.0004749298095703125,
|
| 54 |
+
"step": 20
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.09615384615384616,
|
| 58 |
+
"grad_norm": 40.58215068588761,
|
| 59 |
+
"learning_rate": 1.5957446808510638e-07,
|
| 60 |
+
"logits/chosen": -1.46875,
|
| 61 |
+
"logits/rejected": -1.4140625,
|
| 62 |
+
"logps/chosen": -122.5,
|
| 63 |
+
"logps/rejected": -180.0,
|
| 64 |
+
"loss": 0.6562,
|
| 65 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 66 |
+
"rewards/chosen": 0.11474609375,
|
| 67 |
+
"rewards/margins": 0.08203125,
|
| 68 |
+
"rewards/rejected": 0.03271484375,
|
| 69 |
+
"step": 30
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.1282051282051282,
|
| 73 |
+
"grad_norm": 35.55857804315714,
|
| 74 |
+
"learning_rate": 2.127659574468085e-07,
|
| 75 |
+
"logits/chosen": -1.53125,
|
| 76 |
+
"logits/rejected": -1.4609375,
|
| 77 |
+
"logps/chosen": -127.5,
|
| 78 |
+
"logps/rejected": -180.0,
|
| 79 |
+
"loss": 0.6092,
|
| 80 |
+
"rewards/accuracies": 0.8125,
|
| 81 |
+
"rewards/chosen": 0.259765625,
|
| 82 |
+
"rewards/margins": 0.21484375,
|
| 83 |
+
"rewards/rejected": 0.046142578125,
|
| 84 |
+
"step": 40
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.16025641025641027,
|
| 88 |
+
"grad_norm": 23.579694487139133,
|
| 89 |
+
"learning_rate": 2.659574468085106e-07,
|
| 90 |
+
"logits/chosen": -1.515625,
|
| 91 |
+
"logits/rejected": -1.453125,
|
| 92 |
+
"logps/chosen": -116.5,
|
| 93 |
+
"logps/rejected": -174.0,
|
| 94 |
+
"loss": 0.5579,
|
| 95 |
+
"rewards/accuracies": 0.762499988079071,
|
| 96 |
+
"rewards/chosen": 0.26953125,
|
| 97 |
+
"rewards/margins": 0.365234375,
|
| 98 |
+
"rewards/rejected": -0.095703125,
|
| 99 |
+
"step": 50
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"epoch": 0.19230769230769232,
|
| 103 |
+
"grad_norm": 23.19132387624345,
|
| 104 |
+
"learning_rate": 3.1914893617021275e-07,
|
| 105 |
+
"logits/chosen": -1.453125,
|
| 106 |
+
"logits/rejected": -1.40625,
|
| 107 |
+
"logps/chosen": -99.5,
|
| 108 |
+
"logps/rejected": -155.0,
|
| 109 |
+
"loss": 0.5053,
|
| 110 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 111 |
+
"rewards/chosen": 0.298828125,
|
| 112 |
+
"rewards/margins": 0.6015625,
|
| 113 |
+
"rewards/rejected": -0.30078125,
|
| 114 |
+
"step": 60
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.22435897435897437,
|
| 118 |
+
"grad_norm": 21.818865607569602,
|
| 119 |
+
"learning_rate": 3.7234042553191484e-07,
|
| 120 |
+
"logits/chosen": -1.5625,
|
| 121 |
+
"logits/rejected": -1.484375,
|
| 122 |
+
"logps/chosen": -100.5,
|
| 123 |
+
"logps/rejected": -162.0,
|
| 124 |
+
"loss": 0.3894,
|
| 125 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 126 |
+
"rewards/chosen": 0.3984375,
|
| 127 |
+
"rewards/margins": 1.0703125,
|
| 128 |
+
"rewards/rejected": -0.66796875,
|
| 129 |
+
"step": 70
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"epoch": 0.2564102564102564,
|
| 133 |
+
"grad_norm": 26.315987904129013,
|
| 134 |
+
"learning_rate": 4.25531914893617e-07,
|
| 135 |
+
"logits/chosen": -1.5625,
|
| 136 |
+
"logits/rejected": -1.4921875,
|
| 137 |
+
"logps/chosen": -117.5,
|
| 138 |
+
"logps/rejected": -170.0,
|
| 139 |
+
"loss": 0.3909,
|
| 140 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 141 |
+
"rewards/chosen": 0.026611328125,
|
| 142 |
+
"rewards/margins": 1.1875,
|
| 143 |
+
"rewards/rejected": -1.1640625,
|
| 144 |
+
"step": 80
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 0.28846153846153844,
|
| 148 |
+
"grad_norm": 28.51531224072289,
|
| 149 |
+
"learning_rate": 4.787234042553192e-07,
|
| 150 |
+
"logits/chosen": -1.5625,
|
| 151 |
+
"logits/rejected": -1.515625,
|
| 152 |
+
"logps/chosen": -114.5,
|
| 153 |
+
"logps/rejected": -176.0,
|
| 154 |
+
"loss": 0.387,
|
| 155 |
+
"rewards/accuracies": 0.800000011920929,
|
| 156 |
+
"rewards/chosen": -0.1240234375,
|
| 157 |
+
"rewards/margins": 1.375,
|
| 158 |
+
"rewards/rejected": -1.5,
|
| 159 |
+
"step": 90
|
| 160 |
+
},
|
| 161 |
+
{
|
| 162 |
+
"epoch": 0.32051282051282054,
|
| 163 |
+
"grad_norm": 17.39940529070616,
|
| 164 |
+
"learning_rate": 4.96437054631829e-07,
|
| 165 |
+
"logits/chosen": -1.59375,
|
| 166 |
+
"logits/rejected": -1.5234375,
|
| 167 |
+
"logps/chosen": -105.5,
|
| 168 |
+
"logps/rejected": -169.0,
|
| 169 |
+
"loss": 0.324,
|
| 170 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 171 |
+
"rewards/chosen": 0.390625,
|
| 172 |
+
"rewards/margins": 1.6640625,
|
| 173 |
+
"rewards/rejected": -1.2734375,
|
| 174 |
+
"step": 100
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"epoch": 0.3525641025641026,
|
| 178 |
+
"grad_norm": 28.443023292937838,
|
| 179 |
+
"learning_rate": 4.904988123515439e-07,
|
| 180 |
+
"logits/chosen": -1.6015625,
|
| 181 |
+
"logits/rejected": -1.5546875,
|
| 182 |
+
"logps/chosen": -118.0,
|
| 183 |
+
"logps/rejected": -199.0,
|
| 184 |
+
"loss": 0.3077,
|
| 185 |
+
"rewards/accuracies": 0.875,
|
| 186 |
+
"rewards/chosen": 0.052490234375,
|
| 187 |
+
"rewards/margins": 1.75,
|
| 188 |
+
"rewards/rejected": -1.703125,
|
| 189 |
+
"step": 110
|
| 190 |
+
},
|
| 191 |
+
{
|
| 192 |
+
"epoch": 0.38461538461538464,
|
| 193 |
+
"grad_norm": 19.86248937078093,
|
| 194 |
+
"learning_rate": 4.845605700712589e-07,
|
| 195 |
+
"logits/chosen": -1.640625,
|
| 196 |
+
"logits/rejected": -1.515625,
|
| 197 |
+
"logps/chosen": -118.0,
|
| 198 |
+
"logps/rejected": -182.0,
|
| 199 |
+
"loss": 0.3212,
|
| 200 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 201 |
+
"rewards/chosen": -0.41015625,
|
| 202 |
+
"rewards/margins": 2.0625,
|
| 203 |
+
"rewards/rejected": -2.46875,
|
| 204 |
+
"step": 120
|
| 205 |
+
},
|
| 206 |
+
{
|
| 207 |
+
"epoch": 0.4166666666666667,
|
| 208 |
+
"grad_norm": 15.34168051684897,
|
| 209 |
+
"learning_rate": 4.786223277909738e-07,
|
| 210 |
+
"logits/chosen": -1.625,
|
| 211 |
+
"logits/rejected": -1.515625,
|
| 212 |
+
"logps/chosen": -127.0,
|
| 213 |
+
"logps/rejected": -196.0,
|
| 214 |
+
"loss": 0.3436,
|
| 215 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 216 |
+
"rewards/chosen": -0.73046875,
|
| 217 |
+
"rewards/margins": 2.25,
|
| 218 |
+
"rewards/rejected": -2.984375,
|
| 219 |
+
"step": 130
|
| 220 |
+
},
|
| 221 |
+
{
|
| 222 |
+
"epoch": 0.44871794871794873,
|
| 223 |
+
"grad_norm": 31.34040320177804,
|
| 224 |
+
"learning_rate": 4.7268408551068883e-07,
|
| 225 |
+
"logits/chosen": -1.609375,
|
| 226 |
+
"logits/rejected": -1.515625,
|
| 227 |
+
"logps/chosen": -138.0,
|
| 228 |
+
"logps/rejected": -193.0,
|
| 229 |
+
"loss": 0.335,
|
| 230 |
+
"rewards/accuracies": 0.862500011920929,
|
| 231 |
+
"rewards/chosen": -0.453125,
|
| 232 |
+
"rewards/margins": 1.7578125,
|
| 233 |
+
"rewards/rejected": -2.21875,
|
| 234 |
+
"step": 140
|
| 235 |
+
},
|
| 236 |
+
{
|
| 237 |
+
"epoch": 0.4807692307692308,
|
| 238 |
+
"grad_norm": 39.09253912133659,
|
| 239 |
+
"learning_rate": 4.667458432304038e-07,
|
| 240 |
+
"logits/chosen": -1.640625,
|
| 241 |
+
"logits/rejected": -1.6328125,
|
| 242 |
+
"logps/chosen": -128.0,
|
| 243 |
+
"logps/rejected": -249.0,
|
| 244 |
+
"loss": 0.2853,
|
| 245 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 246 |
+
"rewards/chosen": -0.08056640625,
|
| 247 |
+
"rewards/margins": 2.53125,
|
| 248 |
+
"rewards/rejected": -2.609375,
|
| 249 |
+
"step": 150
|
| 250 |
+
},
|
| 251 |
+
{
|
| 252 |
+
"epoch": 0.5128205128205128,
|
| 253 |
+
"grad_norm": 17.194231471019577,
|
| 254 |
+
"learning_rate": 4.6080760095011875e-07,
|
| 255 |
+
"logits/chosen": -1.6015625,
|
| 256 |
+
"logits/rejected": -1.5859375,
|
| 257 |
+
"logps/chosen": -107.0,
|
| 258 |
+
"logps/rejected": -191.0,
|
| 259 |
+
"loss": 0.2666,
|
| 260 |
+
"rewards/accuracies": 0.925000011920929,
|
| 261 |
+
"rewards/chosen": -0.310546875,
|
| 262 |
+
"rewards/margins": 2.71875,
|
| 263 |
+
"rewards/rejected": -3.015625,
|
| 264 |
+
"step": 160
|
| 265 |
+
},
|
| 266 |
+
{
|
| 267 |
+
"epoch": 0.5448717948717948,
|
| 268 |
+
"grad_norm": 18.722209877304238,
|
| 269 |
+
"learning_rate": 4.548693586698337e-07,
|
| 270 |
+
"logits/chosen": -1.578125,
|
| 271 |
+
"logits/rejected": -1.59375,
|
| 272 |
+
"logps/chosen": -115.0,
|
| 273 |
+
"logps/rejected": -175.0,
|
| 274 |
+
"loss": 0.2696,
|
| 275 |
+
"rewards/accuracies": 0.875,
|
| 276 |
+
"rewards/chosen": -0.150390625,
|
| 277 |
+
"rewards/margins": 2.234375,
|
| 278 |
+
"rewards/rejected": -2.390625,
|
| 279 |
+
"step": 170
|
| 280 |
+
},
|
| 281 |
+
{
|
| 282 |
+
"epoch": 0.5769230769230769,
|
| 283 |
+
"grad_norm": 21.42273705026191,
|
| 284 |
+
"learning_rate": 4.4893111638954866e-07,
|
| 285 |
+
"logits/chosen": -1.5859375,
|
| 286 |
+
"logits/rejected": -1.6171875,
|
| 287 |
+
"logps/chosen": -121.5,
|
| 288 |
+
"logps/rejected": -203.0,
|
| 289 |
+
"loss": 0.2899,
|
| 290 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 291 |
+
"rewards/chosen": -0.248046875,
|
| 292 |
+
"rewards/margins": 2.390625,
|
| 293 |
+
"rewards/rejected": -2.640625,
|
| 294 |
+
"step": 180
|
| 295 |
+
},
|
| 296 |
+
{
|
| 297 |
+
"epoch": 0.6089743589743589,
|
| 298 |
+
"grad_norm": 35.105792045016955,
|
| 299 |
+
"learning_rate": 4.429928741092636e-07,
|
| 300 |
+
"logits/chosen": -1.625,
|
| 301 |
+
"logits/rejected": -1.546875,
|
| 302 |
+
"logps/chosen": -126.5,
|
| 303 |
+
"logps/rejected": -199.0,
|
| 304 |
+
"loss": 0.2834,
|
| 305 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 306 |
+
"rewards/chosen": -0.38671875,
|
| 307 |
+
"rewards/margins": 2.609375,
|
| 308 |
+
"rewards/rejected": -3.0,
|
| 309 |
+
"step": 190
|
| 310 |
+
},
|
| 311 |
+
{
|
| 312 |
+
"epoch": 0.6410256410256411,
|
| 313 |
+
"grad_norm": 25.037051509827627,
|
| 314 |
+
"learning_rate": 4.3705463182897863e-07,
|
| 315 |
+
"logits/chosen": -1.59375,
|
| 316 |
+
"logits/rejected": -1.5078125,
|
| 317 |
+
"logps/chosen": -116.5,
|
| 318 |
+
"logps/rejected": -171.0,
|
| 319 |
+
"loss": 0.3022,
|
| 320 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 321 |
+
"rewards/chosen": -0.59375,
|
| 322 |
+
"rewards/margins": 2.15625,
|
| 323 |
+
"rewards/rejected": -2.75,
|
| 324 |
+
"step": 200
|
| 325 |
+
},
|
| 326 |
+
{
|
| 327 |
+
"epoch": 0.6730769230769231,
|
| 328 |
+
"grad_norm": 25.3426239738337,
|
| 329 |
+
"learning_rate": 4.311163895486936e-07,
|
| 330 |
+
"logits/chosen": -1.5859375,
|
| 331 |
+
"logits/rejected": -1.5546875,
|
| 332 |
+
"logps/chosen": -118.5,
|
| 333 |
+
"logps/rejected": -185.0,
|
| 334 |
+
"loss": 0.3209,
|
| 335 |
+
"rewards/accuracies": 0.925000011920929,
|
| 336 |
+
"rewards/chosen": -1.21875,
|
| 337 |
+
"rewards/margins": 2.1875,
|
| 338 |
+
"rewards/rejected": -3.40625,
|
| 339 |
+
"step": 210
|
| 340 |
+
},
|
| 341 |
+
{
|
| 342 |
+
"epoch": 0.7051282051282052,
|
| 343 |
+
"grad_norm": 41.23383631897795,
|
| 344 |
+
"learning_rate": 4.251781472684085e-07,
|
| 345 |
+
"logits/chosen": -1.6484375,
|
| 346 |
+
"logits/rejected": -1.5546875,
|
| 347 |
+
"logps/chosen": -148.0,
|
| 348 |
+
"logps/rejected": -200.0,
|
| 349 |
+
"loss": 0.2921,
|
| 350 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 351 |
+
"rewards/chosen": -0.609375,
|
| 352 |
+
"rewards/margins": 2.3125,
|
| 353 |
+
"rewards/rejected": -2.921875,
|
| 354 |
+
"step": 220
|
| 355 |
+
},
|
| 356 |
+
{
|
| 357 |
+
"epoch": 0.7371794871794872,
|
| 358 |
+
"grad_norm": 22.289875990424193,
|
| 359 |
+
"learning_rate": 4.192399049881235e-07,
|
| 360 |
+
"logits/chosen": -1.609375,
|
| 361 |
+
"logits/rejected": -1.5859375,
|
| 362 |
+
"logps/chosen": -127.0,
|
| 363 |
+
"logps/rejected": -199.0,
|
| 364 |
+
"loss": 0.2805,
|
| 365 |
+
"rewards/accuracies": 0.925000011920929,
|
| 366 |
+
"rewards/chosen": -0.318359375,
|
| 367 |
+
"rewards/margins": 2.578125,
|
| 368 |
+
"rewards/rejected": -2.90625,
|
| 369 |
+
"step": 230
|
| 370 |
+
},
|
| 371 |
+
{
|
| 372 |
+
"epoch": 0.7692307692307693,
|
| 373 |
+
"grad_norm": 29.25683917677673,
|
| 374 |
+
"learning_rate": 4.1330166270783846e-07,
|
| 375 |
+
"logits/chosen": -1.59375,
|
| 376 |
+
"logits/rejected": -1.5390625,
|
| 377 |
+
"logps/chosen": -108.5,
|
| 378 |
+
"logps/rejected": -180.0,
|
| 379 |
+
"loss": 0.2541,
|
| 380 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 381 |
+
"rewards/chosen": -0.267578125,
|
| 382 |
+
"rewards/margins": 2.765625,
|
| 383 |
+
"rewards/rejected": -3.03125,
|
| 384 |
+
"step": 240
|
| 385 |
+
},
|
| 386 |
+
{
|
| 387 |
+
"epoch": 0.8012820512820513,
|
| 388 |
+
"grad_norm": 28.339641539576366,
|
| 389 |
+
"learning_rate": 4.0736342042755347e-07,
|
| 390 |
+
"logits/chosen": -1.6484375,
|
| 391 |
+
"logits/rejected": -1.5625,
|
| 392 |
+
"logps/chosen": -120.0,
|
| 393 |
+
"logps/rejected": -215.0,
|
| 394 |
+
"loss": 0.257,
|
| 395 |
+
"rewards/accuracies": 0.887499988079071,
|
| 396 |
+
"rewards/chosen": -0.50390625,
|
| 397 |
+
"rewards/margins": 3.140625,
|
| 398 |
+
"rewards/rejected": -3.640625,
|
| 399 |
+
"step": 250
|
| 400 |
+
},
|
| 401 |
+
{
|
| 402 |
+
"epoch": 0.8333333333333334,
|
| 403 |
+
"grad_norm": 28.458003688730955,
|
| 404 |
+
"learning_rate": 4.0142517814726837e-07,
|
| 405 |
+
"logits/chosen": -1.6328125,
|
| 406 |
+
"logits/rejected": -1.5703125,
|
| 407 |
+
"logps/chosen": -124.0,
|
| 408 |
+
"logps/rejected": -195.0,
|
| 409 |
+
"loss": 0.2514,
|
| 410 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 411 |
+
"rewards/chosen": -0.3203125,
|
| 412 |
+
"rewards/margins": 2.6875,
|
| 413 |
+
"rewards/rejected": -3.015625,
|
| 414 |
+
"step": 260
|
| 415 |
+
},
|
| 416 |
+
{
|
| 417 |
+
"epoch": 0.8653846153846154,
|
| 418 |
+
"grad_norm": 19.712649681094366,
|
| 419 |
+
"learning_rate": 3.9548693586698333e-07,
|
| 420 |
+
"logits/chosen": -1.640625,
|
| 421 |
+
"logits/rejected": -1.546875,
|
| 422 |
+
"logps/chosen": -121.0,
|
| 423 |
+
"logps/rejected": -177.0,
|
| 424 |
+
"loss": 0.2635,
|
| 425 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 426 |
+
"rewards/chosen": -0.11083984375,
|
| 427 |
+
"rewards/margins": 2.453125,
|
| 428 |
+
"rewards/rejected": -2.5625,
|
| 429 |
+
"step": 270
|
| 430 |
+
},
|
| 431 |
+
{
|
| 432 |
+
"epoch": 0.8974358974358975,
|
| 433 |
+
"grad_norm": 46.09415755147797,
|
| 434 |
+
"learning_rate": 3.8954869358669834e-07,
|
| 435 |
+
"logits/chosen": -1.6640625,
|
| 436 |
+
"logits/rejected": -1.546875,
|
| 437 |
+
"logps/chosen": -118.5,
|
| 438 |
+
"logps/rejected": -197.0,
|
| 439 |
+
"loss": 0.2671,
|
| 440 |
+
"rewards/accuracies": 0.875,
|
| 441 |
+
"rewards/chosen": 0.0693359375,
|
| 442 |
+
"rewards/margins": 2.78125,
|
| 443 |
+
"rewards/rejected": -2.703125,
|
| 444 |
+
"step": 280
|
| 445 |
+
},
|
| 446 |
+
{
|
| 447 |
+
"epoch": 0.9294871794871795,
|
| 448 |
+
"grad_norm": 25.186336345995475,
|
| 449 |
+
"learning_rate": 3.836104513064133e-07,
|
| 450 |
+
"logits/chosen": -1.5703125,
|
| 451 |
+
"logits/rejected": -1.5390625,
|
| 452 |
+
"logps/chosen": -129.0,
|
| 453 |
+
"logps/rejected": -203.0,
|
| 454 |
+
"loss": 0.2957,
|
| 455 |
+
"rewards/accuracies": 0.824999988079071,
|
| 456 |
+
"rewards/chosen": -1.03125,
|
| 457 |
+
"rewards/margins": 2.21875,
|
| 458 |
+
"rewards/rejected": -3.25,
|
| 459 |
+
"step": 290
|
| 460 |
+
},
|
| 461 |
+
{
|
| 462 |
+
"epoch": 0.9615384615384616,
|
| 463 |
+
"grad_norm": 25.061643380626624,
|
| 464 |
+
"learning_rate": 3.7767220902612825e-07,
|
| 465 |
+
"logits/chosen": -1.609375,
|
| 466 |
+
"logits/rejected": -1.578125,
|
| 467 |
+
"logps/chosen": -134.0,
|
| 468 |
+
"logps/rejected": -219.0,
|
| 469 |
+
"loss": 0.2345,
|
| 470 |
+
"rewards/accuracies": 0.887499988079071,
|
| 471 |
+
"rewards/chosen": -0.90234375,
|
| 472 |
+
"rewards/margins": 2.671875,
|
| 473 |
+
"rewards/rejected": -3.578125,
|
| 474 |
+
"step": 300
|
| 475 |
+
},
|
| 476 |
+
{
|
| 477 |
+
"epoch": 0.9935897435897436,
|
| 478 |
+
"grad_norm": 44.17174735441787,
|
| 479 |
+
"learning_rate": 3.717339667458432e-07,
|
| 480 |
+
"logits/chosen": -1.640625,
|
| 481 |
+
"logits/rejected": -1.5703125,
|
| 482 |
+
"logps/chosen": -119.0,
|
| 483 |
+
"logps/rejected": -192.0,
|
| 484 |
+
"loss": 0.2734,
|
| 485 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 486 |
+
"rewards/chosen": -0.2392578125,
|
| 487 |
+
"rewards/margins": 2.734375,
|
| 488 |
+
"rewards/rejected": -2.96875,
|
| 489 |
+
"step": 310
|
| 490 |
+
},
|
| 491 |
+
{
|
| 492 |
+
"epoch": 1.0,
|
| 493 |
+
"eval_logits/chosen": -1.5546875,
|
| 494 |
+
"eval_logits/rejected": -1.5234375,
|
| 495 |
+
"eval_logps/chosen": -170.0,
|
| 496 |
+
"eval_logps/rejected": -184.0,
|
| 497 |
+
"eval_loss": 0.6114453077316284,
|
| 498 |
+
"eval_rewards/accuracies": 0.5892857313156128,
|
| 499 |
+
"eval_rewards/chosen": -1.53125,
|
| 500 |
+
"eval_rewards/margins": 0.8984375,
|
| 501 |
+
"eval_rewards/rejected": -2.4375,
|
| 502 |
+
"eval_runtime": 14.9255,
|
| 503 |
+
"eval_samples_per_second": 13.4,
|
| 504 |
+
"eval_steps_per_second": 0.469,
|
| 505 |
+
"step": 312
|
| 506 |
+
},
|
| 507 |
+
{
|
| 508 |
+
"epoch": 1.0256410256410255,
|
| 509 |
+
"grad_norm": 16.449029965031666,
|
| 510 |
+
"learning_rate": 3.6579572446555817e-07,
|
| 511 |
+
"logits/chosen": -1.59375,
|
| 512 |
+
"logits/rejected": -1.578125,
|
| 513 |
+
"logps/chosen": -117.0,
|
| 514 |
+
"logps/rejected": -195.0,
|
| 515 |
+
"loss": 0.1531,
|
| 516 |
+
"rewards/accuracies": 0.925000011920929,
|
| 517 |
+
"rewards/chosen": -0.34765625,
|
| 518 |
+
"rewards/margins": 3.1875,
|
| 519 |
+
"rewards/rejected": -3.546875,
|
| 520 |
+
"step": 320
|
| 521 |
+
},
|
| 522 |
+
{
|
| 523 |
+
"epoch": 1.0576923076923077,
|
| 524 |
+
"grad_norm": 6.562166239087373,
|
| 525 |
+
"learning_rate": 3.598574821852731e-07,
|
| 526 |
+
"logits/chosen": -1.59375,
|
| 527 |
+
"logits/rejected": -1.5703125,
|
| 528 |
+
"logps/chosen": -111.0,
|
| 529 |
+
"logps/rejected": -187.0,
|
| 530 |
+
"loss": 0.1452,
|
| 531 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 532 |
+
"rewards/chosen": -0.1923828125,
|
| 533 |
+
"rewards/margins": 3.40625,
|
| 534 |
+
"rewards/rejected": -3.609375,
|
| 535 |
+
"step": 330
|
| 536 |
+
},
|
| 537 |
+
{
|
| 538 |
+
"epoch": 1.0897435897435896,
|
| 539 |
+
"grad_norm": 25.028970432454248,
|
| 540 |
+
"learning_rate": 3.5391923990498813e-07,
|
| 541 |
+
"logits/chosen": -1.6484375,
|
| 542 |
+
"logits/rejected": -1.609375,
|
| 543 |
+
"logps/chosen": -134.0,
|
| 544 |
+
"logps/rejected": -219.0,
|
| 545 |
+
"loss": 0.1345,
|
| 546 |
+
"rewards/accuracies": 0.949999988079071,
|
| 547 |
+
"rewards/chosen": -0.69921875,
|
| 548 |
+
"rewards/margins": 3.609375,
|
| 549 |
+
"rewards/rejected": -4.3125,
|
| 550 |
+
"step": 340
|
| 551 |
+
},
|
| 552 |
+
{
|
| 553 |
+
"epoch": 1.1217948717948718,
|
| 554 |
+
"grad_norm": 15.596444813277682,
|
| 555 |
+
"learning_rate": 3.479809976247031e-07,
|
| 556 |
+
"logits/chosen": -1.65625,
|
| 557 |
+
"logits/rejected": -1.6015625,
|
| 558 |
+
"logps/chosen": -130.0,
|
| 559 |
+
"logps/rejected": -207.0,
|
| 560 |
+
"loss": 0.1092,
|
| 561 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 562 |
+
"rewards/chosen": -0.00677490234375,
|
| 563 |
+
"rewards/margins": 3.828125,
|
| 564 |
+
"rewards/rejected": -3.828125,
|
| 565 |
+
"step": 350
|
| 566 |
+
},
|
| 567 |
+
{
|
| 568 |
+
"epoch": 1.1538461538461537,
|
| 569 |
+
"grad_norm": 22.37742199079234,
|
| 570 |
+
"learning_rate": 3.42042755344418e-07,
|
| 571 |
+
"logits/chosen": -1.5859375,
|
| 572 |
+
"logits/rejected": -1.5703125,
|
| 573 |
+
"logps/chosen": -136.0,
|
| 574 |
+
"logps/rejected": -242.0,
|
| 575 |
+
"loss": 0.0918,
|
| 576 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 577 |
+
"rewards/chosen": -0.640625,
|
| 578 |
+
"rewards/margins": 4.4375,
|
| 579 |
+
"rewards/rejected": -5.09375,
|
| 580 |
+
"step": 360
|
| 581 |
+
},
|
| 582 |
+
{
|
| 583 |
+
"epoch": 1.185897435897436,
|
| 584 |
+
"grad_norm": 27.25128799333949,
|
| 585 |
+
"learning_rate": 3.36104513064133e-07,
|
| 586 |
+
"logits/chosen": -1.609375,
|
| 587 |
+
"logits/rejected": -1.59375,
|
| 588 |
+
"logps/chosen": -154.0,
|
| 589 |
+
"logps/rejected": -221.0,
|
| 590 |
+
"loss": 0.1215,
|
| 591 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 592 |
+
"rewards/chosen": -0.333984375,
|
| 593 |
+
"rewards/margins": 4.46875,
|
| 594 |
+
"rewards/rejected": -4.78125,
|
| 595 |
+
"step": 370
|
| 596 |
+
},
|
| 597 |
+
{
|
| 598 |
+
"epoch": 1.217948717948718,
|
| 599 |
+
"grad_norm": 16.802650370434502,
|
| 600 |
+
"learning_rate": 3.3016627078384796e-07,
|
| 601 |
+
"logits/chosen": -1.625,
|
| 602 |
+
"logits/rejected": -1.515625,
|
| 603 |
+
"logps/chosen": -121.0,
|
| 604 |
+
"logps/rejected": -184.0,
|
| 605 |
+
"loss": 0.1051,
|
| 606 |
+
"rewards/accuracies": 0.949999988079071,
|
| 607 |
+
"rewards/chosen": -0.75390625,
|
| 608 |
+
"rewards/margins": 3.75,
|
| 609 |
+
"rewards/rejected": -4.5,
|
| 610 |
+
"step": 380
|
| 611 |
+
},
|
| 612 |
+
{
|
| 613 |
+
"epoch": 1.25,
|
| 614 |
+
"grad_norm": 18.14599606798184,
|
| 615 |
+
"learning_rate": 3.2422802850356297e-07,
|
| 616 |
+
"logits/chosen": -1.65625,
|
| 617 |
+
"logits/rejected": -1.5625,
|
| 618 |
+
"logps/chosen": -118.0,
|
| 619 |
+
"logps/rejected": -203.0,
|
| 620 |
+
"loss": 0.1691,
|
| 621 |
+
"rewards/accuracies": 0.925000011920929,
|
| 622 |
+
"rewards/chosen": -0.4609375,
|
| 623 |
+
"rewards/margins": 4.4375,
|
| 624 |
+
"rewards/rejected": -4.875,
|
| 625 |
+
"step": 390
|
| 626 |
+
},
|
| 627 |
+
{
|
| 628 |
+
"epoch": 1.282051282051282,
|
| 629 |
+
"grad_norm": 14.341477798166297,
|
| 630 |
+
"learning_rate": 3.182897862232779e-07,
|
| 631 |
+
"logits/chosen": -1.609375,
|
| 632 |
+
"logits/rejected": -1.5546875,
|
| 633 |
+
"logps/chosen": -112.0,
|
| 634 |
+
"logps/rejected": -202.0,
|
| 635 |
+
"loss": 0.1084,
|
| 636 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 637 |
+
"rewards/chosen": -0.2578125,
|
| 638 |
+
"rewards/margins": 4.65625,
|
| 639 |
+
"rewards/rejected": -4.90625,
|
| 640 |
+
"step": 400
|
| 641 |
+
},
|
| 642 |
+
{
|
| 643 |
+
"epoch": 1.314102564102564,
|
| 644 |
+
"grad_norm": 22.538352683910524,
|
| 645 |
+
"learning_rate": 3.1235154394299283e-07,
|
| 646 |
+
"logits/chosen": -1.671875,
|
| 647 |
+
"logits/rejected": -1.625,
|
| 648 |
+
"logps/chosen": -168.0,
|
| 649 |
+
"logps/rejected": -270.0,
|
| 650 |
+
"loss": 0.136,
|
| 651 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 652 |
+
"rewards/chosen": -1.1796875,
|
| 653 |
+
"rewards/margins": 4.71875,
|
| 654 |
+
"rewards/rejected": -5.90625,
|
| 655 |
+
"step": 410
|
| 656 |
+
},
|
| 657 |
+
{
|
| 658 |
+
"epoch": 1.3461538461538463,
|
| 659 |
+
"grad_norm": 7.69063972210861,
|
| 660 |
+
"learning_rate": 3.0641330166270784e-07,
|
| 661 |
+
"logits/chosen": -1.609375,
|
| 662 |
+
"logits/rejected": -1.59375,
|
| 663 |
+
"logps/chosen": -128.0,
|
| 664 |
+
"logps/rejected": -232.0,
|
| 665 |
+
"loss": 0.1123,
|
| 666 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 667 |
+
"rewards/chosen": -1.0859375,
|
| 668 |
+
"rewards/margins": 4.59375,
|
| 669 |
+
"rewards/rejected": -5.6875,
|
| 670 |
+
"step": 420
|
| 671 |
+
},
|
| 672 |
+
{
|
| 673 |
+
"epoch": 1.3782051282051282,
|
| 674 |
+
"grad_norm": 22.81215314331503,
|
| 675 |
+
"learning_rate": 3.004750593824228e-07,
|
| 676 |
+
"logits/chosen": -1.6328125,
|
| 677 |
+
"logits/rejected": -1.5390625,
|
| 678 |
+
"logps/chosen": -122.5,
|
| 679 |
+
"logps/rejected": -209.0,
|
| 680 |
+
"loss": 0.0971,
|
| 681 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 682 |
+
"rewards/chosen": -0.33984375,
|
| 683 |
+
"rewards/margins": 4.65625,
|
| 684 |
+
"rewards/rejected": -5.0,
|
| 685 |
+
"step": 430
|
| 686 |
+
},
|
| 687 |
+
{
|
| 688 |
+
"epoch": 1.4102564102564101,
|
| 689 |
+
"grad_norm": 16.88060293661355,
|
| 690 |
+
"learning_rate": 2.9453681710213776e-07,
|
| 691 |
+
"logits/chosen": -1.5859375,
|
| 692 |
+
"logits/rejected": -1.4765625,
|
| 693 |
+
"logps/chosen": -123.0,
|
| 694 |
+
"logps/rejected": -205.0,
|
| 695 |
+
"loss": 0.11,
|
| 696 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 697 |
+
"rewards/chosen": -0.4296875,
|
| 698 |
+
"rewards/margins": 4.5625,
|
| 699 |
+
"rewards/rejected": -5.0,
|
| 700 |
+
"step": 440
|
| 701 |
+
},
|
| 702 |
+
{
|
| 703 |
+
"epoch": 1.4423076923076923,
|
| 704 |
+
"grad_norm": 14.494261083829612,
|
| 705 |
+
"learning_rate": 2.885985748218527e-07,
|
| 706 |
+
"logits/chosen": -1.546875,
|
| 707 |
+
"logits/rejected": -1.5546875,
|
| 708 |
+
"logps/chosen": -118.5,
|
| 709 |
+
"logps/rejected": -221.0,
|
| 710 |
+
"loss": 0.0896,
|
| 711 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 712 |
+
"rewards/chosen": -1.0859375,
|
| 713 |
+
"rewards/margins": 4.59375,
|
| 714 |
+
"rewards/rejected": -5.6875,
|
| 715 |
+
"step": 450
|
| 716 |
+
},
|
| 717 |
+
{
|
| 718 |
+
"epoch": 1.4743589743589745,
|
| 719 |
+
"grad_norm": 16.655718879118293,
|
| 720 |
+
"learning_rate": 2.8266033254156767e-07,
|
| 721 |
+
"logits/chosen": -1.59375,
|
| 722 |
+
"logits/rejected": -1.5859375,
|
| 723 |
+
"logps/chosen": -142.0,
|
| 724 |
+
"logps/rejected": -256.0,
|
| 725 |
+
"loss": 0.1071,
|
| 726 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 727 |
+
"rewards/chosen": -1.25,
|
| 728 |
+
"rewards/margins": 4.84375,
|
| 729 |
+
"rewards/rejected": -6.09375,
|
| 730 |
+
"step": 460
|
| 731 |
+
},
|
| 732 |
+
{
|
| 733 |
+
"epoch": 1.5064102564102564,
|
| 734 |
+
"grad_norm": 20.51874312804957,
|
| 735 |
+
"learning_rate": 2.7672209026128263e-07,
|
| 736 |
+
"logits/chosen": -1.5625,
|
| 737 |
+
"logits/rejected": -1.53125,
|
| 738 |
+
"logps/chosen": -132.0,
|
| 739 |
+
"logps/rejected": -231.0,
|
| 740 |
+
"loss": 0.1012,
|
| 741 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 742 |
+
"rewards/chosen": -1.015625,
|
| 743 |
+
"rewards/margins": 4.65625,
|
| 744 |
+
"rewards/rejected": -5.6875,
|
| 745 |
+
"step": 470
|
| 746 |
+
},
|
| 747 |
+
{
|
| 748 |
+
"epoch": 1.5384615384615383,
|
| 749 |
+
"grad_norm": 28.03019085733964,
|
| 750 |
+
"learning_rate": 2.7078384798099764e-07,
|
| 751 |
+
"logits/chosen": -1.4453125,
|
| 752 |
+
"logits/rejected": -1.4609375,
|
| 753 |
+
"logps/chosen": -104.5,
|
| 754 |
+
"logps/rejected": -203.0,
|
| 755 |
+
"loss": 0.1074,
|
| 756 |
+
"rewards/accuracies": 0.987500011920929,
|
| 757 |
+
"rewards/chosen": -1.6796875,
|
| 758 |
+
"rewards/margins": 4.59375,
|
| 759 |
+
"rewards/rejected": -6.3125,
|
| 760 |
+
"step": 480
|
| 761 |
+
},
|
| 762 |
+
{
|
| 763 |
+
"epoch": 1.5705128205128205,
|
| 764 |
+
"grad_norm": 14.674518317046845,
|
| 765 |
+
"learning_rate": 2.648456057007126e-07,
|
| 766 |
+
"logits/chosen": -1.5625,
|
| 767 |
+
"logits/rejected": -1.5390625,
|
| 768 |
+
"logps/chosen": -139.0,
|
| 769 |
+
"logps/rejected": -233.0,
|
| 770 |
+
"loss": 0.1003,
|
| 771 |
+
"rewards/accuracies": 0.987500011920929,
|
| 772 |
+
"rewards/chosen": -1.3125,
|
| 773 |
+
"rewards/margins": 4.9375,
|
| 774 |
+
"rewards/rejected": -6.28125,
|
| 775 |
+
"step": 490
|
| 776 |
+
},
|
| 777 |
+
{
|
| 778 |
+
"epoch": 1.6025641025641026,
|
| 779 |
+
"grad_norm": 18.731133870495857,
|
| 780 |
+
"learning_rate": 2.589073634204275e-07,
|
| 781 |
+
"logits/chosen": -1.5703125,
|
| 782 |
+
"logits/rejected": -1.5625,
|
| 783 |
+
"logps/chosen": -144.0,
|
| 784 |
+
"logps/rejected": -251.0,
|
| 785 |
+
"loss": 0.1228,
|
| 786 |
+
"rewards/accuracies": 0.949999988079071,
|
| 787 |
+
"rewards/chosen": -1.109375,
|
| 788 |
+
"rewards/margins": 5.25,
|
| 789 |
+
"rewards/rejected": -6.375,
|
| 790 |
+
"step": 500
|
| 791 |
+
},
|
| 792 |
+
{
|
| 793 |
+
"epoch": 1.6346153846153846,
|
| 794 |
+
"grad_norm": 16.81631331394118,
|
| 795 |
+
"learning_rate": 2.529691211401425e-07,
|
| 796 |
+
"logits/chosen": -1.5625,
|
| 797 |
+
"logits/rejected": -1.5,
|
| 798 |
+
"logps/chosen": -131.0,
|
| 799 |
+
"logps/rejected": -230.0,
|
| 800 |
+
"loss": 0.0819,
|
| 801 |
+
"rewards/accuracies": 0.987500011920929,
|
| 802 |
+
"rewards/chosen": -0.8359375,
|
| 803 |
+
"rewards/margins": 5.375,
|
| 804 |
+
"rewards/rejected": -6.21875,
|
| 805 |
+
"step": 510
|
| 806 |
+
},
|
| 807 |
+
{
|
| 808 |
+
"epoch": 1.6666666666666665,
|
| 809 |
+
"grad_norm": 15.111001669085608,
|
| 810 |
+
"learning_rate": 2.4703087885985747e-07,
|
| 811 |
+
"logits/chosen": -1.546875,
|
| 812 |
+
"logits/rejected": -1.4921875,
|
| 813 |
+
"logps/chosen": -114.0,
|
| 814 |
+
"logps/rejected": -202.0,
|
| 815 |
+
"loss": 0.1091,
|
| 816 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 817 |
+
"rewards/chosen": -0.87109375,
|
| 818 |
+
"rewards/margins": 4.78125,
|
| 819 |
+
"rewards/rejected": -5.625,
|
| 820 |
+
"step": 520
|
| 821 |
+
},
|
| 822 |
+
{
|
| 823 |
+
"epoch": 1.6987179487179487,
|
| 824 |
+
"grad_norm": 26.75975769912583,
|
| 825 |
+
"learning_rate": 2.410926365795724e-07,
|
| 826 |
+
"logits/chosen": -1.5625,
|
| 827 |
+
"logits/rejected": -1.515625,
|
| 828 |
+
"logps/chosen": -140.0,
|
| 829 |
+
"logps/rejected": -221.0,
|
| 830 |
+
"loss": 0.1139,
|
| 831 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 832 |
+
"rewards/chosen": -1.0625,
|
| 833 |
+
"rewards/margins": 4.8125,
|
| 834 |
+
"rewards/rejected": -5.875,
|
| 835 |
+
"step": 530
|
| 836 |
+
},
|
| 837 |
+
{
|
| 838 |
+
"epoch": 1.7307692307692308,
|
| 839 |
+
"grad_norm": 13.759559493772942,
|
| 840 |
+
"learning_rate": 2.351543942992874e-07,
|
| 841 |
+
"logits/chosen": -1.578125,
|
| 842 |
+
"logits/rejected": -1.5,
|
| 843 |
+
"logps/chosen": -143.0,
|
| 844 |
+
"logps/rejected": -229.0,
|
| 845 |
+
"loss": 0.0978,
|
| 846 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 847 |
+
"rewards/chosen": -0.9765625,
|
| 848 |
+
"rewards/margins": 5.09375,
|
| 849 |
+
"rewards/rejected": -6.0625,
|
| 850 |
+
"step": 540
|
| 851 |
+
},
|
| 852 |
+
{
|
| 853 |
+
"epoch": 1.7628205128205128,
|
| 854 |
+
"grad_norm": 18.98900148314797,
|
| 855 |
+
"learning_rate": 2.2921615201900234e-07,
|
| 856 |
+
"logits/chosen": -1.4921875,
|
| 857 |
+
"logits/rejected": -1.5,
|
| 858 |
+
"logps/chosen": -114.0,
|
| 859 |
+
"logps/rejected": -232.0,
|
| 860 |
+
"loss": 0.1103,
|
| 861 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 862 |
+
"rewards/chosen": -0.6953125,
|
| 863 |
+
"rewards/margins": 5.03125,
|
| 864 |
+
"rewards/rejected": -5.71875,
|
| 865 |
+
"step": 550
|
| 866 |
+
},
|
| 867 |
+
{
|
| 868 |
+
"epoch": 1.7948717948717947,
|
| 869 |
+
"grad_norm": 7.478935072745169,
|
| 870 |
+
"learning_rate": 2.2327790973871732e-07,
|
| 871 |
+
"logits/chosen": -1.515625,
|
| 872 |
+
"logits/rejected": -1.4375,
|
| 873 |
+
"logps/chosen": -112.0,
|
| 874 |
+
"logps/rejected": -177.0,
|
| 875 |
+
"loss": 0.1139,
|
| 876 |
+
"rewards/accuracies": 0.9375,
|
| 877 |
+
"rewards/chosen": -0.6640625,
|
| 878 |
+
"rewards/margins": 4.5625,
|
| 879 |
+
"rewards/rejected": -5.21875,
|
| 880 |
+
"step": 560
|
| 881 |
+
},
|
| 882 |
+
{
|
| 883 |
+
"epoch": 1.8269230769230769,
|
| 884 |
+
"grad_norm": 13.728806584543696,
|
| 885 |
+
"learning_rate": 2.173396674584323e-07,
|
| 886 |
+
"logits/chosen": -1.546875,
|
| 887 |
+
"logits/rejected": -1.5546875,
|
| 888 |
+
"logps/chosen": -130.0,
|
| 889 |
+
"logps/rejected": -239.0,
|
| 890 |
+
"loss": 0.1355,
|
| 891 |
+
"rewards/accuracies": 0.949999988079071,
|
| 892 |
+
"rewards/chosen": -0.466796875,
|
| 893 |
+
"rewards/margins": 4.875,
|
| 894 |
+
"rewards/rejected": -5.34375,
|
| 895 |
+
"step": 570
|
| 896 |
+
},
|
| 897 |
+
{
|
| 898 |
+
"epoch": 1.858974358974359,
|
| 899 |
+
"grad_norm": 9.926742756220793,
|
| 900 |
+
"learning_rate": 2.1140142517814726e-07,
|
| 901 |
+
"logits/chosen": -1.609375,
|
| 902 |
+
"logits/rejected": -1.5078125,
|
| 903 |
+
"logps/chosen": -158.0,
|
| 904 |
+
"logps/rejected": -202.0,
|
| 905 |
+
"loss": 0.1118,
|
| 906 |
+
"rewards/accuracies": 0.949999988079071,
|
| 907 |
+
"rewards/chosen": -1.09375,
|
| 908 |
+
"rewards/margins": 4.21875,
|
| 909 |
+
"rewards/rejected": -5.3125,
|
| 910 |
+
"step": 580
|
| 911 |
+
},
|
| 912 |
+
{
|
| 913 |
+
"epoch": 1.891025641025641,
|
| 914 |
+
"grad_norm": 12.431864869622972,
|
| 915 |
+
"learning_rate": 2.0546318289786222e-07,
|
| 916 |
+
"logits/chosen": -1.5546875,
|
| 917 |
+
"logits/rejected": -1.53125,
|
| 918 |
+
"logps/chosen": -122.0,
|
| 919 |
+
"logps/rejected": -202.0,
|
| 920 |
+
"loss": 0.1325,
|
| 921 |
+
"rewards/accuracies": 0.987500011920929,
|
| 922 |
+
"rewards/chosen": -0.61328125,
|
| 923 |
+
"rewards/margins": 4.65625,
|
| 924 |
+
"rewards/rejected": -5.28125,
|
| 925 |
+
"step": 590
|
| 926 |
+
},
|
| 927 |
+
{
|
| 928 |
+
"epoch": 1.9230769230769231,
|
| 929 |
+
"grad_norm": 12.291656293105886,
|
| 930 |
+
"learning_rate": 1.9952494061757718e-07,
|
| 931 |
+
"logits/chosen": -1.5703125,
|
| 932 |
+
"logits/rejected": -1.5625,
|
| 933 |
+
"logps/chosen": -123.0,
|
| 934 |
+
"logps/rejected": -227.0,
|
| 935 |
+
"loss": 0.1085,
|
| 936 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 937 |
+
"rewards/chosen": -0.67578125,
|
| 938 |
+
"rewards/margins": 4.8125,
|
| 939 |
+
"rewards/rejected": -5.5,
|
| 940 |
+
"step": 600
|
| 941 |
+
},
|
| 942 |
+
{
|
| 943 |
+
"epoch": 1.9551282051282053,
|
| 944 |
+
"grad_norm": 6.534649119157792,
|
| 945 |
+
"learning_rate": 1.9358669833729216e-07,
|
| 946 |
+
"logits/chosen": -1.5546875,
|
| 947 |
+
"logits/rejected": -1.484375,
|
| 948 |
+
"logps/chosen": -122.0,
|
| 949 |
+
"logps/rejected": -202.0,
|
| 950 |
+
"loss": 0.0935,
|
| 951 |
+
"rewards/accuracies": 0.949999988079071,
|
| 952 |
+
"rewards/chosen": -0.56640625,
|
| 953 |
+
"rewards/margins": 5.0,
|
| 954 |
+
"rewards/rejected": -5.59375,
|
| 955 |
+
"step": 610
|
| 956 |
+
},
|
| 957 |
+
{
|
| 958 |
+
"epoch": 1.9871794871794872,
|
| 959 |
+
"grad_norm": 10.91488724265466,
|
| 960 |
+
"learning_rate": 1.876484560570071e-07,
|
| 961 |
+
"logits/chosen": -1.5703125,
|
| 962 |
+
"logits/rejected": -1.515625,
|
| 963 |
+
"logps/chosen": -139.0,
|
| 964 |
+
"logps/rejected": -236.0,
|
| 965 |
+
"loss": 0.1071,
|
| 966 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 967 |
+
"rewards/chosen": -1.1953125,
|
| 968 |
+
"rewards/margins": 5.34375,
|
| 969 |
+
"rewards/rejected": -6.53125,
|
| 970 |
+
"step": 620
|
| 971 |
+
},
|
| 972 |
+
{
|
| 973 |
+
"epoch": 2.0,
|
| 974 |
+
"eval_logits/chosen": -1.5,
|
| 975 |
+
"eval_logits/rejected": -1.46875,
|
| 976 |
+
"eval_logps/chosen": -188.0,
|
| 977 |
+
"eval_logps/rejected": -206.0,
|
| 978 |
+
"eval_loss": 0.700976550579071,
|
| 979 |
+
"eval_rewards/accuracies": 0.6428571343421936,
|
| 980 |
+
"eval_rewards/chosen": -3.296875,
|
| 981 |
+
"eval_rewards/margins": 1.3046875,
|
| 982 |
+
"eval_rewards/rejected": -4.59375,
|
| 983 |
+
"eval_runtime": 15.0135,
|
| 984 |
+
"eval_samples_per_second": 13.321,
|
| 985 |
+
"eval_steps_per_second": 0.466,
|
| 986 |
+
"step": 624
|
| 987 |
+
},
|
| 988 |
+
{
|
| 989 |
+
"epoch": 2.019230769230769,
|
| 990 |
+
"grad_norm": 2.9708656343687063,
|
| 991 |
+
"learning_rate": 1.8171021377672207e-07,
|
| 992 |
+
"logits/chosen": -1.515625,
|
| 993 |
+
"logits/rejected": -1.5,
|
| 994 |
+
"logps/chosen": -150.0,
|
| 995 |
+
"logps/rejected": -224.0,
|
| 996 |
+
"loss": 0.071,
|
| 997 |
+
"rewards/accuracies": 0.949999988079071,
|
| 998 |
+
"rewards/chosen": -0.98828125,
|
| 999 |
+
"rewards/margins": 5.40625,
|
| 1000 |
+
"rewards/rejected": -6.40625,
|
| 1001 |
+
"step": 630
|
| 1002 |
+
},
|
| 1003 |
+
{
|
| 1004 |
+
"epoch": 2.051282051282051,
|
| 1005 |
+
"grad_norm": 3.1451139864714266,
|
| 1006 |
+
"learning_rate": 1.7577197149643706e-07,
|
| 1007 |
+
"logits/chosen": -1.5078125,
|
| 1008 |
+
"logits/rejected": -1.4609375,
|
| 1009 |
+
"logps/chosen": -121.5,
|
| 1010 |
+
"logps/rejected": -198.0,
|
| 1011 |
+
"loss": 0.0534,
|
| 1012 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1013 |
+
"rewards/chosen": -0.8203125,
|
| 1014 |
+
"rewards/margins": 4.9375,
|
| 1015 |
+
"rewards/rejected": -5.75,
|
| 1016 |
+
"step": 640
|
| 1017 |
+
},
|
| 1018 |
+
{
|
| 1019 |
+
"epoch": 2.0833333333333335,
|
| 1020 |
+
"grad_norm": 7.314670214652841,
|
| 1021 |
+
"learning_rate": 1.6983372921615202e-07,
|
| 1022 |
+
"logits/chosen": -1.5703125,
|
| 1023 |
+
"logits/rejected": -1.546875,
|
| 1024 |
+
"logps/chosen": -134.0,
|
| 1025 |
+
"logps/rejected": -247.0,
|
| 1026 |
+
"loss": 0.0546,
|
| 1027 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1028 |
+
"rewards/chosen": -0.73046875,
|
| 1029 |
+
"rewards/margins": 5.84375,
|
| 1030 |
+
"rewards/rejected": -6.5625,
|
| 1031 |
+
"step": 650
|
| 1032 |
+
},
|
| 1033 |
+
{
|
| 1034 |
+
"epoch": 2.1153846153846154,
|
| 1035 |
+
"grad_norm": 6.986967407965111,
|
| 1036 |
+
"learning_rate": 1.6389548693586697e-07,
|
| 1037 |
+
"logits/chosen": -1.5625,
|
| 1038 |
+
"logits/rejected": -1.53125,
|
| 1039 |
+
"logps/chosen": -123.5,
|
| 1040 |
+
"logps/rejected": -219.0,
|
| 1041 |
+
"loss": 0.0485,
|
| 1042 |
+
"rewards/accuracies": 1.0,
|
| 1043 |
+
"rewards/chosen": -0.9375,
|
| 1044 |
+
"rewards/margins": 5.5625,
|
| 1045 |
+
"rewards/rejected": -6.53125,
|
| 1046 |
+
"step": 660
|
| 1047 |
+
},
|
| 1048 |
+
{
|
| 1049 |
+
"epoch": 2.1474358974358974,
|
| 1050 |
+
"grad_norm": 3.152252590941172,
|
| 1051 |
+
"learning_rate": 1.5795724465558193e-07,
|
| 1052 |
+
"logits/chosen": -1.515625,
|
| 1053 |
+
"logits/rejected": -1.53125,
|
| 1054 |
+
"logps/chosen": -121.5,
|
| 1055 |
+
"logps/rejected": -264.0,
|
| 1056 |
+
"loss": 0.0391,
|
| 1057 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1058 |
+
"rewards/chosen": -0.97265625,
|
| 1059 |
+
"rewards/margins": 6.40625,
|
| 1060 |
+
"rewards/rejected": -7.375,
|
| 1061 |
+
"step": 670
|
| 1062 |
+
},
|
| 1063 |
+
{
|
| 1064 |
+
"epoch": 2.1794871794871793,
|
| 1065 |
+
"grad_norm": 5.616933414198294,
|
| 1066 |
+
"learning_rate": 1.520190023752969e-07,
|
| 1067 |
+
"logits/chosen": -1.5390625,
|
| 1068 |
+
"logits/rejected": -1.5078125,
|
| 1069 |
+
"logps/chosen": -136.0,
|
| 1070 |
+
"logps/rejected": -225.0,
|
| 1071 |
+
"loss": 0.0523,
|
| 1072 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1073 |
+
"rewards/chosen": -1.265625,
|
| 1074 |
+
"rewards/margins": 5.4375,
|
| 1075 |
+
"rewards/rejected": -6.6875,
|
| 1076 |
+
"step": 680
|
| 1077 |
+
},
|
| 1078 |
+
{
|
| 1079 |
+
"epoch": 2.2115384615384617,
|
| 1080 |
+
"grad_norm": 4.462570602831943,
|
| 1081 |
+
"learning_rate": 1.4608076009501184e-07,
|
| 1082 |
+
"logits/chosen": -1.5390625,
|
| 1083 |
+
"logits/rejected": -1.53125,
|
| 1084 |
+
"logps/chosen": -141.0,
|
| 1085 |
+
"logps/rejected": -247.0,
|
| 1086 |
+
"loss": 0.0548,
|
| 1087 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1088 |
+
"rewards/chosen": -0.9921875,
|
| 1089 |
+
"rewards/margins": 5.90625,
|
| 1090 |
+
"rewards/rejected": -6.90625,
|
| 1091 |
+
"step": 690
|
| 1092 |
+
},
|
| 1093 |
+
{
|
| 1094 |
+
"epoch": 2.2435897435897436,
|
| 1095 |
+
"grad_norm": 6.786342065093859,
|
| 1096 |
+
"learning_rate": 1.4014251781472683e-07,
|
| 1097 |
+
"logits/chosen": -1.5546875,
|
| 1098 |
+
"logits/rejected": -1.546875,
|
| 1099 |
+
"logps/chosen": -115.0,
|
| 1100 |
+
"logps/rejected": -243.0,
|
| 1101 |
+
"loss": 0.0328,
|
| 1102 |
+
"rewards/accuracies": 1.0,
|
| 1103 |
+
"rewards/chosen": -0.7890625,
|
| 1104 |
+
"rewards/margins": 6.25,
|
| 1105 |
+
"rewards/rejected": -7.03125,
|
| 1106 |
+
"step": 700
|
| 1107 |
+
},
|
| 1108 |
+
{
|
| 1109 |
+
"epoch": 2.2756410256410255,
|
| 1110 |
+
"grad_norm": 10.783463457904992,
|
| 1111 |
+
"learning_rate": 1.342042755344418e-07,
|
| 1112 |
+
"logits/chosen": -1.5390625,
|
| 1113 |
+
"logits/rejected": -1.546875,
|
| 1114 |
+
"logps/chosen": -149.0,
|
| 1115 |
+
"logps/rejected": -266.0,
|
| 1116 |
+
"loss": 0.0556,
|
| 1117 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1118 |
+
"rewards/chosen": -1.203125,
|
| 1119 |
+
"rewards/margins": 6.125,
|
| 1120 |
+
"rewards/rejected": -7.3125,
|
| 1121 |
+
"step": 710
|
| 1122 |
+
},
|
| 1123 |
+
{
|
| 1124 |
+
"epoch": 2.3076923076923075,
|
| 1125 |
+
"grad_norm": 9.0679334215508,
|
| 1126 |
+
"learning_rate": 1.2826603325415677e-07,
|
| 1127 |
+
"logits/chosen": -1.53125,
|
| 1128 |
+
"logits/rejected": -1.4765625,
|
| 1129 |
+
"logps/chosen": -121.0,
|
| 1130 |
+
"logps/rejected": -217.0,
|
| 1131 |
+
"loss": 0.0438,
|
| 1132 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1133 |
+
"rewards/chosen": -0.80078125,
|
| 1134 |
+
"rewards/margins": 5.84375,
|
| 1135 |
+
"rewards/rejected": -6.65625,
|
| 1136 |
+
"step": 720
|
| 1137 |
+
},
|
| 1138 |
+
{
|
| 1139 |
+
"epoch": 2.33974358974359,
|
| 1140 |
+
"grad_norm": 2.164285473082669,
|
| 1141 |
+
"learning_rate": 1.2232779097387173e-07,
|
| 1142 |
+
"logits/chosen": -1.5078125,
|
| 1143 |
+
"logits/rejected": -1.5,
|
| 1144 |
+
"logps/chosen": -153.0,
|
| 1145 |
+
"logps/rejected": -237.0,
|
| 1146 |
+
"loss": 0.0414,
|
| 1147 |
+
"rewards/accuracies": 1.0,
|
| 1148 |
+
"rewards/chosen": -1.2421875,
|
| 1149 |
+
"rewards/margins": 6.15625,
|
| 1150 |
+
"rewards/rejected": -7.375,
|
| 1151 |
+
"step": 730
|
| 1152 |
+
},
|
| 1153 |
+
{
|
| 1154 |
+
"epoch": 2.371794871794872,
|
| 1155 |
+
"grad_norm": 8.601824238964523,
|
| 1156 |
+
"learning_rate": 1.163895486935867e-07,
|
| 1157 |
+
"logits/chosen": -1.4921875,
|
| 1158 |
+
"logits/rejected": -1.46875,
|
| 1159 |
+
"logps/chosen": -148.0,
|
| 1160 |
+
"logps/rejected": -238.0,
|
| 1161 |
+
"loss": 0.0358,
|
| 1162 |
+
"rewards/accuracies": 1.0,
|
| 1163 |
+
"rewards/chosen": -1.6953125,
|
| 1164 |
+
"rewards/margins": 5.9375,
|
| 1165 |
+
"rewards/rejected": -7.625,
|
| 1166 |
+
"step": 740
|
| 1167 |
+
},
|
| 1168 |
+
{
|
| 1169 |
+
"epoch": 2.4038461538461537,
|
| 1170 |
+
"grad_norm": 11.674320163023708,
|
| 1171 |
+
"learning_rate": 1.1045130641330165e-07,
|
| 1172 |
+
"logits/chosen": -1.5234375,
|
| 1173 |
+
"logits/rejected": -1.5390625,
|
| 1174 |
+
"logps/chosen": -133.0,
|
| 1175 |
+
"logps/rejected": -258.0,
|
| 1176 |
+
"loss": 0.0523,
|
| 1177 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1178 |
+
"rewards/chosen": -1.9453125,
|
| 1179 |
+
"rewards/margins": 6.4375,
|
| 1180 |
+
"rewards/rejected": -8.375,
|
| 1181 |
+
"step": 750
|
| 1182 |
+
},
|
| 1183 |
+
{
|
| 1184 |
+
"epoch": 2.435897435897436,
|
| 1185 |
+
"grad_norm": 4.6040374115250415,
|
| 1186 |
+
"learning_rate": 1.0451306413301662e-07,
|
| 1187 |
+
"logits/chosen": -1.515625,
|
| 1188 |
+
"logits/rejected": -1.5234375,
|
| 1189 |
+
"logps/chosen": -115.5,
|
| 1190 |
+
"logps/rejected": -258.0,
|
| 1191 |
+
"loss": 0.0425,
|
| 1192 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1193 |
+
"rewards/chosen": -0.984375,
|
| 1194 |
+
"rewards/margins": 7.40625,
|
| 1195 |
+
"rewards/rejected": -8.375,
|
| 1196 |
+
"step": 760
|
| 1197 |
+
},
|
| 1198 |
+
{
|
| 1199 |
+
"epoch": 2.467948717948718,
|
| 1200 |
+
"grad_norm": 6.059248925740205,
|
| 1201 |
+
"learning_rate": 9.857482185273158e-08,
|
| 1202 |
+
"logits/chosen": -1.53125,
|
| 1203 |
+
"logits/rejected": -1.515625,
|
| 1204 |
+
"logps/chosen": -137.0,
|
| 1205 |
+
"logps/rejected": -233.0,
|
| 1206 |
+
"loss": 0.037,
|
| 1207 |
+
"rewards/accuracies": 1.0,
|
| 1208 |
+
"rewards/chosen": -1.2265625,
|
| 1209 |
+
"rewards/margins": 5.90625,
|
| 1210 |
+
"rewards/rejected": -7.125,
|
| 1211 |
+
"step": 770
|
| 1212 |
+
},
|
| 1213 |
+
{
|
| 1214 |
+
"epoch": 2.5,
|
| 1215 |
+
"grad_norm": 5.569807408041334,
|
| 1216 |
+
"learning_rate": 9.263657957244655e-08,
|
| 1217 |
+
"logits/chosen": -1.546875,
|
| 1218 |
+
"logits/rejected": -1.53125,
|
| 1219 |
+
"logps/chosen": -136.0,
|
| 1220 |
+
"logps/rejected": -240.0,
|
| 1221 |
+
"loss": 0.0383,
|
| 1222 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1223 |
+
"rewards/chosen": -1.46875,
|
| 1224 |
+
"rewards/margins": 6.78125,
|
| 1225 |
+
"rewards/rejected": -8.25,
|
| 1226 |
+
"step": 780
|
| 1227 |
+
},
|
| 1228 |
+
{
|
| 1229 |
+
"epoch": 2.532051282051282,
|
| 1230 |
+
"grad_norm": 9.080566390829592,
|
| 1231 |
+
"learning_rate": 8.669833729216151e-08,
|
| 1232 |
+
"logits/chosen": -1.515625,
|
| 1233 |
+
"logits/rejected": -1.4375,
|
| 1234 |
+
"logps/chosen": -138.0,
|
| 1235 |
+
"logps/rejected": -233.0,
|
| 1236 |
+
"loss": 0.043,
|
| 1237 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1238 |
+
"rewards/chosen": -2.125,
|
| 1239 |
+
"rewards/margins": 6.1875,
|
| 1240 |
+
"rewards/rejected": -8.3125,
|
| 1241 |
+
"step": 790
|
| 1242 |
+
},
|
| 1243 |
+
{
|
| 1244 |
+
"epoch": 2.564102564102564,
|
| 1245 |
+
"grad_norm": 4.722752417362213,
|
| 1246 |
+
"learning_rate": 8.076009501187649e-08,
|
| 1247 |
+
"logits/chosen": -1.4609375,
|
| 1248 |
+
"logits/rejected": -1.515625,
|
| 1249 |
+
"logps/chosen": -129.0,
|
| 1250 |
+
"logps/rejected": -244.0,
|
| 1251 |
+
"loss": 0.043,
|
| 1252 |
+
"rewards/accuracies": 1.0,
|
| 1253 |
+
"rewards/chosen": -1.78125,
|
| 1254 |
+
"rewards/margins": 6.5625,
|
| 1255 |
+
"rewards/rejected": -8.375,
|
| 1256 |
+
"step": 800
|
| 1257 |
+
},
|
| 1258 |
+
{
|
| 1259 |
+
"epoch": 2.5961538461538463,
|
| 1260 |
+
"grad_norm": 8.320380078890436,
|
| 1261 |
+
"learning_rate": 7.482185273159145e-08,
|
| 1262 |
+
"logits/chosen": -1.5625,
|
| 1263 |
+
"logits/rejected": -1.484375,
|
| 1264 |
+
"logps/chosen": -152.0,
|
| 1265 |
+
"logps/rejected": -234.0,
|
| 1266 |
+
"loss": 0.0311,
|
| 1267 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1268 |
+
"rewards/chosen": -1.734375,
|
| 1269 |
+
"rewards/margins": 6.25,
|
| 1270 |
+
"rewards/rejected": -7.96875,
|
| 1271 |
+
"step": 810
|
| 1272 |
+
},
|
| 1273 |
+
{
|
| 1274 |
+
"epoch": 2.628205128205128,
|
| 1275 |
+
"grad_norm": 10.713886798327927,
|
| 1276 |
+
"learning_rate": 6.88836104513064e-08,
|
| 1277 |
+
"logits/chosen": -1.5078125,
|
| 1278 |
+
"logits/rejected": -1.4765625,
|
| 1279 |
+
"logps/chosen": -151.0,
|
| 1280 |
+
"logps/rejected": -252.0,
|
| 1281 |
+
"loss": 0.0368,
|
| 1282 |
+
"rewards/accuracies": 1.0,
|
| 1283 |
+
"rewards/chosen": -1.6875,
|
| 1284 |
+
"rewards/margins": 6.25,
|
| 1285 |
+
"rewards/rejected": -7.9375,
|
| 1286 |
+
"step": 820
|
| 1287 |
+
},
|
| 1288 |
+
{
|
| 1289 |
+
"epoch": 2.66025641025641,
|
| 1290 |
+
"grad_norm": 10.956365394123852,
|
| 1291 |
+
"learning_rate": 6.294536817102138e-08,
|
| 1292 |
+
"logits/chosen": -1.484375,
|
| 1293 |
+
"logits/rejected": -1.4375,
|
| 1294 |
+
"logps/chosen": -139.0,
|
| 1295 |
+
"logps/rejected": -224.0,
|
| 1296 |
+
"loss": 0.0434,
|
| 1297 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1298 |
+
"rewards/chosen": -1.7421875,
|
| 1299 |
+
"rewards/margins": 6.15625,
|
| 1300 |
+
"rewards/rejected": -7.875,
|
| 1301 |
+
"step": 830
|
| 1302 |
+
},
|
| 1303 |
+
{
|
| 1304 |
+
"epoch": 2.6923076923076925,
|
| 1305 |
+
"grad_norm": 5.167870965947855,
|
| 1306 |
+
"learning_rate": 5.700712589073634e-08,
|
| 1307 |
+
"logits/chosen": -1.5546875,
|
| 1308 |
+
"logits/rejected": -1.484375,
|
| 1309 |
+
"logps/chosen": -155.0,
|
| 1310 |
+
"logps/rejected": -286.0,
|
| 1311 |
+
"loss": 0.0367,
|
| 1312 |
+
"rewards/accuracies": 1.0,
|
| 1313 |
+
"rewards/chosen": -1.4140625,
|
| 1314 |
+
"rewards/margins": 7.15625,
|
| 1315 |
+
"rewards/rejected": -8.5625,
|
| 1316 |
+
"step": 840
|
| 1317 |
+
},
|
| 1318 |
+
{
|
| 1319 |
+
"epoch": 2.7243589743589745,
|
| 1320 |
+
"grad_norm": 10.357396384526298,
|
| 1321 |
+
"learning_rate": 5.10688836104513e-08,
|
| 1322 |
+
"logits/chosen": -1.4765625,
|
| 1323 |
+
"logits/rejected": -1.5,
|
| 1324 |
+
"logps/chosen": -144.0,
|
| 1325 |
+
"logps/rejected": -245.0,
|
| 1326 |
+
"loss": 0.0381,
|
| 1327 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1328 |
+
"rewards/chosen": -1.8125,
|
| 1329 |
+
"rewards/margins": 6.46875,
|
| 1330 |
+
"rewards/rejected": -8.3125,
|
| 1331 |
+
"step": 850
|
| 1332 |
+
},
|
| 1333 |
+
{
|
| 1334 |
+
"epoch": 2.7564102564102564,
|
| 1335 |
+
"grad_norm": 4.157699399133391,
|
| 1336 |
+
"learning_rate": 4.5130641330166267e-08,
|
| 1337 |
+
"logits/chosen": -1.46875,
|
| 1338 |
+
"logits/rejected": -1.421875,
|
| 1339 |
+
"logps/chosen": -126.0,
|
| 1340 |
+
"logps/rejected": -239.0,
|
| 1341 |
+
"loss": 0.0442,
|
| 1342 |
+
"rewards/accuracies": 0.9375,
|
| 1343 |
+
"rewards/chosen": -1.5859375,
|
| 1344 |
+
"rewards/margins": 6.25,
|
| 1345 |
+
"rewards/rejected": -7.8125,
|
| 1346 |
+
"step": 860
|
| 1347 |
+
},
|
| 1348 |
+
{
|
| 1349 |
+
"epoch": 2.7884615384615383,
|
| 1350 |
+
"grad_norm": 10.34195664309316,
|
| 1351 |
+
"learning_rate": 3.919239904988123e-08,
|
| 1352 |
+
"logits/chosen": -1.5234375,
|
| 1353 |
+
"logits/rejected": -1.5390625,
|
| 1354 |
+
"logps/chosen": -148.0,
|
| 1355 |
+
"logps/rejected": -268.0,
|
| 1356 |
+
"loss": 0.036,
|
| 1357 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1358 |
+
"rewards/chosen": -2.296875,
|
| 1359 |
+
"rewards/margins": 6.5,
|
| 1360 |
+
"rewards/rejected": -8.8125,
|
| 1361 |
+
"step": 870
|
| 1362 |
+
},
|
| 1363 |
+
{
|
| 1364 |
+
"epoch": 2.8205128205128203,
|
| 1365 |
+
"grad_norm": 13.885196758276543,
|
| 1366 |
+
"learning_rate": 3.32541567695962e-08,
|
| 1367 |
+
"logits/chosen": -1.53125,
|
| 1368 |
+
"logits/rejected": -1.4609375,
|
| 1369 |
+
"logps/chosen": -139.0,
|
| 1370 |
+
"logps/rejected": -237.0,
|
| 1371 |
+
"loss": 0.0386,
|
| 1372 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1373 |
+
"rewards/chosen": -1.7890625,
|
| 1374 |
+
"rewards/margins": 6.4375,
|
| 1375 |
+
"rewards/rejected": -8.25,
|
| 1376 |
+
"step": 880
|
| 1377 |
+
},
|
| 1378 |
+
{
|
| 1379 |
+
"epoch": 2.8525641025641026,
|
| 1380 |
+
"grad_norm": 3.298992157382736,
|
| 1381 |
+
"learning_rate": 2.7315914489311164e-08,
|
| 1382 |
+
"logits/chosen": -1.53125,
|
| 1383 |
+
"logits/rejected": -1.484375,
|
| 1384 |
+
"logps/chosen": -142.0,
|
| 1385 |
+
"logps/rejected": -264.0,
|
| 1386 |
+
"loss": 0.0326,
|
| 1387 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1388 |
+
"rewards/chosen": -2.0,
|
| 1389 |
+
"rewards/margins": 6.84375,
|
| 1390 |
+
"rewards/rejected": -8.875,
|
| 1391 |
+
"step": 890
|
| 1392 |
+
},
|
| 1393 |
+
{
|
| 1394 |
+
"epoch": 2.8846153846153846,
|
| 1395 |
+
"grad_norm": 9.55680354164709,
|
| 1396 |
+
"learning_rate": 2.1377672209026125e-08,
|
| 1397 |
+
"logits/chosen": -1.4609375,
|
| 1398 |
+
"logits/rejected": -1.4453125,
|
| 1399 |
+
"logps/chosen": -124.0,
|
| 1400 |
+
"logps/rejected": -247.0,
|
| 1401 |
+
"loss": 0.0386,
|
| 1402 |
+
"rewards/accuracies": 1.0,
|
| 1403 |
+
"rewards/chosen": -1.7734375,
|
| 1404 |
+
"rewards/margins": 7.0,
|
| 1405 |
+
"rewards/rejected": -8.8125,
|
| 1406 |
+
"step": 900
|
| 1407 |
+
},
|
| 1408 |
+
{
|
| 1409 |
+
"epoch": 2.9166666666666665,
|
| 1410 |
+
"grad_norm": 8.674386112287156,
|
| 1411 |
+
"learning_rate": 1.5439429928741092e-08,
|
| 1412 |
+
"logits/chosen": -1.5078125,
|
| 1413 |
+
"logits/rejected": -1.4765625,
|
| 1414 |
+
"logps/chosen": -118.0,
|
| 1415 |
+
"logps/rejected": -240.0,
|
| 1416 |
+
"loss": 0.0517,
|
| 1417 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1418 |
+
"rewards/chosen": -1.3984375,
|
| 1419 |
+
"rewards/margins": 7.0625,
|
| 1420 |
+
"rewards/rejected": -8.4375,
|
| 1421 |
+
"step": 910
|
| 1422 |
+
},
|
| 1423 |
+
{
|
| 1424 |
+
"epoch": 2.948717948717949,
|
| 1425 |
+
"grad_norm": 6.427610044394737,
|
| 1426 |
+
"learning_rate": 9.501187648456057e-09,
|
| 1427 |
+
"logits/chosen": -1.5,
|
| 1428 |
+
"logits/rejected": -1.4765625,
|
| 1429 |
+
"logps/chosen": -148.0,
|
| 1430 |
+
"logps/rejected": -256.0,
|
| 1431 |
+
"loss": 0.0481,
|
| 1432 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1433 |
+
"rewards/chosen": -1.84375,
|
| 1434 |
+
"rewards/margins": 6.375,
|
| 1435 |
+
"rewards/rejected": -8.25,
|
| 1436 |
+
"step": 920
|
| 1437 |
+
},
|
| 1438 |
+
{
|
| 1439 |
+
"epoch": 2.980769230769231,
|
| 1440 |
+
"grad_norm": 20.775958318465445,
|
| 1441 |
+
"learning_rate": 3.562945368171021e-09,
|
| 1442 |
+
"logits/chosen": -1.53125,
|
| 1443 |
+
"logits/rejected": -1.5,
|
| 1444 |
+
"logps/chosen": -141.0,
|
| 1445 |
+
"logps/rejected": -232.0,
|
| 1446 |
+
"loss": 0.0498,
|
| 1447 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1448 |
+
"rewards/chosen": -1.640625,
|
| 1449 |
+
"rewards/margins": 6.0,
|
| 1450 |
+
"rewards/rejected": -7.625,
|
| 1451 |
+
"step": 930
|
| 1452 |
+
},
|
| 1453 |
+
{
|
| 1454 |
+
"epoch": 3.0,
|
| 1455 |
+
"eval_logits/chosen": -1.4921875,
|
| 1456 |
+
"eval_logits/rejected": -1.4609375,
|
| 1457 |
+
"eval_logps/chosen": -199.0,
|
| 1458 |
+
"eval_logps/rejected": -222.0,
|
| 1459 |
+
"eval_loss": 0.830273449420929,
|
| 1460 |
+
"eval_rewards/accuracies": 0.625,
|
| 1461 |
+
"eval_rewards/chosen": -4.4375,
|
| 1462 |
+
"eval_rewards/margins": 1.765625,
|
| 1463 |
+
"eval_rewards/rejected": -6.1875,
|
| 1464 |
+
"eval_runtime": 16.7003,
|
| 1465 |
+
"eval_samples_per_second": 11.976,
|
| 1466 |
+
"eval_steps_per_second": 0.419,
|
| 1467 |
+
"step": 936
|
| 1468 |
+
},
|
| 1469 |
+
{
|
| 1470 |
+
"epoch": 3.0,
|
| 1471 |
+
"step": 936,
|
| 1472 |
+
"total_flos": 0.0,
|
| 1473 |
+
"train_loss": 0.1719489719113733,
|
| 1474 |
+
"train_runtime": 6022.2267,
|
| 1475 |
+
"train_samples_per_second": 4.969,
|
| 1476 |
+
"train_steps_per_second": 0.155
|
| 1477 |
+
}
|
| 1478 |
+
],
|
| 1479 |
+
"logging_steps": 10,
|
| 1480 |
+
"max_steps": 936,
|
| 1481 |
+
"num_input_tokens_seen": 0,
|
| 1482 |
+
"num_train_epochs": 3,
|
| 1483 |
+
"save_steps": 500,
|
| 1484 |
+
"stateful_callbacks": {
|
| 1485 |
+
"TrainerControl": {
|
| 1486 |
+
"args": {
|
| 1487 |
+
"should_epoch_stop": false,
|
| 1488 |
+
"should_evaluate": false,
|
| 1489 |
+
"should_log": false,
|
| 1490 |
+
"should_save": true,
|
| 1491 |
+
"should_training_stop": true
|
| 1492 |
+
},
|
| 1493 |
+
"attributes": {}
|
| 1494 |
+
}
|
| 1495 |
+
},
|
| 1496 |
+
"total_flos": 0.0,
|
| 1497 |
+
"train_batch_size": 8,
|
| 1498 |
+
"trial_name": null,
|
| 1499 |
+
"trial_params": null
|
| 1500 |
+
}
|