Model save
Browse files- README.md +68 -0
- all_results.json +9 -0
- generation_config.json +6 -0
- train_results.json +9 -0
- trainer_state.json +1500 -0
README.md
ADDED
|
@@ -0,0 +1,68 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
base_model: AmberYifan/Mistral-v0.1-7b-sft-ultrachat-hhrlhf
|
| 3 |
+
library_name: transformers
|
| 4 |
+
model_name: Mistral-7B-v0.1-sft-hhrlhf-dpo
|
| 5 |
+
tags:
|
| 6 |
+
- generated_from_trainer
|
| 7 |
+
- trl
|
| 8 |
+
- dpo
|
| 9 |
+
licence: license
|
| 10 |
+
---
|
| 11 |
+
|
| 12 |
+
# Model Card for Mistral-7B-v0.1-sft-hhrlhf-dpo
|
| 13 |
+
|
| 14 |
+
This model is a fine-tuned version of [AmberYifan/Mistral-v0.1-7b-sft-ultrachat-hhrlhf](https://huggingface.co/AmberYifan/Mistral-v0.1-7b-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/Mistral-7B-v0.1-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/enqkbvns)
|
| 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": 0.09964664405788767,
|
| 5 |
+
"train_runtime": 5808.5282,
|
| 6 |
+
"train_samples": 9951,
|
| 7 |
+
"train_samples_per_second": 5.14,
|
| 8 |
+
"train_steps_per_second": 0.161
|
| 9 |
+
}
|
generation_config.json
ADDED
|
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"bos_token_id": 1,
|
| 4 |
+
"eos_token_id": 2,
|
| 5 |
+
"transformers_version": "4.46.3"
|
| 6 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.09964664405788767,
|
| 5 |
+
"train_runtime": 5808.5282,
|
| 6 |
+
"train_samples": 9951,
|
| 7 |
+
"train_samples_per_second": 5.14,
|
| 8 |
+
"train_steps_per_second": 0.161
|
| 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": 933,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.003215434083601286,
|
| 13 |
+
"grad_norm": 79.87917506057431,
|
| 14 |
+
"learning_rate": 5.3191489361702125e-09,
|
| 15 |
+
"logits/chosen": -4.125,
|
| 16 |
+
"logits/rejected": -3.765625,
|
| 17 |
+
"logps/chosen": -132.0,
|
| 18 |
+
"logps/rejected": -141.0,
|
| 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": 0.03215434083601286,
|
| 28 |
+
"grad_norm": 118.20552774316917,
|
| 29 |
+
"learning_rate": 5.3191489361702123e-08,
|
| 30 |
+
"logits/chosen": -4.15625,
|
| 31 |
+
"logits/rejected": -3.8125,
|
| 32 |
+
"logps/chosen": -114.0,
|
| 33 |
+
"logps/rejected": -178.0,
|
| 34 |
+
"loss": 0.6858,
|
| 35 |
+
"rewards/accuracies": 0.2638888955116272,
|
| 36 |
+
"rewards/chosen": 0.0086669921875,
|
| 37 |
+
"rewards/margins": 0.00555419921875,
|
| 38 |
+
"rewards/rejected": 0.00311279296875,
|
| 39 |
+
"step": 10
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.06430868167202572,
|
| 43 |
+
"grad_norm": 66.66640901080964,
|
| 44 |
+
"learning_rate": 1.0638297872340425e-07,
|
| 45 |
+
"logits/chosen": -4.0,
|
| 46 |
+
"logits/rejected": -3.8125,
|
| 47 |
+
"logps/chosen": -122.0,
|
| 48 |
+
"logps/rejected": -203.0,
|
| 49 |
+
"loss": 0.6479,
|
| 50 |
+
"rewards/accuracies": 0.637499988079071,
|
| 51 |
+
"rewards/chosen": 0.05615234375,
|
| 52 |
+
"rewards/margins": 0.09423828125,
|
| 53 |
+
"rewards/rejected": -0.0380859375,
|
| 54 |
+
"step": 20
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.09646302250803858,
|
| 58 |
+
"grad_norm": 81.48311869513411,
|
| 59 |
+
"learning_rate": 1.5957446808510638e-07,
|
| 60 |
+
"logits/chosen": -4.03125,
|
| 61 |
+
"logits/rejected": -3.84375,
|
| 62 |
+
"logps/chosen": -142.0,
|
| 63 |
+
"logps/rejected": -224.0,
|
| 64 |
+
"loss": 0.5284,
|
| 65 |
+
"rewards/accuracies": 0.925000011920929,
|
| 66 |
+
"rewards/chosen": 0.1962890625,
|
| 67 |
+
"rewards/margins": 0.474609375,
|
| 68 |
+
"rewards/rejected": -0.279296875,
|
| 69 |
+
"step": 30
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.12861736334405144,
|
| 73 |
+
"grad_norm": 483.03383426380265,
|
| 74 |
+
"learning_rate": 2.127659574468085e-07,
|
| 75 |
+
"logits/chosen": -3.921875,
|
| 76 |
+
"logits/rejected": -3.859375,
|
| 77 |
+
"logps/chosen": -130.0,
|
| 78 |
+
"logps/rejected": -196.0,
|
| 79 |
+
"loss": 0.4049,
|
| 80 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 81 |
+
"rewards/chosen": 0.04638671875,
|
| 82 |
+
"rewards/margins": 0.9296875,
|
| 83 |
+
"rewards/rejected": -0.8828125,
|
| 84 |
+
"step": 40
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.1607717041800643,
|
| 88 |
+
"grad_norm": 521.8067650467351,
|
| 89 |
+
"learning_rate": 2.659574468085106e-07,
|
| 90 |
+
"logits/chosen": -3.875,
|
| 91 |
+
"logits/rejected": -3.90625,
|
| 92 |
+
"logps/chosen": -130.0,
|
| 93 |
+
"logps/rejected": -214.0,
|
| 94 |
+
"loss": 0.4117,
|
| 95 |
+
"rewards/accuracies": 0.949999988079071,
|
| 96 |
+
"rewards/chosen": 0.259765625,
|
| 97 |
+
"rewards/margins": 2.078125,
|
| 98 |
+
"rewards/rejected": -1.8203125,
|
| 99 |
+
"step": 50
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"epoch": 0.19292604501607716,
|
| 103 |
+
"grad_norm": 122.5164041936752,
|
| 104 |
+
"learning_rate": 3.1914893617021275e-07,
|
| 105 |
+
"logits/chosen": -3.90625,
|
| 106 |
+
"logits/rejected": -3.953125,
|
| 107 |
+
"logps/chosen": -116.0,
|
| 108 |
+
"logps/rejected": -214.0,
|
| 109 |
+
"loss": 0.2465,
|
| 110 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 111 |
+
"rewards/chosen": -0.248046875,
|
| 112 |
+
"rewards/margins": 2.34375,
|
| 113 |
+
"rewards/rejected": -2.59375,
|
| 114 |
+
"step": 60
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.22508038585209003,
|
| 118 |
+
"grad_norm": 98.58564731221452,
|
| 119 |
+
"learning_rate": 3.7234042553191484e-07,
|
| 120 |
+
"logits/chosen": -3.84375,
|
| 121 |
+
"logits/rejected": -3.90625,
|
| 122 |
+
"logps/chosen": -130.0,
|
| 123 |
+
"logps/rejected": -274.0,
|
| 124 |
+
"loss": 0.2316,
|
| 125 |
+
"rewards/accuracies": 0.925000011920929,
|
| 126 |
+
"rewards/chosen": -0.1923828125,
|
| 127 |
+
"rewards/margins": 3.359375,
|
| 128 |
+
"rewards/rejected": -3.546875,
|
| 129 |
+
"step": 70
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"epoch": 0.2572347266881029,
|
| 133 |
+
"grad_norm": 58.684567343268625,
|
| 134 |
+
"learning_rate": 4.25531914893617e-07,
|
| 135 |
+
"logits/chosen": -3.6875,
|
| 136 |
+
"logits/rejected": -3.828125,
|
| 137 |
+
"logps/chosen": -144.0,
|
| 138 |
+
"logps/rejected": -209.0,
|
| 139 |
+
"loss": 0.3018,
|
| 140 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 141 |
+
"rewards/chosen": -0.50390625,
|
| 142 |
+
"rewards/margins": 2.765625,
|
| 143 |
+
"rewards/rejected": -3.265625,
|
| 144 |
+
"step": 80
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 0.28938906752411575,
|
| 148 |
+
"grad_norm": 44.1285015532565,
|
| 149 |
+
"learning_rate": 4.787234042553192e-07,
|
| 150 |
+
"logits/chosen": -3.484375,
|
| 151 |
+
"logits/rejected": -3.65625,
|
| 152 |
+
"logps/chosen": -121.0,
|
| 153 |
+
"logps/rejected": -198.0,
|
| 154 |
+
"loss": 0.3009,
|
| 155 |
+
"rewards/accuracies": 0.925000011920929,
|
| 156 |
+
"rewards/chosen": -0.181640625,
|
| 157 |
+
"rewards/margins": 2.4375,
|
| 158 |
+
"rewards/rejected": -2.625,
|
| 159 |
+
"step": 90
|
| 160 |
+
},
|
| 161 |
+
{
|
| 162 |
+
"epoch": 0.3215434083601286,
|
| 163 |
+
"grad_norm": 50.263563110645,
|
| 164 |
+
"learning_rate": 4.964243146603099e-07,
|
| 165 |
+
"logits/chosen": -3.5,
|
| 166 |
+
"logits/rejected": -3.609375,
|
| 167 |
+
"logps/chosen": -150.0,
|
| 168 |
+
"logps/rejected": -222.0,
|
| 169 |
+
"loss": 0.2831,
|
| 170 |
+
"rewards/accuracies": 0.875,
|
| 171 |
+
"rewards/chosen": -4.887580871582031e-05,
|
| 172 |
+
"rewards/margins": 3.09375,
|
| 173 |
+
"rewards/rejected": -3.09375,
|
| 174 |
+
"step": 100
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"epoch": 0.3536977491961415,
|
| 178 |
+
"grad_norm": 77.17996002738214,
|
| 179 |
+
"learning_rate": 4.904648390941597e-07,
|
| 180 |
+
"logits/chosen": -3.578125,
|
| 181 |
+
"logits/rejected": -3.703125,
|
| 182 |
+
"logps/chosen": -114.5,
|
| 183 |
+
"logps/rejected": -233.0,
|
| 184 |
+
"loss": 0.1551,
|
| 185 |
+
"rewards/accuracies": 0.949999988079071,
|
| 186 |
+
"rewards/chosen": -0.0654296875,
|
| 187 |
+
"rewards/margins": 3.828125,
|
| 188 |
+
"rewards/rejected": -3.890625,
|
| 189 |
+
"step": 110
|
| 190 |
+
},
|
| 191 |
+
{
|
| 192 |
+
"epoch": 0.3858520900321543,
|
| 193 |
+
"grad_norm": 50.629644956690505,
|
| 194 |
+
"learning_rate": 4.845053635280095e-07,
|
| 195 |
+
"logits/chosen": -3.59375,
|
| 196 |
+
"logits/rejected": -3.8125,
|
| 197 |
+
"logps/chosen": -120.5,
|
| 198 |
+
"logps/rejected": -262.0,
|
| 199 |
+
"loss": 0.1768,
|
| 200 |
+
"rewards/accuracies": 0.887499988079071,
|
| 201 |
+
"rewards/chosen": -0.388671875,
|
| 202 |
+
"rewards/margins": 4.15625,
|
| 203 |
+
"rewards/rejected": -4.53125,
|
| 204 |
+
"step": 120
|
| 205 |
+
},
|
| 206 |
+
{
|
| 207 |
+
"epoch": 0.4180064308681672,
|
| 208 |
+
"grad_norm": 87.75654163889804,
|
| 209 |
+
"learning_rate": 4.785458879618593e-07,
|
| 210 |
+
"logits/chosen": -3.484375,
|
| 211 |
+
"logits/rejected": -3.640625,
|
| 212 |
+
"logps/chosen": -153.0,
|
| 213 |
+
"logps/rejected": -280.0,
|
| 214 |
+
"loss": 0.1777,
|
| 215 |
+
"rewards/accuracies": 0.949999988079071,
|
| 216 |
+
"rewards/chosen": -0.66796875,
|
| 217 |
+
"rewards/margins": 4.25,
|
| 218 |
+
"rewards/rejected": -4.90625,
|
| 219 |
+
"step": 130
|
| 220 |
+
},
|
| 221 |
+
{
|
| 222 |
+
"epoch": 0.45016077170418006,
|
| 223 |
+
"grad_norm": 48.30101230741645,
|
| 224 |
+
"learning_rate": 4.7258641239570916e-07,
|
| 225 |
+
"logits/chosen": -3.640625,
|
| 226 |
+
"logits/rejected": -3.765625,
|
| 227 |
+
"logps/chosen": -119.0,
|
| 228 |
+
"logps/rejected": -233.0,
|
| 229 |
+
"loss": 0.1792,
|
| 230 |
+
"rewards/accuracies": 0.925000011920929,
|
| 231 |
+
"rewards/chosen": 0.142578125,
|
| 232 |
+
"rewards/margins": 3.859375,
|
| 233 |
+
"rewards/rejected": -3.71875,
|
| 234 |
+
"step": 140
|
| 235 |
+
},
|
| 236 |
+
{
|
| 237 |
+
"epoch": 0.48231511254019294,
|
| 238 |
+
"grad_norm": 160.46320495134455,
|
| 239 |
+
"learning_rate": 4.66626936829559e-07,
|
| 240 |
+
"logits/chosen": -3.75,
|
| 241 |
+
"logits/rejected": -3.875,
|
| 242 |
+
"logps/chosen": -111.0,
|
| 243 |
+
"logps/rejected": -248.0,
|
| 244 |
+
"loss": 0.3135,
|
| 245 |
+
"rewards/accuracies": 0.887499988079071,
|
| 246 |
+
"rewards/chosen": 0.1611328125,
|
| 247 |
+
"rewards/margins": 4.03125,
|
| 248 |
+
"rewards/rejected": -3.875,
|
| 249 |
+
"step": 150
|
| 250 |
+
},
|
| 251 |
+
{
|
| 252 |
+
"epoch": 0.5144694533762058,
|
| 253 |
+
"grad_norm": 27.202159849500408,
|
| 254 |
+
"learning_rate": 4.606674612634088e-07,
|
| 255 |
+
"logits/chosen": -3.5625,
|
| 256 |
+
"logits/rejected": -3.765625,
|
| 257 |
+
"logps/chosen": -132.0,
|
| 258 |
+
"logps/rejected": -260.0,
|
| 259 |
+
"loss": 0.1584,
|
| 260 |
+
"rewards/accuracies": 0.9375,
|
| 261 |
+
"rewards/chosen": 0.181640625,
|
| 262 |
+
"rewards/margins": 4.34375,
|
| 263 |
+
"rewards/rejected": -4.15625,
|
| 264 |
+
"step": 160
|
| 265 |
+
},
|
| 266 |
+
{
|
| 267 |
+
"epoch": 0.5466237942122186,
|
| 268 |
+
"grad_norm": 59.16977770326313,
|
| 269 |
+
"learning_rate": 4.547079856972586e-07,
|
| 270 |
+
"logits/chosen": -3.734375,
|
| 271 |
+
"logits/rejected": -3.796875,
|
| 272 |
+
"logps/chosen": -140.0,
|
| 273 |
+
"logps/rejected": -241.0,
|
| 274 |
+
"loss": 0.1732,
|
| 275 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 276 |
+
"rewards/chosen": -0.1953125,
|
| 277 |
+
"rewards/margins": 3.8125,
|
| 278 |
+
"rewards/rejected": -4.0,
|
| 279 |
+
"step": 170
|
| 280 |
+
},
|
| 281 |
+
{
|
| 282 |
+
"epoch": 0.5787781350482315,
|
| 283 |
+
"grad_norm": 54.881650170057746,
|
| 284 |
+
"learning_rate": 4.487485101311084e-07,
|
| 285 |
+
"logits/chosen": -3.625,
|
| 286 |
+
"logits/rejected": -3.75,
|
| 287 |
+
"logps/chosen": -122.0,
|
| 288 |
+
"logps/rejected": -239.0,
|
| 289 |
+
"loss": 0.2094,
|
| 290 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 291 |
+
"rewards/chosen": -0.58203125,
|
| 292 |
+
"rewards/margins": 3.734375,
|
| 293 |
+
"rewards/rejected": -4.34375,
|
| 294 |
+
"step": 180
|
| 295 |
+
},
|
| 296 |
+
{
|
| 297 |
+
"epoch": 0.6109324758842444,
|
| 298 |
+
"grad_norm": 46.393130621023865,
|
| 299 |
+
"learning_rate": 4.4278903456495827e-07,
|
| 300 |
+
"logits/chosen": -3.5,
|
| 301 |
+
"logits/rejected": -3.734375,
|
| 302 |
+
"logps/chosen": -105.5,
|
| 303 |
+
"logps/rejected": -248.0,
|
| 304 |
+
"loss": 0.1894,
|
| 305 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 306 |
+
"rewards/chosen": 0.248046875,
|
| 307 |
+
"rewards/margins": 4.0,
|
| 308 |
+
"rewards/rejected": -3.765625,
|
| 309 |
+
"step": 190
|
| 310 |
+
},
|
| 311 |
+
{
|
| 312 |
+
"epoch": 0.6430868167202572,
|
| 313 |
+
"grad_norm": 61.77038241219252,
|
| 314 |
+
"learning_rate": 4.368295589988081e-07,
|
| 315 |
+
"logits/chosen": -3.546875,
|
| 316 |
+
"logits/rejected": -3.609375,
|
| 317 |
+
"logps/chosen": -121.5,
|
| 318 |
+
"logps/rejected": -246.0,
|
| 319 |
+
"loss": 0.2004,
|
| 320 |
+
"rewards/accuracies": 0.9375,
|
| 321 |
+
"rewards/chosen": 0.43359375,
|
| 322 |
+
"rewards/margins": 3.84375,
|
| 323 |
+
"rewards/rejected": -3.421875,
|
| 324 |
+
"step": 200
|
| 325 |
+
},
|
| 326 |
+
{
|
| 327 |
+
"epoch": 0.6752411575562701,
|
| 328 |
+
"grad_norm": 54.13127427585697,
|
| 329 |
+
"learning_rate": 4.308700834326579e-07,
|
| 330 |
+
"logits/chosen": -3.546875,
|
| 331 |
+
"logits/rejected": -3.8125,
|
| 332 |
+
"logps/chosen": -134.0,
|
| 333 |
+
"logps/rejected": -258.0,
|
| 334 |
+
"loss": 0.1556,
|
| 335 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 336 |
+
"rewards/chosen": 0.259765625,
|
| 337 |
+
"rewards/margins": 4.46875,
|
| 338 |
+
"rewards/rejected": -4.1875,
|
| 339 |
+
"step": 210
|
| 340 |
+
},
|
| 341 |
+
{
|
| 342 |
+
"epoch": 0.707395498392283,
|
| 343 |
+
"grad_norm": 39.76866338516343,
|
| 344 |
+
"learning_rate": 4.249106078665077e-07,
|
| 345 |
+
"logits/chosen": -3.734375,
|
| 346 |
+
"logits/rejected": -3.84375,
|
| 347 |
+
"logps/chosen": -114.5,
|
| 348 |
+
"logps/rejected": -254.0,
|
| 349 |
+
"loss": 0.1614,
|
| 350 |
+
"rewards/accuracies": 0.9375,
|
| 351 |
+
"rewards/chosen": 0.41796875,
|
| 352 |
+
"rewards/margins": 4.9375,
|
| 353 |
+
"rewards/rejected": -4.53125,
|
| 354 |
+
"step": 220
|
| 355 |
+
},
|
| 356 |
+
{
|
| 357 |
+
"epoch": 0.7395498392282959,
|
| 358 |
+
"grad_norm": 19.376631378140395,
|
| 359 |
+
"learning_rate": 4.1895113230035757e-07,
|
| 360 |
+
"logits/chosen": -3.65625,
|
| 361 |
+
"logits/rejected": -3.703125,
|
| 362 |
+
"logps/chosen": -125.5,
|
| 363 |
+
"logps/rejected": -266.0,
|
| 364 |
+
"loss": 0.2333,
|
| 365 |
+
"rewards/accuracies": 0.949999988079071,
|
| 366 |
+
"rewards/chosen": 0.96484375,
|
| 367 |
+
"rewards/margins": 4.5,
|
| 368 |
+
"rewards/rejected": -3.515625,
|
| 369 |
+
"step": 230
|
| 370 |
+
},
|
| 371 |
+
{
|
| 372 |
+
"epoch": 0.7717041800643086,
|
| 373 |
+
"grad_norm": 38.972752947728395,
|
| 374 |
+
"learning_rate": 4.129916567342074e-07,
|
| 375 |
+
"logits/chosen": -3.53125,
|
| 376 |
+
"logits/rejected": -3.5625,
|
| 377 |
+
"logps/chosen": -133.0,
|
| 378 |
+
"logps/rejected": -266.0,
|
| 379 |
+
"loss": 0.2215,
|
| 380 |
+
"rewards/accuracies": 0.949999988079071,
|
| 381 |
+
"rewards/chosen": 1.46875,
|
| 382 |
+
"rewards/margins": 5.03125,
|
| 383 |
+
"rewards/rejected": -3.578125,
|
| 384 |
+
"step": 240
|
| 385 |
+
},
|
| 386 |
+
{
|
| 387 |
+
"epoch": 0.8038585209003215,
|
| 388 |
+
"grad_norm": 25.549122278044088,
|
| 389 |
+
"learning_rate": 4.070321811680572e-07,
|
| 390 |
+
"logits/chosen": -3.546875,
|
| 391 |
+
"logits/rejected": -3.671875,
|
| 392 |
+
"logps/chosen": -139.0,
|
| 393 |
+
"logps/rejected": -238.0,
|
| 394 |
+
"loss": 0.1699,
|
| 395 |
+
"rewards/accuracies": 0.887499988079071,
|
| 396 |
+
"rewards/chosen": -0.59375,
|
| 397 |
+
"rewards/margins": 4.46875,
|
| 398 |
+
"rewards/rejected": -5.09375,
|
| 399 |
+
"step": 250
|
| 400 |
+
},
|
| 401 |
+
{
|
| 402 |
+
"epoch": 0.8360128617363344,
|
| 403 |
+
"grad_norm": 36.7792748916066,
|
| 404 |
+
"learning_rate": 4.0107270560190706e-07,
|
| 405 |
+
"logits/chosen": -3.640625,
|
| 406 |
+
"logits/rejected": -3.84375,
|
| 407 |
+
"logps/chosen": -125.0,
|
| 408 |
+
"logps/rejected": -248.0,
|
| 409 |
+
"loss": 0.1649,
|
| 410 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 411 |
+
"rewards/chosen": 0.51171875,
|
| 412 |
+
"rewards/margins": 4.53125,
|
| 413 |
+
"rewards/rejected": -4.03125,
|
| 414 |
+
"step": 260
|
| 415 |
+
},
|
| 416 |
+
{
|
| 417 |
+
"epoch": 0.8681672025723473,
|
| 418 |
+
"grad_norm": 10.885169604932974,
|
| 419 |
+
"learning_rate": 3.9511323003575687e-07,
|
| 420 |
+
"logits/chosen": -3.59375,
|
| 421 |
+
"logits/rejected": -3.640625,
|
| 422 |
+
"logps/chosen": -145.0,
|
| 423 |
+
"logps/rejected": -246.0,
|
| 424 |
+
"loss": 0.1556,
|
| 425 |
+
"rewards/accuracies": 0.925000011920929,
|
| 426 |
+
"rewards/chosen": 0.396484375,
|
| 427 |
+
"rewards/margins": 4.53125,
|
| 428 |
+
"rewards/rejected": -4.125,
|
| 429 |
+
"step": 270
|
| 430 |
+
},
|
| 431 |
+
{
|
| 432 |
+
"epoch": 0.9003215434083601,
|
| 433 |
+
"grad_norm": 33.68141687625864,
|
| 434 |
+
"learning_rate": 3.8915375446960663e-07,
|
| 435 |
+
"logits/chosen": -3.5625,
|
| 436 |
+
"logits/rejected": -3.734375,
|
| 437 |
+
"logps/chosen": -132.0,
|
| 438 |
+
"logps/rejected": -276.0,
|
| 439 |
+
"loss": 0.13,
|
| 440 |
+
"rewards/accuracies": 0.949999988079071,
|
| 441 |
+
"rewards/chosen": -0.203125,
|
| 442 |
+
"rewards/margins": 5.125,
|
| 443 |
+
"rewards/rejected": -5.34375,
|
| 444 |
+
"step": 280
|
| 445 |
+
},
|
| 446 |
+
{
|
| 447 |
+
"epoch": 0.932475884244373,
|
| 448 |
+
"grad_norm": 38.54727132055094,
|
| 449 |
+
"learning_rate": 3.8319427890345644e-07,
|
| 450 |
+
"logits/chosen": -3.453125,
|
| 451 |
+
"logits/rejected": -3.578125,
|
| 452 |
+
"logps/chosen": -136.0,
|
| 453 |
+
"logps/rejected": -266.0,
|
| 454 |
+
"loss": 0.1626,
|
| 455 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 456 |
+
"rewards/chosen": -0.1279296875,
|
| 457 |
+
"rewards/margins": 4.96875,
|
| 458 |
+
"rewards/rejected": -5.09375,
|
| 459 |
+
"step": 290
|
| 460 |
+
},
|
| 461 |
+
{
|
| 462 |
+
"epoch": 0.9646302250803859,
|
| 463 |
+
"grad_norm": 33.100977527402115,
|
| 464 |
+
"learning_rate": 3.772348033373063e-07,
|
| 465 |
+
"logits/chosen": -3.4375,
|
| 466 |
+
"logits/rejected": -3.65625,
|
| 467 |
+
"logps/chosen": -137.0,
|
| 468 |
+
"logps/rejected": -256.0,
|
| 469 |
+
"loss": 0.1537,
|
| 470 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 471 |
+
"rewards/chosen": 0.031982421875,
|
| 472 |
+
"rewards/margins": 4.96875,
|
| 473 |
+
"rewards/rejected": -4.9375,
|
| 474 |
+
"step": 300
|
| 475 |
+
},
|
| 476 |
+
{
|
| 477 |
+
"epoch": 0.9967845659163987,
|
| 478 |
+
"grad_norm": 67.60299631407663,
|
| 479 |
+
"learning_rate": 3.712753277711561e-07,
|
| 480 |
+
"logits/chosen": -3.484375,
|
| 481 |
+
"logits/rejected": -3.703125,
|
| 482 |
+
"logps/chosen": -143.0,
|
| 483 |
+
"logps/rejected": -296.0,
|
| 484 |
+
"loss": 0.1769,
|
| 485 |
+
"rewards/accuracies": 0.949999988079071,
|
| 486 |
+
"rewards/chosen": 0.10888671875,
|
| 487 |
+
"rewards/margins": 5.4375,
|
| 488 |
+
"rewards/rejected": -5.34375,
|
| 489 |
+
"step": 310
|
| 490 |
+
},
|
| 491 |
+
{
|
| 492 |
+
"epoch": 1.0,
|
| 493 |
+
"eval_logits/chosen": -3.78125,
|
| 494 |
+
"eval_logits/rejected": -3.78125,
|
| 495 |
+
"eval_logps/chosen": -206.0,
|
| 496 |
+
"eval_logps/rejected": -234.0,
|
| 497 |
+
"eval_loss": 0.5869876742362976,
|
| 498 |
+
"eval_rewards/accuracies": 0.7678571343421936,
|
| 499 |
+
"eval_rewards/chosen": -2.078125,
|
| 500 |
+
"eval_rewards/margins": 2.265625,
|
| 501 |
+
"eval_rewards/rejected": -4.34375,
|
| 502 |
+
"eval_runtime": 14.2416,
|
| 503 |
+
"eval_samples_per_second": 13.973,
|
| 504 |
+
"eval_steps_per_second": 0.492,
|
| 505 |
+
"step": 311
|
| 506 |
+
},
|
| 507 |
+
{
|
| 508 |
+
"epoch": 1.0289389067524115,
|
| 509 |
+
"grad_norm": 17.744370347927422,
|
| 510 |
+
"learning_rate": 3.6531585220500593e-07,
|
| 511 |
+
"logits/chosen": -3.546875,
|
| 512 |
+
"logits/rejected": -3.71875,
|
| 513 |
+
"logps/chosen": -104.0,
|
| 514 |
+
"logps/rejected": -240.0,
|
| 515 |
+
"loss": 0.0588,
|
| 516 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 517 |
+
"rewards/chosen": 0.56640625,
|
| 518 |
+
"rewards/margins": 6.75,
|
| 519 |
+
"rewards/rejected": -6.1875,
|
| 520 |
+
"step": 320
|
| 521 |
+
},
|
| 522 |
+
{
|
| 523 |
+
"epoch": 1.0610932475884245,
|
| 524 |
+
"grad_norm": 4.160707504041419,
|
| 525 |
+
"learning_rate": 3.5935637663885575e-07,
|
| 526 |
+
"logits/chosen": -3.484375,
|
| 527 |
+
"logits/rejected": -3.65625,
|
| 528 |
+
"logps/chosen": -113.0,
|
| 529 |
+
"logps/rejected": -286.0,
|
| 530 |
+
"loss": 0.0337,
|
| 531 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 532 |
+
"rewards/chosen": 1.2265625,
|
| 533 |
+
"rewards/margins": 8.4375,
|
| 534 |
+
"rewards/rejected": -7.1875,
|
| 535 |
+
"step": 330
|
| 536 |
+
},
|
| 537 |
+
{
|
| 538 |
+
"epoch": 1.0932475884244373,
|
| 539 |
+
"grad_norm": 17.23951731292116,
|
| 540 |
+
"learning_rate": 3.533969010727056e-07,
|
| 541 |
+
"logits/chosen": -3.640625,
|
| 542 |
+
"logits/rejected": -3.796875,
|
| 543 |
+
"logps/chosen": -115.5,
|
| 544 |
+
"logps/rejected": -252.0,
|
| 545 |
+
"loss": 0.0347,
|
| 546 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 547 |
+
"rewards/chosen": 1.1875,
|
| 548 |
+
"rewards/margins": 8.375,
|
| 549 |
+
"rewards/rejected": -7.1875,
|
| 550 |
+
"step": 340
|
| 551 |
+
},
|
| 552 |
+
{
|
| 553 |
+
"epoch": 1.1254019292604502,
|
| 554 |
+
"grad_norm": 49.64075213527915,
|
| 555 |
+
"learning_rate": 3.474374255065554e-07,
|
| 556 |
+
"logits/chosen": -3.40625,
|
| 557 |
+
"logits/rejected": -3.5625,
|
| 558 |
+
"logps/chosen": -140.0,
|
| 559 |
+
"logps/rejected": -300.0,
|
| 560 |
+
"loss": 0.0189,
|
| 561 |
+
"rewards/accuracies": 0.987500011920929,
|
| 562 |
+
"rewards/chosen": 1.0390625,
|
| 563 |
+
"rewards/margins": 8.625,
|
| 564 |
+
"rewards/rejected": -7.625,
|
| 565 |
+
"step": 350
|
| 566 |
+
},
|
| 567 |
+
{
|
| 568 |
+
"epoch": 1.157556270096463,
|
| 569 |
+
"grad_norm": 7.306932203883141,
|
| 570 |
+
"learning_rate": 3.4147794994040524e-07,
|
| 571 |
+
"logits/chosen": -3.171875,
|
| 572 |
+
"logits/rejected": -3.421875,
|
| 573 |
+
"logps/chosen": -109.5,
|
| 574 |
+
"logps/rejected": -274.0,
|
| 575 |
+
"loss": 0.0278,
|
| 576 |
+
"rewards/accuracies": 1.0,
|
| 577 |
+
"rewards/chosen": 1.4609375,
|
| 578 |
+
"rewards/margins": 9.5625,
|
| 579 |
+
"rewards/rejected": -8.125,
|
| 580 |
+
"step": 360
|
| 581 |
+
},
|
| 582 |
+
{
|
| 583 |
+
"epoch": 1.189710610932476,
|
| 584 |
+
"grad_norm": 35.16526862248255,
|
| 585 |
+
"learning_rate": 3.3551847437425505e-07,
|
| 586 |
+
"logits/chosen": -3.25,
|
| 587 |
+
"logits/rejected": -3.421875,
|
| 588 |
+
"logps/chosen": -120.5,
|
| 589 |
+
"logps/rejected": -282.0,
|
| 590 |
+
"loss": 0.0399,
|
| 591 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 592 |
+
"rewards/chosen": 0.44140625,
|
| 593 |
+
"rewards/margins": 8.8125,
|
| 594 |
+
"rewards/rejected": -8.375,
|
| 595 |
+
"step": 370
|
| 596 |
+
},
|
| 597 |
+
{
|
| 598 |
+
"epoch": 1.2218649517684887,
|
| 599 |
+
"grad_norm": 7.575200512680006,
|
| 600 |
+
"learning_rate": 3.295589988081049e-07,
|
| 601 |
+
"logits/chosen": -3.359375,
|
| 602 |
+
"logits/rejected": -3.515625,
|
| 603 |
+
"logps/chosen": -120.0,
|
| 604 |
+
"logps/rejected": -300.0,
|
| 605 |
+
"loss": 0.0216,
|
| 606 |
+
"rewards/accuracies": 0.987500011920929,
|
| 607 |
+
"rewards/chosen": 0.9140625,
|
| 608 |
+
"rewards/margins": 9.3125,
|
| 609 |
+
"rewards/rejected": -8.4375,
|
| 610 |
+
"step": 380
|
| 611 |
+
},
|
| 612 |
+
{
|
| 613 |
+
"epoch": 1.2540192926045015,
|
| 614 |
+
"grad_norm": 0.6452900307805328,
|
| 615 |
+
"learning_rate": 3.235995232419547e-07,
|
| 616 |
+
"logits/chosen": -3.265625,
|
| 617 |
+
"logits/rejected": -3.453125,
|
| 618 |
+
"logps/chosen": -132.0,
|
| 619 |
+
"logps/rejected": -322.0,
|
| 620 |
+
"loss": 0.0475,
|
| 621 |
+
"rewards/accuracies": 0.987500011920929,
|
| 622 |
+
"rewards/chosen": 0.8515625,
|
| 623 |
+
"rewards/margins": 10.0,
|
| 624 |
+
"rewards/rejected": -9.1875,
|
| 625 |
+
"step": 390
|
| 626 |
+
},
|
| 627 |
+
{
|
| 628 |
+
"epoch": 1.2861736334405145,
|
| 629 |
+
"grad_norm": 7.420661461533486,
|
| 630 |
+
"learning_rate": 3.176400476758045e-07,
|
| 631 |
+
"logits/chosen": -3.15625,
|
| 632 |
+
"logits/rejected": -3.328125,
|
| 633 |
+
"logps/chosen": -105.5,
|
| 634 |
+
"logps/rejected": -272.0,
|
| 635 |
+
"loss": 0.0335,
|
| 636 |
+
"rewards/accuracies": 0.987500011920929,
|
| 637 |
+
"rewards/chosen": 0.90625,
|
| 638 |
+
"rewards/margins": 9.1875,
|
| 639 |
+
"rewards/rejected": -8.25,
|
| 640 |
+
"step": 400
|
| 641 |
+
},
|
| 642 |
+
{
|
| 643 |
+
"epoch": 1.3183279742765273,
|
| 644 |
+
"grad_norm": 27.92454063116238,
|
| 645 |
+
"learning_rate": 3.116805721096543e-07,
|
| 646 |
+
"logits/chosen": -3.28125,
|
| 647 |
+
"logits/rejected": -3.484375,
|
| 648 |
+
"logps/chosen": -129.0,
|
| 649 |
+
"logps/rejected": -282.0,
|
| 650 |
+
"loss": 0.0376,
|
| 651 |
+
"rewards/accuracies": 0.987500011920929,
|
| 652 |
+
"rewards/chosen": 0.023193359375,
|
| 653 |
+
"rewards/margins": 9.0625,
|
| 654 |
+
"rewards/rejected": -9.0625,
|
| 655 |
+
"step": 410
|
| 656 |
+
},
|
| 657 |
+
{
|
| 658 |
+
"epoch": 1.3504823151125402,
|
| 659 |
+
"grad_norm": 48.832852324456454,
|
| 660 |
+
"learning_rate": 3.0572109654350416e-07,
|
| 661 |
+
"logits/chosen": -3.203125,
|
| 662 |
+
"logits/rejected": -3.34375,
|
| 663 |
+
"logps/chosen": -134.0,
|
| 664 |
+
"logps/rejected": -282.0,
|
| 665 |
+
"loss": 0.0261,
|
| 666 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 667 |
+
"rewards/chosen": 0.173828125,
|
| 668 |
+
"rewards/margins": 9.0625,
|
| 669 |
+
"rewards/rejected": -8.875,
|
| 670 |
+
"step": 420
|
| 671 |
+
},
|
| 672 |
+
{
|
| 673 |
+
"epoch": 1.382636655948553,
|
| 674 |
+
"grad_norm": 65.54781768372352,
|
| 675 |
+
"learning_rate": 2.99761620977354e-07,
|
| 676 |
+
"logits/chosen": -3.28125,
|
| 677 |
+
"logits/rejected": -3.46875,
|
| 678 |
+
"logps/chosen": -117.0,
|
| 679 |
+
"logps/rejected": -276.0,
|
| 680 |
+
"loss": 0.0396,
|
| 681 |
+
"rewards/accuracies": 0.987500011920929,
|
| 682 |
+
"rewards/chosen": 0.494140625,
|
| 683 |
+
"rewards/margins": 9.1875,
|
| 684 |
+
"rewards/rejected": -8.6875,
|
| 685 |
+
"step": 430
|
| 686 |
+
},
|
| 687 |
+
{
|
| 688 |
+
"epoch": 1.414790996784566,
|
| 689 |
+
"grad_norm": 25.7310269564794,
|
| 690 |
+
"learning_rate": 2.938021454112038e-07,
|
| 691 |
+
"logits/chosen": -3.25,
|
| 692 |
+
"logits/rejected": -3.484375,
|
| 693 |
+
"logps/chosen": -156.0,
|
| 694 |
+
"logps/rejected": -316.0,
|
| 695 |
+
"loss": 0.0399,
|
| 696 |
+
"rewards/accuracies": 0.987500011920929,
|
| 697 |
+
"rewards/chosen": 0.78515625,
|
| 698 |
+
"rewards/margins": 9.125,
|
| 699 |
+
"rewards/rejected": -8.375,
|
| 700 |
+
"step": 440
|
| 701 |
+
},
|
| 702 |
+
{
|
| 703 |
+
"epoch": 1.4469453376205788,
|
| 704 |
+
"grad_norm": 8.455918132650991,
|
| 705 |
+
"learning_rate": 2.878426698450536e-07,
|
| 706 |
+
"logits/chosen": -3.25,
|
| 707 |
+
"logits/rejected": -3.484375,
|
| 708 |
+
"logps/chosen": -110.5,
|
| 709 |
+
"logps/rejected": -255.0,
|
| 710 |
+
"loss": 0.0372,
|
| 711 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 712 |
+
"rewards/chosen": 0.6640625,
|
| 713 |
+
"rewards/margins": 8.75,
|
| 714 |
+
"rewards/rejected": -8.0625,
|
| 715 |
+
"step": 450
|
| 716 |
+
},
|
| 717 |
+
{
|
| 718 |
+
"epoch": 1.4790996784565915,
|
| 719 |
+
"grad_norm": 23.483398578435676,
|
| 720 |
+
"learning_rate": 2.8188319427890346e-07,
|
| 721 |
+
"logits/chosen": -3.296875,
|
| 722 |
+
"logits/rejected": -3.515625,
|
| 723 |
+
"logps/chosen": -115.5,
|
| 724 |
+
"logps/rejected": -304.0,
|
| 725 |
+
"loss": 0.0359,
|
| 726 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 727 |
+
"rewards/chosen": 0.91796875,
|
| 728 |
+
"rewards/margins": 9.5,
|
| 729 |
+
"rewards/rejected": -8.5625,
|
| 730 |
+
"step": 460
|
| 731 |
+
},
|
| 732 |
+
{
|
| 733 |
+
"epoch": 1.5112540192926045,
|
| 734 |
+
"grad_norm": 1.560695472946775,
|
| 735 |
+
"learning_rate": 2.759237187127533e-07,
|
| 736 |
+
"logits/chosen": -3.296875,
|
| 737 |
+
"logits/rejected": -3.46875,
|
| 738 |
+
"logps/chosen": -103.0,
|
| 739 |
+
"logps/rejected": -294.0,
|
| 740 |
+
"loss": 0.0296,
|
| 741 |
+
"rewards/accuracies": 0.987500011920929,
|
| 742 |
+
"rewards/chosen": 0.91015625,
|
| 743 |
+
"rewards/margins": 10.125,
|
| 744 |
+
"rewards/rejected": -9.25,
|
| 745 |
+
"step": 470
|
| 746 |
+
},
|
| 747 |
+
{
|
| 748 |
+
"epoch": 1.5434083601286175,
|
| 749 |
+
"grad_norm": 4.044291002647325,
|
| 750 |
+
"learning_rate": 2.699642431466031e-07,
|
| 751 |
+
"logits/chosen": -3.40625,
|
| 752 |
+
"logits/rejected": -3.625,
|
| 753 |
+
"logps/chosen": -119.0,
|
| 754 |
+
"logps/rejected": -292.0,
|
| 755 |
+
"loss": 0.0338,
|
| 756 |
+
"rewards/accuracies": 0.987500011920929,
|
| 757 |
+
"rewards/chosen": 0.72265625,
|
| 758 |
+
"rewards/margins": 10.125,
|
| 759 |
+
"rewards/rejected": -9.4375,
|
| 760 |
+
"step": 480
|
| 761 |
+
},
|
| 762 |
+
{
|
| 763 |
+
"epoch": 1.5755627009646302,
|
| 764 |
+
"grad_norm": 52.234565040636085,
|
| 765 |
+
"learning_rate": 2.640047675804529e-07,
|
| 766 |
+
"logits/chosen": -3.421875,
|
| 767 |
+
"logits/rejected": -3.53125,
|
| 768 |
+
"logps/chosen": -133.0,
|
| 769 |
+
"logps/rejected": -296.0,
|
| 770 |
+
"loss": 0.0613,
|
| 771 |
+
"rewards/accuracies": 0.949999988079071,
|
| 772 |
+
"rewards/chosen": 0.9375,
|
| 773 |
+
"rewards/margins": 9.3125,
|
| 774 |
+
"rewards/rejected": -8.375,
|
| 775 |
+
"step": 490
|
| 776 |
+
},
|
| 777 |
+
{
|
| 778 |
+
"epoch": 1.607717041800643,
|
| 779 |
+
"grad_norm": 3.80360585724896,
|
| 780 |
+
"learning_rate": 2.5804529201430277e-07,
|
| 781 |
+
"logits/chosen": -3.34375,
|
| 782 |
+
"logits/rejected": -3.578125,
|
| 783 |
+
"logps/chosen": -128.0,
|
| 784 |
+
"logps/rejected": -300.0,
|
| 785 |
+
"loss": 0.1382,
|
| 786 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 787 |
+
"rewards/chosen": 0.16796875,
|
| 788 |
+
"rewards/margins": 8.9375,
|
| 789 |
+
"rewards/rejected": -8.8125,
|
| 790 |
+
"step": 500
|
| 791 |
+
},
|
| 792 |
+
{
|
| 793 |
+
"epoch": 1.639871382636656,
|
| 794 |
+
"grad_norm": 2.651977727973715,
|
| 795 |
+
"learning_rate": 2.520858164481526e-07,
|
| 796 |
+
"logits/chosen": -3.421875,
|
| 797 |
+
"logits/rejected": -3.65625,
|
| 798 |
+
"logps/chosen": -119.0,
|
| 799 |
+
"logps/rejected": -276.0,
|
| 800 |
+
"loss": 0.0504,
|
| 801 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 802 |
+
"rewards/chosen": 0.37890625,
|
| 803 |
+
"rewards/margins": 8.6875,
|
| 804 |
+
"rewards/rejected": -8.3125,
|
| 805 |
+
"step": 510
|
| 806 |
+
},
|
| 807 |
+
{
|
| 808 |
+
"epoch": 1.6720257234726688,
|
| 809 |
+
"grad_norm": 28.125055557089983,
|
| 810 |
+
"learning_rate": 2.461263408820024e-07,
|
| 811 |
+
"logits/chosen": -3.34375,
|
| 812 |
+
"logits/rejected": -3.53125,
|
| 813 |
+
"logps/chosen": -100.0,
|
| 814 |
+
"logps/rejected": -276.0,
|
| 815 |
+
"loss": 0.0269,
|
| 816 |
+
"rewards/accuracies": 0.987500011920929,
|
| 817 |
+
"rewards/chosen": 1.5390625,
|
| 818 |
+
"rewards/margins": 10.25,
|
| 819 |
+
"rewards/rejected": -8.75,
|
| 820 |
+
"step": 520
|
| 821 |
+
},
|
| 822 |
+
{
|
| 823 |
+
"epoch": 1.7041800643086815,
|
| 824 |
+
"grad_norm": 20.361634972075212,
|
| 825 |
+
"learning_rate": 2.401668653158522e-07,
|
| 826 |
+
"logits/chosen": -3.4375,
|
| 827 |
+
"logits/rejected": -3.59375,
|
| 828 |
+
"logps/chosen": -135.0,
|
| 829 |
+
"logps/rejected": -310.0,
|
| 830 |
+
"loss": 0.0232,
|
| 831 |
+
"rewards/accuracies": 0.987500011920929,
|
| 832 |
+
"rewards/chosen": 0.75390625,
|
| 833 |
+
"rewards/margins": 10.875,
|
| 834 |
+
"rewards/rejected": -10.125,
|
| 835 |
+
"step": 530
|
| 836 |
+
},
|
| 837 |
+
{
|
| 838 |
+
"epoch": 1.7363344051446945,
|
| 839 |
+
"grad_norm": 29.37467125959926,
|
| 840 |
+
"learning_rate": 2.3420738974970201e-07,
|
| 841 |
+
"logits/chosen": -3.578125,
|
| 842 |
+
"logits/rejected": -3.734375,
|
| 843 |
+
"logps/chosen": -130.0,
|
| 844 |
+
"logps/rejected": -300.0,
|
| 845 |
+
"loss": 0.0688,
|
| 846 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 847 |
+
"rewards/chosen": -0.859375,
|
| 848 |
+
"rewards/margins": 9.0625,
|
| 849 |
+
"rewards/rejected": -9.9375,
|
| 850 |
+
"step": 540
|
| 851 |
+
},
|
| 852 |
+
{
|
| 853 |
+
"epoch": 1.7684887459807075,
|
| 854 |
+
"grad_norm": 16.99690727375027,
|
| 855 |
+
"learning_rate": 2.2824791418355183e-07,
|
| 856 |
+
"logits/chosen": -3.453125,
|
| 857 |
+
"logits/rejected": -3.6875,
|
| 858 |
+
"logps/chosen": -145.0,
|
| 859 |
+
"logps/rejected": -312.0,
|
| 860 |
+
"loss": 0.0231,
|
| 861 |
+
"rewards/accuracies": 1.0,
|
| 862 |
+
"rewards/chosen": -0.048828125,
|
| 863 |
+
"rewards/margins": 9.5,
|
| 864 |
+
"rewards/rejected": -9.5,
|
| 865 |
+
"step": 550
|
| 866 |
+
},
|
| 867 |
+
{
|
| 868 |
+
"epoch": 1.8006430868167203,
|
| 869 |
+
"grad_norm": 49.05866215948682,
|
| 870 |
+
"learning_rate": 2.2228843861740164e-07,
|
| 871 |
+
"logits/chosen": -3.4375,
|
| 872 |
+
"logits/rejected": -3.609375,
|
| 873 |
+
"logps/chosen": -126.5,
|
| 874 |
+
"logps/rejected": -284.0,
|
| 875 |
+
"loss": 0.0162,
|
| 876 |
+
"rewards/accuracies": 1.0,
|
| 877 |
+
"rewards/chosen": 0.453125,
|
| 878 |
+
"rewards/margins": 10.0,
|
| 879 |
+
"rewards/rejected": -9.5625,
|
| 880 |
+
"step": 560
|
| 881 |
+
},
|
| 882 |
+
{
|
| 883 |
+
"epoch": 1.832797427652733,
|
| 884 |
+
"grad_norm": 17.76760727253399,
|
| 885 |
+
"learning_rate": 2.1632896305125148e-07,
|
| 886 |
+
"logits/chosen": -3.515625,
|
| 887 |
+
"logits/rejected": -3.71875,
|
| 888 |
+
"logps/chosen": -126.5,
|
| 889 |
+
"logps/rejected": -298.0,
|
| 890 |
+
"loss": 0.0305,
|
| 891 |
+
"rewards/accuracies": 0.987500011920929,
|
| 892 |
+
"rewards/chosen": 0.3125,
|
| 893 |
+
"rewards/margins": 10.1875,
|
| 894 |
+
"rewards/rejected": -9.875,
|
| 895 |
+
"step": 570
|
| 896 |
+
},
|
| 897 |
+
{
|
| 898 |
+
"epoch": 1.864951768488746,
|
| 899 |
+
"grad_norm": 6.0736731293937,
|
| 900 |
+
"learning_rate": 2.1036948748510132e-07,
|
| 901 |
+
"logits/chosen": -3.4375,
|
| 902 |
+
"logits/rejected": -3.5625,
|
| 903 |
+
"logps/chosen": -127.0,
|
| 904 |
+
"logps/rejected": -312.0,
|
| 905 |
+
"loss": 0.054,
|
| 906 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 907 |
+
"rewards/chosen": -0.1064453125,
|
| 908 |
+
"rewards/margins": 10.25,
|
| 909 |
+
"rewards/rejected": -10.375,
|
| 910 |
+
"step": 580
|
| 911 |
+
},
|
| 912 |
+
{
|
| 913 |
+
"epoch": 1.897106109324759,
|
| 914 |
+
"grad_norm": 39.17766360963206,
|
| 915 |
+
"learning_rate": 2.0441001191895113e-07,
|
| 916 |
+
"logits/chosen": -3.59375,
|
| 917 |
+
"logits/rejected": -3.796875,
|
| 918 |
+
"logps/chosen": -137.0,
|
| 919 |
+
"logps/rejected": -300.0,
|
| 920 |
+
"loss": 0.0353,
|
| 921 |
+
"rewards/accuracies": 0.987500011920929,
|
| 922 |
+
"rewards/chosen": -0.1953125,
|
| 923 |
+
"rewards/margins": 9.8125,
|
| 924 |
+
"rewards/rejected": -10.0,
|
| 925 |
+
"step": 590
|
| 926 |
+
},
|
| 927 |
+
{
|
| 928 |
+
"epoch": 1.9292604501607717,
|
| 929 |
+
"grad_norm": 38.14836445184288,
|
| 930 |
+
"learning_rate": 1.9845053635280097e-07,
|
| 931 |
+
"logits/chosen": -3.34375,
|
| 932 |
+
"logits/rejected": -3.53125,
|
| 933 |
+
"logps/chosen": -138.0,
|
| 934 |
+
"logps/rejected": -324.0,
|
| 935 |
+
"loss": 0.0299,
|
| 936 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 937 |
+
"rewards/chosen": 0.458984375,
|
| 938 |
+
"rewards/margins": 11.0625,
|
| 939 |
+
"rewards/rejected": -10.625,
|
| 940 |
+
"step": 600
|
| 941 |
+
},
|
| 942 |
+
{
|
| 943 |
+
"epoch": 1.9614147909967845,
|
| 944 |
+
"grad_norm": 55.90659543384472,
|
| 945 |
+
"learning_rate": 1.9249106078665075e-07,
|
| 946 |
+
"logits/chosen": -3.3125,
|
| 947 |
+
"logits/rejected": -3.546875,
|
| 948 |
+
"logps/chosen": -113.0,
|
| 949 |
+
"logps/rejected": -272.0,
|
| 950 |
+
"loss": 0.0322,
|
| 951 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 952 |
+
"rewards/chosen": 1.09375,
|
| 953 |
+
"rewards/margins": 9.375,
|
| 954 |
+
"rewards/rejected": -8.25,
|
| 955 |
+
"step": 610
|
| 956 |
+
},
|
| 957 |
+
{
|
| 958 |
+
"epoch": 1.9935691318327975,
|
| 959 |
+
"grad_norm": 0.34892216328376285,
|
| 960 |
+
"learning_rate": 1.865315852205006e-07,
|
| 961 |
+
"logits/chosen": -3.296875,
|
| 962 |
+
"logits/rejected": -3.46875,
|
| 963 |
+
"logps/chosen": -150.0,
|
| 964 |
+
"logps/rejected": -334.0,
|
| 965 |
+
"loss": 0.018,
|
| 966 |
+
"rewards/accuracies": 1.0,
|
| 967 |
+
"rewards/chosen": 0.484375,
|
| 968 |
+
"rewards/margins": 10.9375,
|
| 969 |
+
"rewards/rejected": -10.4375,
|
| 970 |
+
"step": 620
|
| 971 |
+
},
|
| 972 |
+
{
|
| 973 |
+
"epoch": 2.0,
|
| 974 |
+
"eval_logits/chosen": -3.65625,
|
| 975 |
+
"eval_logits/rejected": -3.734375,
|
| 976 |
+
"eval_logps/chosen": -226.0,
|
| 977 |
+
"eval_logps/rejected": -266.0,
|
| 978 |
+
"eval_loss": 0.9209327697753906,
|
| 979 |
+
"eval_rewards/accuracies": 0.7142857313156128,
|
| 980 |
+
"eval_rewards/chosen": -4.09375,
|
| 981 |
+
"eval_rewards/margins": 3.375,
|
| 982 |
+
"eval_rewards/rejected": -7.46875,
|
| 983 |
+
"eval_runtime": 14.1621,
|
| 984 |
+
"eval_samples_per_second": 14.052,
|
| 985 |
+
"eval_steps_per_second": 0.494,
|
| 986 |
+
"step": 622
|
| 987 |
+
},
|
| 988 |
+
{
|
| 989 |
+
"epoch": 2.0257234726688105,
|
| 990 |
+
"grad_norm": 1.0084688310317689,
|
| 991 |
+
"learning_rate": 1.805721096543504e-07,
|
| 992 |
+
"logits/chosen": -3.265625,
|
| 993 |
+
"logits/rejected": -3.53125,
|
| 994 |
+
"logps/chosen": -138.0,
|
| 995 |
+
"logps/rejected": -296.0,
|
| 996 |
+
"loss": 0.0133,
|
| 997 |
+
"rewards/accuracies": 1.0,
|
| 998 |
+
"rewards/chosen": 0.82421875,
|
| 999 |
+
"rewards/margins": 10.125,
|
| 1000 |
+
"rewards/rejected": -9.3125,
|
| 1001 |
+
"step": 630
|
| 1002 |
+
},
|
| 1003 |
+
{
|
| 1004 |
+
"epoch": 2.057877813504823,
|
| 1005 |
+
"grad_norm": 1.0564688005496572,
|
| 1006 |
+
"learning_rate": 1.7461263408820024e-07,
|
| 1007 |
+
"logits/chosen": -3.28125,
|
| 1008 |
+
"logits/rejected": -3.46875,
|
| 1009 |
+
"logps/chosen": -123.5,
|
| 1010 |
+
"logps/rejected": -278.0,
|
| 1011 |
+
"loss": 0.0128,
|
| 1012 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1013 |
+
"rewards/chosen": 0.5,
|
| 1014 |
+
"rewards/margins": 10.0,
|
| 1015 |
+
"rewards/rejected": -9.5,
|
| 1016 |
+
"step": 640
|
| 1017 |
+
},
|
| 1018 |
+
{
|
| 1019 |
+
"epoch": 2.090032154340836,
|
| 1020 |
+
"grad_norm": 2.814703376282905,
|
| 1021 |
+
"learning_rate": 1.6865315852205006e-07,
|
| 1022 |
+
"logits/chosen": -3.171875,
|
| 1023 |
+
"logits/rejected": -3.390625,
|
| 1024 |
+
"logps/chosen": -122.5,
|
| 1025 |
+
"logps/rejected": -330.0,
|
| 1026 |
+
"loss": 0.0096,
|
| 1027 |
+
"rewards/accuracies": 1.0,
|
| 1028 |
+
"rewards/chosen": 1.4296875,
|
| 1029 |
+
"rewards/margins": 13.4375,
|
| 1030 |
+
"rewards/rejected": -12.0,
|
| 1031 |
+
"step": 650
|
| 1032 |
+
},
|
| 1033 |
+
{
|
| 1034 |
+
"epoch": 2.122186495176849,
|
| 1035 |
+
"grad_norm": 9.832993297771846,
|
| 1036 |
+
"learning_rate": 1.626936829558999e-07,
|
| 1037 |
+
"logits/chosen": -3.25,
|
| 1038 |
+
"logits/rejected": -3.53125,
|
| 1039 |
+
"logps/chosen": -138.0,
|
| 1040 |
+
"logps/rejected": -330.0,
|
| 1041 |
+
"loss": 0.0111,
|
| 1042 |
+
"rewards/accuracies": 1.0,
|
| 1043 |
+
"rewards/chosen": 0.1591796875,
|
| 1044 |
+
"rewards/margins": 12.3125,
|
| 1045 |
+
"rewards/rejected": -12.1875,
|
| 1046 |
+
"step": 660
|
| 1047 |
+
},
|
| 1048 |
+
{
|
| 1049 |
+
"epoch": 2.154340836012862,
|
| 1050 |
+
"grad_norm": 0.21098065783878342,
|
| 1051 |
+
"learning_rate": 1.5673420738974968e-07,
|
| 1052 |
+
"logits/chosen": -3.234375,
|
| 1053 |
+
"logits/rejected": -3.34375,
|
| 1054 |
+
"logps/chosen": -109.0,
|
| 1055 |
+
"logps/rejected": -276.0,
|
| 1056 |
+
"loss": 0.0034,
|
| 1057 |
+
"rewards/accuracies": 1.0,
|
| 1058 |
+
"rewards/chosen": 1.03125,
|
| 1059 |
+
"rewards/margins": 12.3125,
|
| 1060 |
+
"rewards/rejected": -11.3125,
|
| 1061 |
+
"step": 670
|
| 1062 |
+
},
|
| 1063 |
+
{
|
| 1064 |
+
"epoch": 2.1864951768488745,
|
| 1065 |
+
"grad_norm": 2.476443587212778,
|
| 1066 |
+
"learning_rate": 1.5077473182359952e-07,
|
| 1067 |
+
"logits/chosen": -3.09375,
|
| 1068 |
+
"logits/rejected": -3.375,
|
| 1069 |
+
"logps/chosen": -106.5,
|
| 1070 |
+
"logps/rejected": -320.0,
|
| 1071 |
+
"loss": 0.0043,
|
| 1072 |
+
"rewards/accuracies": 1.0,
|
| 1073 |
+
"rewards/chosen": 0.546875,
|
| 1074 |
+
"rewards/margins": 12.75,
|
| 1075 |
+
"rewards/rejected": -12.1875,
|
| 1076 |
+
"step": 680
|
| 1077 |
+
},
|
| 1078 |
+
{
|
| 1079 |
+
"epoch": 2.2186495176848875,
|
| 1080 |
+
"grad_norm": 11.363132455431685,
|
| 1081 |
+
"learning_rate": 1.4481525625744933e-07,
|
| 1082 |
+
"logits/chosen": -3.125,
|
| 1083 |
+
"logits/rejected": -3.40625,
|
| 1084 |
+
"logps/chosen": -134.0,
|
| 1085 |
+
"logps/rejected": -348.0,
|
| 1086 |
+
"loss": 0.0122,
|
| 1087 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1088 |
+
"rewards/chosen": 0.7421875,
|
| 1089 |
+
"rewards/margins": 12.9375,
|
| 1090 |
+
"rewards/rejected": -12.1875,
|
| 1091 |
+
"step": 690
|
| 1092 |
+
},
|
| 1093 |
+
{
|
| 1094 |
+
"epoch": 2.2508038585209005,
|
| 1095 |
+
"grad_norm": 2.5108344224003334,
|
| 1096 |
+
"learning_rate": 1.3885578069129917e-07,
|
| 1097 |
+
"logits/chosen": -3.203125,
|
| 1098 |
+
"logits/rejected": -3.5,
|
| 1099 |
+
"logps/chosen": -123.5,
|
| 1100 |
+
"logps/rejected": -318.0,
|
| 1101 |
+
"loss": 0.0063,
|
| 1102 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1103 |
+
"rewards/chosen": 0.13671875,
|
| 1104 |
+
"rewards/margins": 12.5625,
|
| 1105 |
+
"rewards/rejected": -12.375,
|
| 1106 |
+
"step": 700
|
| 1107 |
+
},
|
| 1108 |
+
{
|
| 1109 |
+
"epoch": 2.282958199356913,
|
| 1110 |
+
"grad_norm": 0.16002189769618574,
|
| 1111 |
+
"learning_rate": 1.3289630512514898e-07,
|
| 1112 |
+
"logits/chosen": -3.0625,
|
| 1113 |
+
"logits/rejected": -3.28125,
|
| 1114 |
+
"logps/chosen": -119.0,
|
| 1115 |
+
"logps/rejected": -340.0,
|
| 1116 |
+
"loss": 0.0073,
|
| 1117 |
+
"rewards/accuracies": 1.0,
|
| 1118 |
+
"rewards/chosen": 0.96484375,
|
| 1119 |
+
"rewards/margins": 13.3125,
|
| 1120 |
+
"rewards/rejected": -12.3125,
|
| 1121 |
+
"step": 710
|
| 1122 |
+
},
|
| 1123 |
+
{
|
| 1124 |
+
"epoch": 2.315112540192926,
|
| 1125 |
+
"grad_norm": 0.24278926553916283,
|
| 1126 |
+
"learning_rate": 1.2693682955899882e-07,
|
| 1127 |
+
"logits/chosen": -3.125,
|
| 1128 |
+
"logits/rejected": -3.296875,
|
| 1129 |
+
"logps/chosen": -136.0,
|
| 1130 |
+
"logps/rejected": -342.0,
|
| 1131 |
+
"loss": 0.001,
|
| 1132 |
+
"rewards/accuracies": 1.0,
|
| 1133 |
+
"rewards/chosen": -0.1201171875,
|
| 1134 |
+
"rewards/margins": 12.875,
|
| 1135 |
+
"rewards/rejected": -13.0,
|
| 1136 |
+
"step": 720
|
| 1137 |
+
},
|
| 1138 |
+
{
|
| 1139 |
+
"epoch": 2.347266881028939,
|
| 1140 |
+
"grad_norm": 8.462048605522668,
|
| 1141 |
+
"learning_rate": 1.2097735399284863e-07,
|
| 1142 |
+
"logits/chosen": -3.15625,
|
| 1143 |
+
"logits/rejected": -3.328125,
|
| 1144 |
+
"logps/chosen": -114.5,
|
| 1145 |
+
"logps/rejected": -298.0,
|
| 1146 |
+
"loss": 0.0085,
|
| 1147 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1148 |
+
"rewards/chosen": 0.5546875,
|
| 1149 |
+
"rewards/margins": 12.125,
|
| 1150 |
+
"rewards/rejected": -11.5625,
|
| 1151 |
+
"step": 730
|
| 1152 |
+
},
|
| 1153 |
+
{
|
| 1154 |
+
"epoch": 2.379421221864952,
|
| 1155 |
+
"grad_norm": 0.2836935848609144,
|
| 1156 |
+
"learning_rate": 1.1501787842669844e-07,
|
| 1157 |
+
"logits/chosen": -3.1875,
|
| 1158 |
+
"logits/rejected": -3.4375,
|
| 1159 |
+
"logps/chosen": -149.0,
|
| 1160 |
+
"logps/rejected": -342.0,
|
| 1161 |
+
"loss": 0.0095,
|
| 1162 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1163 |
+
"rewards/chosen": 0.388671875,
|
| 1164 |
+
"rewards/margins": 12.625,
|
| 1165 |
+
"rewards/rejected": -12.25,
|
| 1166 |
+
"step": 740
|
| 1167 |
+
},
|
| 1168 |
+
{
|
| 1169 |
+
"epoch": 2.4115755627009645,
|
| 1170 |
+
"grad_norm": 0.20905746208537096,
|
| 1171 |
+
"learning_rate": 1.0905840286054827e-07,
|
| 1172 |
+
"logits/chosen": -3.21875,
|
| 1173 |
+
"logits/rejected": -3.4375,
|
| 1174 |
+
"logps/chosen": -144.0,
|
| 1175 |
+
"logps/rejected": -336.0,
|
| 1176 |
+
"loss": 0.0024,
|
| 1177 |
+
"rewards/accuracies": 1.0,
|
| 1178 |
+
"rewards/chosen": 0.66015625,
|
| 1179 |
+
"rewards/margins": 13.125,
|
| 1180 |
+
"rewards/rejected": -12.5,
|
| 1181 |
+
"step": 750
|
| 1182 |
+
},
|
| 1183 |
+
{
|
| 1184 |
+
"epoch": 2.4437299035369775,
|
| 1185 |
+
"grad_norm": 0.13532730518065741,
|
| 1186 |
+
"learning_rate": 1.030989272943981e-07,
|
| 1187 |
+
"logits/chosen": -3.15625,
|
| 1188 |
+
"logits/rejected": -3.4375,
|
| 1189 |
+
"logps/chosen": -135.0,
|
| 1190 |
+
"logps/rejected": -340.0,
|
| 1191 |
+
"loss": 0.0052,
|
| 1192 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1193 |
+
"rewards/chosen": 0.07666015625,
|
| 1194 |
+
"rewards/margins": 12.9375,
|
| 1195 |
+
"rewards/rejected": -12.875,
|
| 1196 |
+
"step": 760
|
| 1197 |
+
},
|
| 1198 |
+
{
|
| 1199 |
+
"epoch": 2.4758842443729905,
|
| 1200 |
+
"grad_norm": 1.0359164426855658,
|
| 1201 |
+
"learning_rate": 9.713945172824791e-08,
|
| 1202 |
+
"logits/chosen": -3.09375,
|
| 1203 |
+
"logits/rejected": -3.40625,
|
| 1204 |
+
"logps/chosen": -124.5,
|
| 1205 |
+
"logps/rejected": -330.0,
|
| 1206 |
+
"loss": 0.0052,
|
| 1207 |
+
"rewards/accuracies": 1.0,
|
| 1208 |
+
"rewards/chosen": -0.034912109375,
|
| 1209 |
+
"rewards/margins": 13.3125,
|
| 1210 |
+
"rewards/rejected": -13.375,
|
| 1211 |
+
"step": 770
|
| 1212 |
+
},
|
| 1213 |
+
{
|
| 1214 |
+
"epoch": 2.508038585209003,
|
| 1215 |
+
"grad_norm": 0.02411362444063136,
|
| 1216 |
+
"learning_rate": 9.117997616209773e-08,
|
| 1217 |
+
"logits/chosen": -3.140625,
|
| 1218 |
+
"logits/rejected": -3.390625,
|
| 1219 |
+
"logps/chosen": -146.0,
|
| 1220 |
+
"logps/rejected": -356.0,
|
| 1221 |
+
"loss": 0.0025,
|
| 1222 |
+
"rewards/accuracies": 1.0,
|
| 1223 |
+
"rewards/chosen": 0.3046875,
|
| 1224 |
+
"rewards/margins": 13.75,
|
| 1225 |
+
"rewards/rejected": -13.5,
|
| 1226 |
+
"step": 780
|
| 1227 |
+
},
|
| 1228 |
+
{
|
| 1229 |
+
"epoch": 2.540192926045016,
|
| 1230 |
+
"grad_norm": 0.14361577541299492,
|
| 1231 |
+
"learning_rate": 8.522050059594756e-08,
|
| 1232 |
+
"logits/chosen": -3.15625,
|
| 1233 |
+
"logits/rejected": -3.40625,
|
| 1234 |
+
"logps/chosen": -148.0,
|
| 1235 |
+
"logps/rejected": -340.0,
|
| 1236 |
+
"loss": 0.0008,
|
| 1237 |
+
"rewards/accuracies": 1.0,
|
| 1238 |
+
"rewards/chosen": 0.2734375,
|
| 1239 |
+
"rewards/margins": 13.5,
|
| 1240 |
+
"rewards/rejected": -13.25,
|
| 1241 |
+
"step": 790
|
| 1242 |
+
},
|
| 1243 |
+
{
|
| 1244 |
+
"epoch": 2.572347266881029,
|
| 1245 |
+
"grad_norm": 3.7709142939626923,
|
| 1246 |
+
"learning_rate": 7.926102502979737e-08,
|
| 1247 |
+
"logits/chosen": -3.171875,
|
| 1248 |
+
"logits/rejected": -3.40625,
|
| 1249 |
+
"logps/chosen": -126.0,
|
| 1250 |
+
"logps/rejected": -344.0,
|
| 1251 |
+
"loss": 0.0013,
|
| 1252 |
+
"rewards/accuracies": 1.0,
|
| 1253 |
+
"rewards/chosen": 0.62890625,
|
| 1254 |
+
"rewards/margins": 13.6875,
|
| 1255 |
+
"rewards/rejected": -13.0625,
|
| 1256 |
+
"step": 800
|
| 1257 |
+
},
|
| 1258 |
+
{
|
| 1259 |
+
"epoch": 2.604501607717042,
|
| 1260 |
+
"grad_norm": 0.6815929170420622,
|
| 1261 |
+
"learning_rate": 7.33015494636472e-08,
|
| 1262 |
+
"logits/chosen": -3.171875,
|
| 1263 |
+
"logits/rejected": -3.421875,
|
| 1264 |
+
"logps/chosen": -122.0,
|
| 1265 |
+
"logps/rejected": -306.0,
|
| 1266 |
+
"loss": 0.0094,
|
| 1267 |
+
"rewards/accuracies": 1.0,
|
| 1268 |
+
"rewards/chosen": 0.006866455078125,
|
| 1269 |
+
"rewards/margins": 12.75,
|
| 1270 |
+
"rewards/rejected": -12.75,
|
| 1271 |
+
"step": 810
|
| 1272 |
+
},
|
| 1273 |
+
{
|
| 1274 |
+
"epoch": 2.6366559485530545,
|
| 1275 |
+
"grad_norm": 1.1968332787733935,
|
| 1276 |
+
"learning_rate": 6.734207389749702e-08,
|
| 1277 |
+
"logits/chosen": -3.09375,
|
| 1278 |
+
"logits/rejected": -3.4375,
|
| 1279 |
+
"logps/chosen": -110.0,
|
| 1280 |
+
"logps/rejected": -320.0,
|
| 1281 |
+
"loss": 0.0075,
|
| 1282 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1283 |
+
"rewards/chosen": 0.94140625,
|
| 1284 |
+
"rewards/margins": 13.75,
|
| 1285 |
+
"rewards/rejected": -12.75,
|
| 1286 |
+
"step": 820
|
| 1287 |
+
},
|
| 1288 |
+
{
|
| 1289 |
+
"epoch": 2.6688102893890675,
|
| 1290 |
+
"grad_norm": 0.2654328339493102,
|
| 1291 |
+
"learning_rate": 6.138259833134683e-08,
|
| 1292 |
+
"logits/chosen": -3.15625,
|
| 1293 |
+
"logits/rejected": -3.484375,
|
| 1294 |
+
"logps/chosen": -127.5,
|
| 1295 |
+
"logps/rejected": -328.0,
|
| 1296 |
+
"loss": 0.0081,
|
| 1297 |
+
"rewards/accuracies": 1.0,
|
| 1298 |
+
"rewards/chosen": 0.453125,
|
| 1299 |
+
"rewards/margins": 13.9375,
|
| 1300 |
+
"rewards/rejected": -13.5,
|
| 1301 |
+
"step": 830
|
| 1302 |
+
},
|
| 1303 |
+
{
|
| 1304 |
+
"epoch": 2.7009646302250805,
|
| 1305 |
+
"grad_norm": 0.2744508216663156,
|
| 1306 |
+
"learning_rate": 5.542312276519666e-08,
|
| 1307 |
+
"logits/chosen": -3.140625,
|
| 1308 |
+
"logits/rejected": -3.375,
|
| 1309 |
+
"logps/chosen": -119.5,
|
| 1310 |
+
"logps/rejected": -346.0,
|
| 1311 |
+
"loss": 0.005,
|
| 1312 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1313 |
+
"rewards/chosen": 0.10107421875,
|
| 1314 |
+
"rewards/margins": 13.125,
|
| 1315 |
+
"rewards/rejected": -13.0625,
|
| 1316 |
+
"step": 840
|
| 1317 |
+
},
|
| 1318 |
+
{
|
| 1319 |
+
"epoch": 2.733118971061093,
|
| 1320 |
+
"grad_norm": 0.03947959588228437,
|
| 1321 |
+
"learning_rate": 4.9463647199046485e-08,
|
| 1322 |
+
"logits/chosen": -3.203125,
|
| 1323 |
+
"logits/rejected": -3.34375,
|
| 1324 |
+
"logps/chosen": -120.5,
|
| 1325 |
+
"logps/rejected": -316.0,
|
| 1326 |
+
"loss": 0.0073,
|
| 1327 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1328 |
+
"rewards/chosen": 0.224609375,
|
| 1329 |
+
"rewards/margins": 12.875,
|
| 1330 |
+
"rewards/rejected": -12.6875,
|
| 1331 |
+
"step": 850
|
| 1332 |
+
},
|
| 1333 |
+
{
|
| 1334 |
+
"epoch": 2.765273311897106,
|
| 1335 |
+
"grad_norm": 0.6863210482116098,
|
| 1336 |
+
"learning_rate": 4.3504171632896303e-08,
|
| 1337 |
+
"logits/chosen": -3.25,
|
| 1338 |
+
"logits/rejected": -3.5,
|
| 1339 |
+
"logps/chosen": -136.0,
|
| 1340 |
+
"logps/rejected": -354.0,
|
| 1341 |
+
"loss": 0.0105,
|
| 1342 |
+
"rewards/accuracies": 1.0,
|
| 1343 |
+
"rewards/chosen": 0.59375,
|
| 1344 |
+
"rewards/margins": 13.75,
|
| 1345 |
+
"rewards/rejected": -13.125,
|
| 1346 |
+
"step": 860
|
| 1347 |
+
},
|
| 1348 |
+
{
|
| 1349 |
+
"epoch": 2.797427652733119,
|
| 1350 |
+
"grad_norm": 2.8280463279363106,
|
| 1351 |
+
"learning_rate": 3.754469606674612e-08,
|
| 1352 |
+
"logits/chosen": -3.0625,
|
| 1353 |
+
"logits/rejected": -3.390625,
|
| 1354 |
+
"logps/chosen": -116.5,
|
| 1355 |
+
"logps/rejected": -318.0,
|
| 1356 |
+
"loss": 0.0078,
|
| 1357 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1358 |
+
"rewards/chosen": 0.255859375,
|
| 1359 |
+
"rewards/margins": 13.25,
|
| 1360 |
+
"rewards/rejected": -13.0,
|
| 1361 |
+
"step": 870
|
| 1362 |
+
},
|
| 1363 |
+
{
|
| 1364 |
+
"epoch": 2.829581993569132,
|
| 1365 |
+
"grad_norm": 4.677533654072525,
|
| 1366 |
+
"learning_rate": 3.158522050059595e-08,
|
| 1367 |
+
"logits/chosen": -3.21875,
|
| 1368 |
+
"logits/rejected": -3.421875,
|
| 1369 |
+
"logps/chosen": -147.0,
|
| 1370 |
+
"logps/rejected": -312.0,
|
| 1371 |
+
"loss": 0.0227,
|
| 1372 |
+
"rewards/accuracies": 0.987500011920929,
|
| 1373 |
+
"rewards/chosen": 0.1044921875,
|
| 1374 |
+
"rewards/margins": 12.375,
|
| 1375 |
+
"rewards/rejected": -12.25,
|
| 1376 |
+
"step": 880
|
| 1377 |
+
},
|
| 1378 |
+
{
|
| 1379 |
+
"epoch": 2.861736334405145,
|
| 1380 |
+
"grad_norm": 4.075516801733505,
|
| 1381 |
+
"learning_rate": 2.562574493444577e-08,
|
| 1382 |
+
"logits/chosen": -3.140625,
|
| 1383 |
+
"logits/rejected": -3.390625,
|
| 1384 |
+
"logps/chosen": -130.0,
|
| 1385 |
+
"logps/rejected": -328.0,
|
| 1386 |
+
"loss": 0.0112,
|
| 1387 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1388 |
+
"rewards/chosen": 0.1201171875,
|
| 1389 |
+
"rewards/margins": 13.125,
|
| 1390 |
+
"rewards/rejected": -13.0,
|
| 1391 |
+
"step": 890
|
| 1392 |
+
},
|
| 1393 |
+
{
|
| 1394 |
+
"epoch": 2.8938906752411575,
|
| 1395 |
+
"grad_norm": 0.31091685277937586,
|
| 1396 |
+
"learning_rate": 1.966626936829559e-08,
|
| 1397 |
+
"logits/chosen": -3.171875,
|
| 1398 |
+
"logits/rejected": -3.453125,
|
| 1399 |
+
"logps/chosen": -114.5,
|
| 1400 |
+
"logps/rejected": -316.0,
|
| 1401 |
+
"loss": 0.0116,
|
| 1402 |
+
"rewards/accuracies": 0.9750000238418579,
|
| 1403 |
+
"rewards/chosen": 0.498046875,
|
| 1404 |
+
"rewards/margins": 13.3125,
|
| 1405 |
+
"rewards/rejected": -12.8125,
|
| 1406 |
+
"step": 900
|
| 1407 |
+
},
|
| 1408 |
+
{
|
| 1409 |
+
"epoch": 2.9260450160771705,
|
| 1410 |
+
"grad_norm": 0.008132094709855322,
|
| 1411 |
+
"learning_rate": 1.370679380214541e-08,
|
| 1412 |
+
"logits/chosen": -3.125,
|
| 1413 |
+
"logits/rejected": -3.5,
|
| 1414 |
+
"logps/chosen": -124.0,
|
| 1415 |
+
"logps/rejected": -356.0,
|
| 1416 |
+
"loss": 0.0056,
|
| 1417 |
+
"rewards/accuracies": 1.0,
|
| 1418 |
+
"rewards/chosen": 0.0111083984375,
|
| 1419 |
+
"rewards/margins": 13.625,
|
| 1420 |
+
"rewards/rejected": -13.625,
|
| 1421 |
+
"step": 910
|
| 1422 |
+
},
|
| 1423 |
+
{
|
| 1424 |
+
"epoch": 2.958199356913183,
|
| 1425 |
+
"grad_norm": 10.508744046977588,
|
| 1426 |
+
"learning_rate": 7.747318235995233e-09,
|
| 1427 |
+
"logits/chosen": -3.28125,
|
| 1428 |
+
"logits/rejected": -3.5625,
|
| 1429 |
+
"logps/chosen": -115.0,
|
| 1430 |
+
"logps/rejected": -316.0,
|
| 1431 |
+
"loss": 0.0031,
|
| 1432 |
+
"rewards/accuracies": 1.0,
|
| 1433 |
+
"rewards/chosen": 0.1298828125,
|
| 1434 |
+
"rewards/margins": 13.125,
|
| 1435 |
+
"rewards/rejected": -13.0,
|
| 1436 |
+
"step": 920
|
| 1437 |
+
},
|
| 1438 |
+
{
|
| 1439 |
+
"epoch": 2.990353697749196,
|
| 1440 |
+
"grad_norm": 0.053140421396559286,
|
| 1441 |
+
"learning_rate": 1.7878426698450536e-09,
|
| 1442 |
+
"logits/chosen": -3.1875,
|
| 1443 |
+
"logits/rejected": -3.40625,
|
| 1444 |
+
"logps/chosen": -117.5,
|
| 1445 |
+
"logps/rejected": -308.0,
|
| 1446 |
+
"loss": 0.0051,
|
| 1447 |
+
"rewards/accuracies": 1.0,
|
| 1448 |
+
"rewards/chosen": 0.203125,
|
| 1449 |
+
"rewards/margins": 12.9375,
|
| 1450 |
+
"rewards/rejected": -12.75,
|
| 1451 |
+
"step": 930
|
| 1452 |
+
},
|
| 1453 |
+
{
|
| 1454 |
+
"epoch": 3.0,
|
| 1455 |
+
"eval_logits/chosen": -3.515625,
|
| 1456 |
+
"eval_logits/rejected": -3.59375,
|
| 1457 |
+
"eval_logps/chosen": -245.0,
|
| 1458 |
+
"eval_logps/rejected": -292.0,
|
| 1459 |
+
"eval_loss": 1.0541672706604004,
|
| 1460 |
+
"eval_rewards/accuracies": 0.6964285969734192,
|
| 1461 |
+
"eval_rewards/chosen": -5.96875,
|
| 1462 |
+
"eval_rewards/margins": 4.125,
|
| 1463 |
+
"eval_rewards/rejected": -10.0625,
|
| 1464 |
+
"eval_runtime": 15.892,
|
| 1465 |
+
"eval_samples_per_second": 12.522,
|
| 1466 |
+
"eval_steps_per_second": 0.44,
|
| 1467 |
+
"step": 933
|
| 1468 |
+
},
|
| 1469 |
+
{
|
| 1470 |
+
"epoch": 3.0,
|
| 1471 |
+
"step": 933,
|
| 1472 |
+
"total_flos": 0.0,
|
| 1473 |
+
"train_loss": 0.09964664405788767,
|
| 1474 |
+
"train_runtime": 5808.5282,
|
| 1475 |
+
"train_samples_per_second": 5.14,
|
| 1476 |
+
"train_steps_per_second": 0.161
|
| 1477 |
+
}
|
| 1478 |
+
],
|
| 1479 |
+
"logging_steps": 10,
|
| 1480 |
+
"max_steps": 933,
|
| 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 |
+
}
|