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-7B-v0.3-sft-ultrachat-hhrlhf
|
| 3 |
+
library_name: transformers
|
| 4 |
+
model_name: Mistral-7B-v0.3-sft-ultrachat-hhrlhf-dpo
|
| 5 |
+
tags:
|
| 6 |
+
- generated_from_trainer
|
| 7 |
+
- trl
|
| 8 |
+
- dpo
|
| 9 |
+
licence: license
|
| 10 |
+
---
|
| 11 |
+
|
| 12 |
+
# Model Card for Mistral-7B-v0.3-sft-ultrachat-hhrlhf-dpo
|
| 13 |
+
|
| 14 |
+
This model is a fine-tuned version of [AmberYifan/Mistral-7B-v0.3-sft-ultrachat-hhrlhf](https://huggingface.co/AmberYifan/Mistral-7B-v0.3-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="jyc0325/Mistral-7B-v0.3-sft-ultrachat-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 |
+
|
| 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: 2.1.2
|
| 39 |
+
- Datasets: 3.3.1
|
| 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": 0.5581134729111157,
|
| 5 |
+
"train_runtime": 5075.7144,
|
| 6 |
+
"train_samples": 10000,
|
| 7 |
+
"train_samples_per_second": 5.91,
|
| 8 |
+
"train_steps_per_second": 0.185
|
| 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": 0.5581134729111157,
|
| 5 |
+
"train_runtime": 5075.7144,
|
| 6 |
+
"train_samples": 10000,
|
| 7 |
+
"train_samples_per_second": 5.91,
|
| 8 |
+
"train_steps_per_second": 0.185
|
| 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": 939,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.003194888178913738,
|
| 13 |
+
"grad_norm": 22.587236912668185,
|
| 14 |
+
"learning_rate": 5.3191489361702125e-09,
|
| 15 |
+
"logits/chosen": -3.890625,
|
| 16 |
+
"logits/rejected": -3.890625,
|
| 17 |
+
"logps/chosen": -104.0,
|
| 18 |
+
"logps/rejected": -105.5,
|
| 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": 0.03194888178913738,
|
| 28 |
+
"grad_norm": 20.504214733036054,
|
| 29 |
+
"learning_rate": 5.3191489361702123e-08,
|
| 30 |
+
"logits/chosen": -3.78125,
|
| 31 |
+
"logits/rejected": -3.84375,
|
| 32 |
+
"logps/chosen": -112.5,
|
| 33 |
+
"logps/rejected": -127.5,
|
| 34 |
+
"loss": 0.6933,
|
| 35 |
+
"rewards/accuracies": 0.0416666679084301,
|
| 36 |
+
"rewards/chosen": -0.0045166015625,
|
| 37 |
+
"rewards/margins": -0.0020904541015625,
|
| 38 |
+
"rewards/rejected": -0.0024261474609375,
|
| 39 |
+
"step": 10
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.06389776357827476,
|
| 43 |
+
"grad_norm": 26.543460167609325,
|
| 44 |
+
"learning_rate": 1.0638297872340425e-07,
|
| 45 |
+
"logits/chosen": -3.890625,
|
| 46 |
+
"logits/rejected": -3.921875,
|
| 47 |
+
"logps/chosen": -131.0,
|
| 48 |
+
"logps/rejected": -141.0,
|
| 49 |
+
"loss": 0.6905,
|
| 50 |
+
"rewards/accuracies": 0.012500000186264515,
|
| 51 |
+
"rewards/chosen": 0.000308990478515625,
|
| 52 |
+
"rewards/margins": -0.00093841552734375,
|
| 53 |
+
"rewards/rejected": 0.001251220703125,
|
| 54 |
+
"step": 20
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.09584664536741214,
|
| 58 |
+
"grad_norm": 90.02360712811911,
|
| 59 |
+
"learning_rate": 1.5957446808510638e-07,
|
| 60 |
+
"logits/chosen": -3.859375,
|
| 61 |
+
"logits/rejected": -3.859375,
|
| 62 |
+
"logps/chosen": -126.5,
|
| 63 |
+
"logps/rejected": -132.0,
|
| 64 |
+
"loss": 0.6906,
|
| 65 |
+
"rewards/accuracies": 0.10000000149011612,
|
| 66 |
+
"rewards/chosen": 0.0128173828125,
|
| 67 |
+
"rewards/margins": 0.002197265625,
|
| 68 |
+
"rewards/rejected": 0.0106201171875,
|
| 69 |
+
"step": 30
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.12779552715654952,
|
| 73 |
+
"grad_norm": 17.123125909083377,
|
| 74 |
+
"learning_rate": 2.127659574468085e-07,
|
| 75 |
+
"logits/chosen": -3.875,
|
| 76 |
+
"logits/rejected": -3.859375,
|
| 77 |
+
"logps/chosen": -135.0,
|
| 78 |
+
"logps/rejected": -133.0,
|
| 79 |
+
"loss": 0.6896,
|
| 80 |
+
"rewards/accuracies": 0.13750000298023224,
|
| 81 |
+
"rewards/chosen": 0.034423828125,
|
| 82 |
+
"rewards/margins": 0.00689697265625,
|
| 83 |
+
"rewards/rejected": 0.027587890625,
|
| 84 |
+
"step": 40
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.1597444089456869,
|
| 88 |
+
"grad_norm": 31.72995333133769,
|
| 89 |
+
"learning_rate": 2.659574468085106e-07,
|
| 90 |
+
"logits/chosen": -3.84375,
|
| 91 |
+
"logits/rejected": -3.859375,
|
| 92 |
+
"logps/chosen": -116.0,
|
| 93 |
+
"logps/rejected": -130.0,
|
| 94 |
+
"loss": 0.6862,
|
| 95 |
+
"rewards/accuracies": 0.21250000596046448,
|
| 96 |
+
"rewards/chosen": 0.051513671875,
|
| 97 |
+
"rewards/margins": 0.015625,
|
| 98 |
+
"rewards/rejected": 0.035888671875,
|
| 99 |
+
"step": 50
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"epoch": 0.19169329073482427,
|
| 103 |
+
"grad_norm": 27.598228962388703,
|
| 104 |
+
"learning_rate": 3.1914893617021275e-07,
|
| 105 |
+
"logits/chosen": -3.9375,
|
| 106 |
+
"logits/rejected": -3.890625,
|
| 107 |
+
"logps/chosen": -123.0,
|
| 108 |
+
"logps/rejected": -124.0,
|
| 109 |
+
"loss": 0.6864,
|
| 110 |
+
"rewards/accuracies": 0.20000000298023224,
|
| 111 |
+
"rewards/chosen": 0.1298828125,
|
| 112 |
+
"rewards/margins": 0.0147705078125,
|
| 113 |
+
"rewards/rejected": 0.115234375,
|
| 114 |
+
"step": 60
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.22364217252396165,
|
| 118 |
+
"grad_norm": 24.68090794587246,
|
| 119 |
+
"learning_rate": 3.7234042553191484e-07,
|
| 120 |
+
"logits/chosen": -3.84375,
|
| 121 |
+
"logits/rejected": -3.875,
|
| 122 |
+
"logps/chosen": -114.0,
|
| 123 |
+
"logps/rejected": -120.5,
|
| 124 |
+
"loss": 0.6794,
|
| 125 |
+
"rewards/accuracies": 0.1875,
|
| 126 |
+
"rewards/chosen": 0.2099609375,
|
| 127 |
+
"rewards/margins": 0.0250244140625,
|
| 128 |
+
"rewards/rejected": 0.185546875,
|
| 129 |
+
"step": 70
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"epoch": 0.25559105431309903,
|
| 133 |
+
"grad_norm": 32.31032873231678,
|
| 134 |
+
"learning_rate": 4.25531914893617e-07,
|
| 135 |
+
"logits/chosen": -3.84375,
|
| 136 |
+
"logits/rejected": -3.859375,
|
| 137 |
+
"logps/chosen": -128.0,
|
| 138 |
+
"logps/rejected": -141.0,
|
| 139 |
+
"loss": 0.6746,
|
| 140 |
+
"rewards/accuracies": 0.21250000596046448,
|
| 141 |
+
"rewards/chosen": 0.359375,
|
| 142 |
+
"rewards/margins": 0.042724609375,
|
| 143 |
+
"rewards/rejected": 0.31640625,
|
| 144 |
+
"step": 80
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 0.28753993610223644,
|
| 148 |
+
"grad_norm": 25.62788513297419,
|
| 149 |
+
"learning_rate": 4.787234042553192e-07,
|
| 150 |
+
"logits/chosen": -3.921875,
|
| 151 |
+
"logits/rejected": -3.90625,
|
| 152 |
+
"logps/chosen": -125.5,
|
| 153 |
+
"logps/rejected": -134.0,
|
| 154 |
+
"loss": 0.6756,
|
| 155 |
+
"rewards/accuracies": 0.1875,
|
| 156 |
+
"rewards/chosen": 0.33203125,
|
| 157 |
+
"rewards/margins": 0.07373046875,
|
| 158 |
+
"rewards/rejected": 0.259765625,
|
| 159 |
+
"step": 90
|
| 160 |
+
},
|
| 161 |
+
{
|
| 162 |
+
"epoch": 0.3194888178913738,
|
| 163 |
+
"grad_norm": 34.430976745441534,
|
| 164 |
+
"learning_rate": 4.964497041420119e-07,
|
| 165 |
+
"logits/chosen": -3.890625,
|
| 166 |
+
"logits/rejected": -3.859375,
|
| 167 |
+
"logps/chosen": -115.0,
|
| 168 |
+
"logps/rejected": -116.0,
|
| 169 |
+
"loss": 0.6704,
|
| 170 |
+
"rewards/accuracies": 0.125,
|
| 171 |
+
"rewards/chosen": 0.30078125,
|
| 172 |
+
"rewards/margins": 0.027587890625,
|
| 173 |
+
"rewards/rejected": 0.2734375,
|
| 174 |
+
"step": 100
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"epoch": 0.3514376996805112,
|
| 178 |
+
"grad_norm": 19.726440993478906,
|
| 179 |
+
"learning_rate": 4.905325443786982e-07,
|
| 180 |
+
"logits/chosen": -3.90625,
|
| 181 |
+
"logits/rejected": -3.875,
|
| 182 |
+
"logps/chosen": -123.0,
|
| 183 |
+
"logps/rejected": -132.0,
|
| 184 |
+
"loss": 0.6666,
|
| 185 |
+
"rewards/accuracies": 0.17499999701976776,
|
| 186 |
+
"rewards/chosen": 0.361328125,
|
| 187 |
+
"rewards/margins": 0.126953125,
|
| 188 |
+
"rewards/rejected": 0.2353515625,
|
| 189 |
+
"step": 110
|
| 190 |
+
},
|
| 191 |
+
{
|
| 192 |
+
"epoch": 0.38338658146964855,
|
| 193 |
+
"grad_norm": 22.606814351335732,
|
| 194 |
+
"learning_rate": 4.846153846153846e-07,
|
| 195 |
+
"logits/chosen": -3.859375,
|
| 196 |
+
"logits/rejected": -3.875,
|
| 197 |
+
"logps/chosen": -116.5,
|
| 198 |
+
"logps/rejected": -120.5,
|
| 199 |
+
"loss": 0.6611,
|
| 200 |
+
"rewards/accuracies": 0.15000000596046448,
|
| 201 |
+
"rewards/chosen": 0.36328125,
|
| 202 |
+
"rewards/margins": 0.07373046875,
|
| 203 |
+
"rewards/rejected": 0.2890625,
|
| 204 |
+
"step": 120
|
| 205 |
+
},
|
| 206 |
+
{
|
| 207 |
+
"epoch": 0.41533546325878595,
|
| 208 |
+
"grad_norm": 30.108481813541545,
|
| 209 |
+
"learning_rate": 4.78698224852071e-07,
|
| 210 |
+
"logits/chosen": -3.84375,
|
| 211 |
+
"logits/rejected": -3.875,
|
| 212 |
+
"logps/chosen": -108.5,
|
| 213 |
+
"logps/rejected": -133.0,
|
| 214 |
+
"loss": 0.6624,
|
| 215 |
+
"rewards/accuracies": 0.2874999940395355,
|
| 216 |
+
"rewards/chosen": 0.462890625,
|
| 217 |
+
"rewards/margins": 0.2060546875,
|
| 218 |
+
"rewards/rejected": 0.255859375,
|
| 219 |
+
"step": 130
|
| 220 |
+
},
|
| 221 |
+
{
|
| 222 |
+
"epoch": 0.4472843450479233,
|
| 223 |
+
"grad_norm": 42.69476922343966,
|
| 224 |
+
"learning_rate": 4.727810650887574e-07,
|
| 225 |
+
"logits/chosen": -3.84375,
|
| 226 |
+
"logits/rejected": -3.875,
|
| 227 |
+
"logps/chosen": -122.0,
|
| 228 |
+
"logps/rejected": -131.0,
|
| 229 |
+
"loss": 0.6584,
|
| 230 |
+
"rewards/accuracies": 0.17499999701976776,
|
| 231 |
+
"rewards/chosen": 0.07666015625,
|
| 232 |
+
"rewards/margins": 0.09033203125,
|
| 233 |
+
"rewards/rejected": -0.013427734375,
|
| 234 |
+
"step": 140
|
| 235 |
+
},
|
| 236 |
+
{
|
| 237 |
+
"epoch": 0.4792332268370607,
|
| 238 |
+
"grad_norm": 29.2950267437935,
|
| 239 |
+
"learning_rate": 4.668639053254438e-07,
|
| 240 |
+
"logits/chosen": -3.859375,
|
| 241 |
+
"logits/rejected": -3.875,
|
| 242 |
+
"logps/chosen": -122.0,
|
| 243 |
+
"logps/rejected": -129.0,
|
| 244 |
+
"loss": 0.6454,
|
| 245 |
+
"rewards/accuracies": 0.21250000596046448,
|
| 246 |
+
"rewards/chosen": 0.0849609375,
|
| 247 |
+
"rewards/margins": 0.265625,
|
| 248 |
+
"rewards/rejected": -0.1806640625,
|
| 249 |
+
"step": 150
|
| 250 |
+
},
|
| 251 |
+
{
|
| 252 |
+
"epoch": 0.5111821086261981,
|
| 253 |
+
"grad_norm": 29.18218174433763,
|
| 254 |
+
"learning_rate": 4.6094674556213014e-07,
|
| 255 |
+
"logits/chosen": -3.890625,
|
| 256 |
+
"logits/rejected": -3.875,
|
| 257 |
+
"logps/chosen": -126.0,
|
| 258 |
+
"logps/rejected": -135.0,
|
| 259 |
+
"loss": 0.6396,
|
| 260 |
+
"rewards/accuracies": 0.20000000298023224,
|
| 261 |
+
"rewards/chosen": -0.1611328125,
|
| 262 |
+
"rewards/margins": 0.2197265625,
|
| 263 |
+
"rewards/rejected": -0.380859375,
|
| 264 |
+
"step": 160
|
| 265 |
+
},
|
| 266 |
+
{
|
| 267 |
+
"epoch": 0.5431309904153354,
|
| 268 |
+
"grad_norm": 24.071324961932216,
|
| 269 |
+
"learning_rate": 4.5502958579881655e-07,
|
| 270 |
+
"logits/chosen": -3.90625,
|
| 271 |
+
"logits/rejected": -3.921875,
|
| 272 |
+
"logps/chosen": -102.5,
|
| 273 |
+
"logps/rejected": -104.0,
|
| 274 |
+
"loss": 0.6336,
|
| 275 |
+
"rewards/accuracies": 0.20000000298023224,
|
| 276 |
+
"rewards/chosen": 0.31640625,
|
| 277 |
+
"rewards/margins": 0.19140625,
|
| 278 |
+
"rewards/rejected": 0.125,
|
| 279 |
+
"step": 170
|
| 280 |
+
},
|
| 281 |
+
{
|
| 282 |
+
"epoch": 0.5750798722044729,
|
| 283 |
+
"grad_norm": 15.807029915813834,
|
| 284 |
+
"learning_rate": 4.491124260355029e-07,
|
| 285 |
+
"logits/chosen": -3.859375,
|
| 286 |
+
"logits/rejected": -3.875,
|
| 287 |
+
"logps/chosen": -115.5,
|
| 288 |
+
"logps/rejected": -122.5,
|
| 289 |
+
"loss": 0.63,
|
| 290 |
+
"rewards/accuracies": 0.25,
|
| 291 |
+
"rewards/chosen": -0.45703125,
|
| 292 |
+
"rewards/margins": 0.3359375,
|
| 293 |
+
"rewards/rejected": -0.79296875,
|
| 294 |
+
"step": 180
|
| 295 |
+
},
|
| 296 |
+
{
|
| 297 |
+
"epoch": 0.6070287539936102,
|
| 298 |
+
"grad_norm": 25.188943189422474,
|
| 299 |
+
"learning_rate": 4.4319526627218936e-07,
|
| 300 |
+
"logits/chosen": -3.890625,
|
| 301 |
+
"logits/rejected": -3.90625,
|
| 302 |
+
"logps/chosen": -123.5,
|
| 303 |
+
"logps/rejected": -137.0,
|
| 304 |
+
"loss": 0.6489,
|
| 305 |
+
"rewards/accuracies": 0.21250000596046448,
|
| 306 |
+
"rewards/chosen": -0.81640625,
|
| 307 |
+
"rewards/margins": 0.1787109375,
|
| 308 |
+
"rewards/rejected": -0.99609375,
|
| 309 |
+
"step": 190
|
| 310 |
+
},
|
| 311 |
+
{
|
| 312 |
+
"epoch": 0.6389776357827476,
|
| 313 |
+
"grad_norm": 27.085036945419034,
|
| 314 |
+
"learning_rate": 4.372781065088757e-07,
|
| 315 |
+
"logits/chosen": -3.859375,
|
| 316 |
+
"logits/rejected": -3.84375,
|
| 317 |
+
"logps/chosen": -116.0,
|
| 318 |
+
"logps/rejected": -123.0,
|
| 319 |
+
"loss": 0.6206,
|
| 320 |
+
"rewards/accuracies": 0.22499999403953552,
|
| 321 |
+
"rewards/chosen": -0.357421875,
|
| 322 |
+
"rewards/margins": 0.2431640625,
|
| 323 |
+
"rewards/rejected": -0.6015625,
|
| 324 |
+
"step": 200
|
| 325 |
+
},
|
| 326 |
+
{
|
| 327 |
+
"epoch": 0.670926517571885,
|
| 328 |
+
"grad_norm": 59.95656852986459,
|
| 329 |
+
"learning_rate": 4.313609467455621e-07,
|
| 330 |
+
"logits/chosen": -3.90625,
|
| 331 |
+
"logits/rejected": -3.90625,
|
| 332 |
+
"logps/chosen": -120.0,
|
| 333 |
+
"logps/rejected": -122.0,
|
| 334 |
+
"loss": 0.6614,
|
| 335 |
+
"rewards/accuracies": 0.1875,
|
| 336 |
+
"rewards/chosen": -0.2216796875,
|
| 337 |
+
"rewards/margins": 0.1650390625,
|
| 338 |
+
"rewards/rejected": -0.38671875,
|
| 339 |
+
"step": 210
|
| 340 |
+
},
|
| 341 |
+
{
|
| 342 |
+
"epoch": 0.7028753993610224,
|
| 343 |
+
"grad_norm": 10.470332523245947,
|
| 344 |
+
"learning_rate": 4.2544378698224847e-07,
|
| 345 |
+
"logits/chosen": -3.96875,
|
| 346 |
+
"logits/rejected": -3.984375,
|
| 347 |
+
"logps/chosen": -122.5,
|
| 348 |
+
"logps/rejected": -133.0,
|
| 349 |
+
"loss": 0.6319,
|
| 350 |
+
"rewards/accuracies": 0.20000000298023224,
|
| 351 |
+
"rewards/chosen": 0.0751953125,
|
| 352 |
+
"rewards/margins": 0.197265625,
|
| 353 |
+
"rewards/rejected": -0.1220703125,
|
| 354 |
+
"step": 220
|
| 355 |
+
},
|
| 356 |
+
{
|
| 357 |
+
"epoch": 0.7348242811501597,
|
| 358 |
+
"grad_norm": 29.016293729576255,
|
| 359 |
+
"learning_rate": 4.195266272189349e-07,
|
| 360 |
+
"logits/chosen": -3.984375,
|
| 361 |
+
"logits/rejected": -3.984375,
|
| 362 |
+
"logps/chosen": -146.0,
|
| 363 |
+
"logps/rejected": -135.0,
|
| 364 |
+
"loss": 0.6889,
|
| 365 |
+
"rewards/accuracies": 0.15000000596046448,
|
| 366 |
+
"rewards/chosen": -0.05224609375,
|
| 367 |
+
"rewards/margins": 0.10205078125,
|
| 368 |
+
"rewards/rejected": -0.154296875,
|
| 369 |
+
"step": 230
|
| 370 |
+
},
|
| 371 |
+
{
|
| 372 |
+
"epoch": 0.7667731629392971,
|
| 373 |
+
"grad_norm": 29.554961998726093,
|
| 374 |
+
"learning_rate": 4.1360946745562133e-07,
|
| 375 |
+
"logits/chosen": -3.953125,
|
| 376 |
+
"logits/rejected": -3.9375,
|
| 377 |
+
"logps/chosen": -112.5,
|
| 378 |
+
"logps/rejected": -125.0,
|
| 379 |
+
"loss": 0.6383,
|
| 380 |
+
"rewards/accuracies": 0.13750000298023224,
|
| 381 |
+
"rewards/chosen": 0.00262451171875,
|
| 382 |
+
"rewards/margins": 0.1708984375,
|
| 383 |
+
"rewards/rejected": -0.16796875,
|
| 384 |
+
"step": 240
|
| 385 |
+
},
|
| 386 |
+
{
|
| 387 |
+
"epoch": 0.7987220447284346,
|
| 388 |
+
"grad_norm": 30.050484609489388,
|
| 389 |
+
"learning_rate": 4.076923076923077e-07,
|
| 390 |
+
"logits/chosen": -3.953125,
|
| 391 |
+
"logits/rejected": -3.984375,
|
| 392 |
+
"logps/chosen": -109.0,
|
| 393 |
+
"logps/rejected": -119.5,
|
| 394 |
+
"loss": 0.6651,
|
| 395 |
+
"rewards/accuracies": 0.2874999940395355,
|
| 396 |
+
"rewards/chosen": 0.35546875,
|
| 397 |
+
"rewards/margins": 0.310546875,
|
| 398 |
+
"rewards/rejected": 0.046630859375,
|
| 399 |
+
"step": 250
|
| 400 |
+
},
|
| 401 |
+
{
|
| 402 |
+
"epoch": 0.8306709265175719,
|
| 403 |
+
"grad_norm": 50.35008935486711,
|
| 404 |
+
"learning_rate": 4.017751479289941e-07,
|
| 405 |
+
"logits/chosen": -3.9375,
|
| 406 |
+
"logits/rejected": -3.953125,
|
| 407 |
+
"logps/chosen": -123.5,
|
| 408 |
+
"logps/rejected": -135.0,
|
| 409 |
+
"loss": 0.6779,
|
| 410 |
+
"rewards/accuracies": 0.21250000596046448,
|
| 411 |
+
"rewards/chosen": 0.455078125,
|
| 412 |
+
"rewards/margins": 0.1943359375,
|
| 413 |
+
"rewards/rejected": 0.26171875,
|
| 414 |
+
"step": 260
|
| 415 |
+
},
|
| 416 |
+
{
|
| 417 |
+
"epoch": 0.8626198083067093,
|
| 418 |
+
"grad_norm": 33.675183816708014,
|
| 419 |
+
"learning_rate": 3.9585798816568044e-07,
|
| 420 |
+
"logits/chosen": -3.953125,
|
| 421 |
+
"logits/rejected": -3.96875,
|
| 422 |
+
"logps/chosen": -104.5,
|
| 423 |
+
"logps/rejected": -109.0,
|
| 424 |
+
"loss": 0.6701,
|
| 425 |
+
"rewards/accuracies": 0.1875,
|
| 426 |
+
"rewards/chosen": 0.6171875,
|
| 427 |
+
"rewards/margins": 0.1103515625,
|
| 428 |
+
"rewards/rejected": 0.5078125,
|
| 429 |
+
"step": 270
|
| 430 |
+
},
|
| 431 |
+
{
|
| 432 |
+
"epoch": 0.8945686900958466,
|
| 433 |
+
"grad_norm": 21.039116456131303,
|
| 434 |
+
"learning_rate": 3.8994082840236685e-07,
|
| 435 |
+
"logits/chosen": -3.921875,
|
| 436 |
+
"logits/rejected": -3.921875,
|
| 437 |
+
"logps/chosen": -110.0,
|
| 438 |
+
"logps/rejected": -112.5,
|
| 439 |
+
"loss": 0.6536,
|
| 440 |
+
"rewards/accuracies": 0.20000000298023224,
|
| 441 |
+
"rewards/chosen": 0.5546875,
|
| 442 |
+
"rewards/margins": 0.0791015625,
|
| 443 |
+
"rewards/rejected": 0.474609375,
|
| 444 |
+
"step": 280
|
| 445 |
+
},
|
| 446 |
+
{
|
| 447 |
+
"epoch": 0.9265175718849841,
|
| 448 |
+
"grad_norm": 29.646688523670253,
|
| 449 |
+
"learning_rate": 3.840236686390532e-07,
|
| 450 |
+
"logits/chosen": -3.953125,
|
| 451 |
+
"logits/rejected": -3.953125,
|
| 452 |
+
"logps/chosen": -125.5,
|
| 453 |
+
"logps/rejected": -135.0,
|
| 454 |
+
"loss": 0.6689,
|
| 455 |
+
"rewards/accuracies": 0.21250000596046448,
|
| 456 |
+
"rewards/chosen": -0.06982421875,
|
| 457 |
+
"rewards/margins": 0.228515625,
|
| 458 |
+
"rewards/rejected": -0.296875,
|
| 459 |
+
"step": 290
|
| 460 |
+
},
|
| 461 |
+
{
|
| 462 |
+
"epoch": 0.9584664536741214,
|
| 463 |
+
"grad_norm": 24.41030402543526,
|
| 464 |
+
"learning_rate": 3.7810650887573966e-07,
|
| 465 |
+
"logits/chosen": -3.984375,
|
| 466 |
+
"logits/rejected": -4.0,
|
| 467 |
+
"logps/chosen": -131.0,
|
| 468 |
+
"logps/rejected": -136.0,
|
| 469 |
+
"loss": 0.6724,
|
| 470 |
+
"rewards/accuracies": 0.16249999403953552,
|
| 471 |
+
"rewards/chosen": -0.2734375,
|
| 472 |
+
"rewards/margins": 0.08837890625,
|
| 473 |
+
"rewards/rejected": -0.36328125,
|
| 474 |
+
"step": 300
|
| 475 |
+
},
|
| 476 |
+
{
|
| 477 |
+
"epoch": 0.9904153354632588,
|
| 478 |
+
"grad_norm": 19.87954624614089,
|
| 479 |
+
"learning_rate": 3.72189349112426e-07,
|
| 480 |
+
"logits/chosen": -3.953125,
|
| 481 |
+
"logits/rejected": -3.9375,
|
| 482 |
+
"logps/chosen": -133.0,
|
| 483 |
+
"logps/rejected": -137.0,
|
| 484 |
+
"loss": 0.6607,
|
| 485 |
+
"rewards/accuracies": 0.17499999701976776,
|
| 486 |
+
"rewards/chosen": -0.05224609375,
|
| 487 |
+
"rewards/margins": 0.1044921875,
|
| 488 |
+
"rewards/rejected": -0.15625,
|
| 489 |
+
"step": 310
|
| 490 |
+
},
|
| 491 |
+
{
|
| 492 |
+
"epoch": 1.0,
|
| 493 |
+
"eval_logits/chosen": -3.90625,
|
| 494 |
+
"eval_logits/rejected": -3.921875,
|
| 495 |
+
"eval_logps/chosen": -104.5,
|
| 496 |
+
"eval_logps/rejected": -109.5,
|
| 497 |
+
"eval_loss": 0.6517187356948853,
|
| 498 |
+
"eval_rewards/accuracies": 0.1964285671710968,
|
| 499 |
+
"eval_rewards/chosen": -0.1142578125,
|
| 500 |
+
"eval_rewards/margins": 0.1513671875,
|
| 501 |
+
"eval_rewards/rejected": -0.265625,
|
| 502 |
+
"eval_runtime": 12.1044,
|
| 503 |
+
"eval_samples_per_second": 16.523,
|
| 504 |
+
"eval_steps_per_second": 0.578,
|
| 505 |
+
"step": 313
|
| 506 |
+
},
|
| 507 |
+
{
|
| 508 |
+
"epoch": 1.0223642172523961,
|
| 509 |
+
"grad_norm": 10.097984976496987,
|
| 510 |
+
"learning_rate": 3.662721893491124e-07,
|
| 511 |
+
"logits/chosen": -3.984375,
|
| 512 |
+
"logits/rejected": -3.984375,
|
| 513 |
+
"logps/chosen": -135.0,
|
| 514 |
+
"logps/rejected": -144.0,
|
| 515 |
+
"loss": 0.5903,
|
| 516 |
+
"rewards/accuracies": 0.26249998807907104,
|
| 517 |
+
"rewards/chosen": 0.1611328125,
|
| 518 |
+
"rewards/margins": 0.86328125,
|
| 519 |
+
"rewards/rejected": -0.703125,
|
| 520 |
+
"step": 320
|
| 521 |
+
},
|
| 522 |
+
{
|
| 523 |
+
"epoch": 1.0543130990415335,
|
| 524 |
+
"grad_norm": 1.1859208612609746,
|
| 525 |
+
"learning_rate": 3.603550295857988e-07,
|
| 526 |
+
"logits/chosen": -4.25,
|
| 527 |
+
"logits/rejected": -4.25,
|
| 528 |
+
"logps/chosen": -130.0,
|
| 529 |
+
"logps/rejected": -151.0,
|
| 530 |
+
"loss": 0.5324,
|
| 531 |
+
"rewards/accuracies": 0.21250000596046448,
|
| 532 |
+
"rewards/chosen": -1.359375,
|
| 533 |
+
"rewards/margins": 1.1484375,
|
| 534 |
+
"rewards/rejected": -2.515625,
|
| 535 |
+
"step": 330
|
| 536 |
+
},
|
| 537 |
+
{
|
| 538 |
+
"epoch": 1.0862619808306708,
|
| 539 |
+
"grad_norm": 2.418489775814248,
|
| 540 |
+
"learning_rate": 3.544378698224852e-07,
|
| 541 |
+
"logits/chosen": -4.46875,
|
| 542 |
+
"logits/rejected": -4.5,
|
| 543 |
+
"logps/chosen": -139.0,
|
| 544 |
+
"logps/rejected": -154.0,
|
| 545 |
+
"loss": 0.5059,
|
| 546 |
+
"rewards/accuracies": 0.30000001192092896,
|
| 547 |
+
"rewards/chosen": -1.703125,
|
| 548 |
+
"rewards/margins": 1.3671875,
|
| 549 |
+
"rewards/rejected": -3.0625,
|
| 550 |
+
"step": 340
|
| 551 |
+
},
|
| 552 |
+
{
|
| 553 |
+
"epoch": 1.1182108626198084,
|
| 554 |
+
"grad_norm": 14.519066358162673,
|
| 555 |
+
"learning_rate": 3.485207100591716e-07,
|
| 556 |
+
"logits/chosen": -4.15625,
|
| 557 |
+
"logits/rejected": -4.15625,
|
| 558 |
+
"logps/chosen": -133.0,
|
| 559 |
+
"logps/rejected": -169.0,
|
| 560 |
+
"loss": 0.4855,
|
| 561 |
+
"rewards/accuracies": 0.2874999940395355,
|
| 562 |
+
"rewards/chosen": -1.6796875,
|
| 563 |
+
"rewards/margins": 1.9609375,
|
| 564 |
+
"rewards/rejected": -3.640625,
|
| 565 |
+
"step": 350
|
| 566 |
+
},
|
| 567 |
+
{
|
| 568 |
+
"epoch": 1.1501597444089458,
|
| 569 |
+
"grad_norm": 10.64966368268835,
|
| 570 |
+
"learning_rate": 3.42603550295858e-07,
|
| 571 |
+
"logits/chosen": -4.125,
|
| 572 |
+
"logits/rejected": -4.1875,
|
| 573 |
+
"logps/chosen": -148.0,
|
| 574 |
+
"logps/rejected": -186.0,
|
| 575 |
+
"loss": 0.535,
|
| 576 |
+
"rewards/accuracies": 0.22499999403953552,
|
| 577 |
+
"rewards/chosen": -2.265625,
|
| 578 |
+
"rewards/margins": 1.671875,
|
| 579 |
+
"rewards/rejected": -3.9375,
|
| 580 |
+
"step": 360
|
| 581 |
+
},
|
| 582 |
+
{
|
| 583 |
+
"epoch": 1.182108626198083,
|
| 584 |
+
"grad_norm": 16.274756270248496,
|
| 585 |
+
"learning_rate": 3.366863905325444e-07,
|
| 586 |
+
"logits/chosen": -4.1875,
|
| 587 |
+
"logits/rejected": -4.21875,
|
| 588 |
+
"logps/chosen": -153.0,
|
| 589 |
+
"logps/rejected": -170.0,
|
| 590 |
+
"loss": 0.5207,
|
| 591 |
+
"rewards/accuracies": 0.21250000596046448,
|
| 592 |
+
"rewards/chosen": -2.046875,
|
| 593 |
+
"rewards/margins": 1.5,
|
| 594 |
+
"rewards/rejected": -3.546875,
|
| 595 |
+
"step": 370
|
| 596 |
+
},
|
| 597 |
+
{
|
| 598 |
+
"epoch": 1.2140575079872205,
|
| 599 |
+
"grad_norm": 2.316873389478435,
|
| 600 |
+
"learning_rate": 3.3076923076923075e-07,
|
| 601 |
+
"logits/chosen": -4.09375,
|
| 602 |
+
"logits/rejected": -4.0625,
|
| 603 |
+
"logps/chosen": -130.0,
|
| 604 |
+
"logps/rejected": -145.0,
|
| 605 |
+
"loss": 0.5208,
|
| 606 |
+
"rewards/accuracies": 0.20000000298023224,
|
| 607 |
+
"rewards/chosen": -1.8359375,
|
| 608 |
+
"rewards/margins": 1.2890625,
|
| 609 |
+
"rewards/rejected": -3.125,
|
| 610 |
+
"step": 380
|
| 611 |
+
},
|
| 612 |
+
{
|
| 613 |
+
"epoch": 1.2460063897763578,
|
| 614 |
+
"grad_norm": 13.428350268463888,
|
| 615 |
+
"learning_rate": 3.2485207100591716e-07,
|
| 616 |
+
"logits/chosen": -4.09375,
|
| 617 |
+
"logits/rejected": -4.125,
|
| 618 |
+
"logps/chosen": -144.0,
|
| 619 |
+
"logps/rejected": -167.0,
|
| 620 |
+
"loss": 0.5454,
|
| 621 |
+
"rewards/accuracies": 0.25,
|
| 622 |
+
"rewards/chosen": -1.71875,
|
| 623 |
+
"rewards/margins": 1.4453125,
|
| 624 |
+
"rewards/rejected": -3.15625,
|
| 625 |
+
"step": 390
|
| 626 |
+
},
|
| 627 |
+
{
|
| 628 |
+
"epoch": 1.2779552715654952,
|
| 629 |
+
"grad_norm": 1.0689515086010029,
|
| 630 |
+
"learning_rate": 3.189349112426035e-07,
|
| 631 |
+
"logits/chosen": -4.09375,
|
| 632 |
+
"logits/rejected": -4.09375,
|
| 633 |
+
"logps/chosen": -159.0,
|
| 634 |
+
"logps/rejected": -182.0,
|
| 635 |
+
"loss": 0.5154,
|
| 636 |
+
"rewards/accuracies": 0.30000001192092896,
|
| 637 |
+
"rewards/chosen": -2.34375,
|
| 638 |
+
"rewards/margins": 1.8828125,
|
| 639 |
+
"rewards/rejected": -4.21875,
|
| 640 |
+
"step": 400
|
| 641 |
+
},
|
| 642 |
+
{
|
| 643 |
+
"epoch": 1.3099041533546325,
|
| 644 |
+
"grad_norm": 23.626902495164178,
|
| 645 |
+
"learning_rate": 3.130177514792899e-07,
|
| 646 |
+
"logits/chosen": -4.125,
|
| 647 |
+
"logits/rejected": -4.09375,
|
| 648 |
+
"logps/chosen": -151.0,
|
| 649 |
+
"logps/rejected": -172.0,
|
| 650 |
+
"loss": 0.4904,
|
| 651 |
+
"rewards/accuracies": 0.32499998807907104,
|
| 652 |
+
"rewards/chosen": -2.453125,
|
| 653 |
+
"rewards/margins": 1.7578125,
|
| 654 |
+
"rewards/rejected": -4.1875,
|
| 655 |
+
"step": 410
|
| 656 |
+
},
|
| 657 |
+
{
|
| 658 |
+
"epoch": 1.34185303514377,
|
| 659 |
+
"grad_norm": 11.414992983326627,
|
| 660 |
+
"learning_rate": 3.071005917159763e-07,
|
| 661 |
+
"logits/chosen": -4.03125,
|
| 662 |
+
"logits/rejected": -4.125,
|
| 663 |
+
"logps/chosen": -151.0,
|
| 664 |
+
"logps/rejected": -184.0,
|
| 665 |
+
"loss": 0.4783,
|
| 666 |
+
"rewards/accuracies": 0.38749998807907104,
|
| 667 |
+
"rewards/chosen": -2.1875,
|
| 668 |
+
"rewards/margins": 2.390625,
|
| 669 |
+
"rewards/rejected": -4.5625,
|
| 670 |
+
"step": 420
|
| 671 |
+
},
|
| 672 |
+
{
|
| 673 |
+
"epoch": 1.3738019169329074,
|
| 674 |
+
"grad_norm": 0.7721195435055699,
|
| 675 |
+
"learning_rate": 3.011834319526627e-07,
|
| 676 |
+
"logits/chosen": -4.0625,
|
| 677 |
+
"logits/rejected": -4.125,
|
| 678 |
+
"logps/chosen": -142.0,
|
| 679 |
+
"logps/rejected": -173.0,
|
| 680 |
+
"loss": 0.5043,
|
| 681 |
+
"rewards/accuracies": 0.2750000059604645,
|
| 682 |
+
"rewards/chosen": -2.796875,
|
| 683 |
+
"rewards/margins": 1.796875,
|
| 684 |
+
"rewards/rejected": -4.59375,
|
| 685 |
+
"step": 430
|
| 686 |
+
},
|
| 687 |
+
{
|
| 688 |
+
"epoch": 1.4057507987220448,
|
| 689 |
+
"grad_norm": 31.030290011978142,
|
| 690 |
+
"learning_rate": 2.952662721893491e-07,
|
| 691 |
+
"logits/chosen": -4.0625,
|
| 692 |
+
"logits/rejected": -4.0625,
|
| 693 |
+
"logps/chosen": -153.0,
|
| 694 |
+
"logps/rejected": -170.0,
|
| 695 |
+
"loss": 0.5087,
|
| 696 |
+
"rewards/accuracies": 0.23749999701976776,
|
| 697 |
+
"rewards/chosen": -2.953125,
|
| 698 |
+
"rewards/margins": 1.5234375,
|
| 699 |
+
"rewards/rejected": -4.46875,
|
| 700 |
+
"step": 440
|
| 701 |
+
},
|
| 702 |
+
{
|
| 703 |
+
"epoch": 1.4376996805111821,
|
| 704 |
+
"grad_norm": 56.770434850318985,
|
| 705 |
+
"learning_rate": 2.893491124260355e-07,
|
| 706 |
+
"logits/chosen": -4.0625,
|
| 707 |
+
"logits/rejected": -4.0625,
|
| 708 |
+
"logps/chosen": -145.0,
|
| 709 |
+
"logps/rejected": -178.0,
|
| 710 |
+
"loss": 0.5115,
|
| 711 |
+
"rewards/accuracies": 0.32499998807907104,
|
| 712 |
+
"rewards/chosen": -2.3125,
|
| 713 |
+
"rewards/margins": 2.03125,
|
| 714 |
+
"rewards/rejected": -4.34375,
|
| 715 |
+
"step": 450
|
| 716 |
+
},
|
| 717 |
+
{
|
| 718 |
+
"epoch": 1.4696485623003195,
|
| 719 |
+
"grad_norm": 4.7100434984872575,
|
| 720 |
+
"learning_rate": 2.834319526627219e-07,
|
| 721 |
+
"logits/chosen": -4.09375,
|
| 722 |
+
"logits/rejected": -4.09375,
|
| 723 |
+
"logps/chosen": -161.0,
|
| 724 |
+
"logps/rejected": -177.0,
|
| 725 |
+
"loss": 0.4992,
|
| 726 |
+
"rewards/accuracies": 0.22499999403953552,
|
| 727 |
+
"rewards/chosen": -3.484375,
|
| 728 |
+
"rewards/margins": 1.5546875,
|
| 729 |
+
"rewards/rejected": -5.03125,
|
| 730 |
+
"step": 460
|
| 731 |
+
},
|
| 732 |
+
{
|
| 733 |
+
"epoch": 1.5015974440894568,
|
| 734 |
+
"grad_norm": 15.230105133264576,
|
| 735 |
+
"learning_rate": 2.7751479289940824e-07,
|
| 736 |
+
"logits/chosen": -4.1875,
|
| 737 |
+
"logits/rejected": -4.15625,
|
| 738 |
+
"logps/chosen": -148.0,
|
| 739 |
+
"logps/rejected": -178.0,
|
| 740 |
+
"loss": 0.5108,
|
| 741 |
+
"rewards/accuracies": 0.3375000059604645,
|
| 742 |
+
"rewards/chosen": -3.625,
|
| 743 |
+
"rewards/margins": 2.140625,
|
| 744 |
+
"rewards/rejected": -5.78125,
|
| 745 |
+
"step": 470
|
| 746 |
+
},
|
| 747 |
+
{
|
| 748 |
+
"epoch": 1.5335463258785942,
|
| 749 |
+
"grad_norm": 10.678316470819308,
|
| 750 |
+
"learning_rate": 2.715976331360947e-07,
|
| 751 |
+
"logits/chosen": -4.15625,
|
| 752 |
+
"logits/rejected": -4.15625,
|
| 753 |
+
"logps/chosen": -151.0,
|
| 754 |
+
"logps/rejected": -173.0,
|
| 755 |
+
"loss": 0.521,
|
| 756 |
+
"rewards/accuracies": 0.21250000596046448,
|
| 757 |
+
"rewards/chosen": -3.796875,
|
| 758 |
+
"rewards/margins": 1.4765625,
|
| 759 |
+
"rewards/rejected": -5.28125,
|
| 760 |
+
"step": 480
|
| 761 |
+
},
|
| 762 |
+
{
|
| 763 |
+
"epoch": 1.5654952076677318,
|
| 764 |
+
"grad_norm": 8.439588364665873,
|
| 765 |
+
"learning_rate": 2.6568047337278106e-07,
|
| 766 |
+
"logits/chosen": -4.1875,
|
| 767 |
+
"logits/rejected": -4.21875,
|
| 768 |
+
"logps/chosen": -152.0,
|
| 769 |
+
"logps/rejected": -178.0,
|
| 770 |
+
"loss": 0.5203,
|
| 771 |
+
"rewards/accuracies": 0.26249998807907104,
|
| 772 |
+
"rewards/chosen": -3.296875,
|
| 773 |
+
"rewards/margins": 2.28125,
|
| 774 |
+
"rewards/rejected": -5.5625,
|
| 775 |
+
"step": 490
|
| 776 |
+
},
|
| 777 |
+
{
|
| 778 |
+
"epoch": 1.5974440894568689,
|
| 779 |
+
"grad_norm": 7.316694492795425,
|
| 780 |
+
"learning_rate": 2.5976331360946746e-07,
|
| 781 |
+
"logits/chosen": -4.125,
|
| 782 |
+
"logits/rejected": -4.15625,
|
| 783 |
+
"logps/chosen": -172.0,
|
| 784 |
+
"logps/rejected": -205.0,
|
| 785 |
+
"loss": 0.5227,
|
| 786 |
+
"rewards/accuracies": 0.2750000059604645,
|
| 787 |
+
"rewards/chosen": -3.28125,
|
| 788 |
+
"rewards/margins": 1.984375,
|
| 789 |
+
"rewards/rejected": -5.28125,
|
| 790 |
+
"step": 500
|
| 791 |
+
},
|
| 792 |
+
{
|
| 793 |
+
"epoch": 1.6293929712460065,
|
| 794 |
+
"grad_norm": 8.243916218487573,
|
| 795 |
+
"learning_rate": 2.538461538461538e-07,
|
| 796 |
+
"logits/chosen": -4.125,
|
| 797 |
+
"logits/rejected": -4.125,
|
| 798 |
+
"logps/chosen": -167.0,
|
| 799 |
+
"logps/rejected": -188.0,
|
| 800 |
+
"loss": 0.5044,
|
| 801 |
+
"rewards/accuracies": 0.2874999940395355,
|
| 802 |
+
"rewards/chosen": -3.453125,
|
| 803 |
+
"rewards/margins": 2.125,
|
| 804 |
+
"rewards/rejected": -5.5625,
|
| 805 |
+
"step": 510
|
| 806 |
+
},
|
| 807 |
+
{
|
| 808 |
+
"epoch": 1.6613418530351438,
|
| 809 |
+
"grad_norm": 7.359358015998256,
|
| 810 |
+
"learning_rate": 2.479289940828402e-07,
|
| 811 |
+
"logits/chosen": -4.0625,
|
| 812 |
+
"logits/rejected": -4.125,
|
| 813 |
+
"logps/chosen": -130.0,
|
| 814 |
+
"logps/rejected": -168.0,
|
| 815 |
+
"loss": 0.5075,
|
| 816 |
+
"rewards/accuracies": 0.3375000059604645,
|
| 817 |
+
"rewards/chosen": -2.015625,
|
| 818 |
+
"rewards/margins": 2.84375,
|
| 819 |
+
"rewards/rejected": -4.84375,
|
| 820 |
+
"step": 520
|
| 821 |
+
},
|
| 822 |
+
{
|
| 823 |
+
"epoch": 1.6932907348242812,
|
| 824 |
+
"grad_norm": 12.343229925422747,
|
| 825 |
+
"learning_rate": 2.420118343195266e-07,
|
| 826 |
+
"logits/chosen": -4.09375,
|
| 827 |
+
"logits/rejected": -4.09375,
|
| 828 |
+
"logps/chosen": -143.0,
|
| 829 |
+
"logps/rejected": -160.0,
|
| 830 |
+
"loss": 0.5023,
|
| 831 |
+
"rewards/accuracies": 0.26249998807907104,
|
| 832 |
+
"rewards/chosen": -2.140625,
|
| 833 |
+
"rewards/margins": 1.328125,
|
| 834 |
+
"rewards/rejected": -3.484375,
|
| 835 |
+
"step": 530
|
| 836 |
+
},
|
| 837 |
+
{
|
| 838 |
+
"epoch": 1.7252396166134185,
|
| 839 |
+
"grad_norm": 12.271822104112585,
|
| 840 |
+
"learning_rate": 2.36094674556213e-07,
|
| 841 |
+
"logits/chosen": -4.09375,
|
| 842 |
+
"logits/rejected": -4.125,
|
| 843 |
+
"logps/chosen": -160.0,
|
| 844 |
+
"logps/rejected": -185.0,
|
| 845 |
+
"loss": 0.5209,
|
| 846 |
+
"rewards/accuracies": 0.23749999701976776,
|
| 847 |
+
"rewards/chosen": -3.421875,
|
| 848 |
+
"rewards/margins": 1.609375,
|
| 849 |
+
"rewards/rejected": -5.03125,
|
| 850 |
+
"step": 540
|
| 851 |
+
},
|
| 852 |
+
{
|
| 853 |
+
"epoch": 1.7571884984025559,
|
| 854 |
+
"grad_norm": 1.7194656245016546,
|
| 855 |
+
"learning_rate": 2.301775147928994e-07,
|
| 856 |
+
"logits/chosen": -4.03125,
|
| 857 |
+
"logits/rejected": -4.03125,
|
| 858 |
+
"logps/chosen": -153.0,
|
| 859 |
+
"logps/rejected": -180.0,
|
| 860 |
+
"loss": 0.4985,
|
| 861 |
+
"rewards/accuracies": 0.2750000059604645,
|
| 862 |
+
"rewards/chosen": -2.515625,
|
| 863 |
+
"rewards/margins": 2.3125,
|
| 864 |
+
"rewards/rejected": -4.8125,
|
| 865 |
+
"step": 550
|
| 866 |
+
},
|
| 867 |
+
{
|
| 868 |
+
"epoch": 1.7891373801916934,
|
| 869 |
+
"grad_norm": 9.434438571448306,
|
| 870 |
+
"learning_rate": 2.242603550295858e-07,
|
| 871 |
+
"logits/chosen": -4.0625,
|
| 872 |
+
"logits/rejected": -4.0625,
|
| 873 |
+
"logps/chosen": -153.0,
|
| 874 |
+
"logps/rejected": -206.0,
|
| 875 |
+
"loss": 0.5254,
|
| 876 |
+
"rewards/accuracies": 0.3125,
|
| 877 |
+
"rewards/chosen": -3.234375,
|
| 878 |
+
"rewards/margins": 2.53125,
|
| 879 |
+
"rewards/rejected": -5.75,
|
| 880 |
+
"step": 560
|
| 881 |
+
},
|
| 882 |
+
{
|
| 883 |
+
"epoch": 1.8210862619808306,
|
| 884 |
+
"grad_norm": 3.770352466427391,
|
| 885 |
+
"learning_rate": 2.1834319526627217e-07,
|
| 886 |
+
"logits/chosen": -4.0,
|
| 887 |
+
"logits/rejected": -4.09375,
|
| 888 |
+
"logps/chosen": -140.0,
|
| 889 |
+
"logps/rejected": -175.0,
|
| 890 |
+
"loss": 0.5318,
|
| 891 |
+
"rewards/accuracies": 0.26249998807907104,
|
| 892 |
+
"rewards/chosen": -2.703125,
|
| 893 |
+
"rewards/margins": 1.9765625,
|
| 894 |
+
"rewards/rejected": -4.6875,
|
| 895 |
+
"step": 570
|
| 896 |
+
},
|
| 897 |
+
{
|
| 898 |
+
"epoch": 1.8530351437699681,
|
| 899 |
+
"grad_norm": 40.85212730947507,
|
| 900 |
+
"learning_rate": 2.1242603550295858e-07,
|
| 901 |
+
"logits/chosen": -4.03125,
|
| 902 |
+
"logits/rejected": -4.0,
|
| 903 |
+
"logps/chosen": -144.0,
|
| 904 |
+
"logps/rejected": -154.0,
|
| 905 |
+
"loss": 0.5218,
|
| 906 |
+
"rewards/accuracies": 0.25,
|
| 907 |
+
"rewards/chosen": -2.984375,
|
| 908 |
+
"rewards/margins": 1.53125,
|
| 909 |
+
"rewards/rejected": -4.5,
|
| 910 |
+
"step": 580
|
| 911 |
+
},
|
| 912 |
+
{
|
| 913 |
+
"epoch": 1.8849840255591053,
|
| 914 |
+
"grad_norm": 9.19758791601065,
|
| 915 |
+
"learning_rate": 2.0650887573964496e-07,
|
| 916 |
+
"logits/chosen": -4.0625,
|
| 917 |
+
"logits/rejected": -4.0625,
|
| 918 |
+
"logps/chosen": -154.0,
|
| 919 |
+
"logps/rejected": -176.0,
|
| 920 |
+
"loss": 0.5163,
|
| 921 |
+
"rewards/accuracies": 0.26249998807907104,
|
| 922 |
+
"rewards/chosen": -3.046875,
|
| 923 |
+
"rewards/margins": 2.1875,
|
| 924 |
+
"rewards/rejected": -5.21875,
|
| 925 |
+
"step": 590
|
| 926 |
+
},
|
| 927 |
+
{
|
| 928 |
+
"epoch": 1.9169329073482428,
|
| 929 |
+
"grad_norm": 1.8606016727764927,
|
| 930 |
+
"learning_rate": 2.0059171597633133e-07,
|
| 931 |
+
"logits/chosen": -4.09375,
|
| 932 |
+
"logits/rejected": -4.125,
|
| 933 |
+
"logps/chosen": -154.0,
|
| 934 |
+
"logps/rejected": -185.0,
|
| 935 |
+
"loss": 0.5158,
|
| 936 |
+
"rewards/accuracies": 0.30000001192092896,
|
| 937 |
+
"rewards/chosen": -4.15625,
|
| 938 |
+
"rewards/margins": 2.046875,
|
| 939 |
+
"rewards/rejected": -6.21875,
|
| 940 |
+
"step": 600
|
| 941 |
+
},
|
| 942 |
+
{
|
| 943 |
+
"epoch": 1.9488817891373802,
|
| 944 |
+
"grad_norm": 6.5440298020464915,
|
| 945 |
+
"learning_rate": 1.9467455621301774e-07,
|
| 946 |
+
"logits/chosen": -4.09375,
|
| 947 |
+
"logits/rejected": -4.09375,
|
| 948 |
+
"logps/chosen": -156.0,
|
| 949 |
+
"logps/rejected": -169.0,
|
| 950 |
+
"loss": 0.5599,
|
| 951 |
+
"rewards/accuracies": 0.25,
|
| 952 |
+
"rewards/chosen": -3.703125,
|
| 953 |
+
"rewards/margins": 1.6484375,
|
| 954 |
+
"rewards/rejected": -5.34375,
|
| 955 |
+
"step": 610
|
| 956 |
+
},
|
| 957 |
+
{
|
| 958 |
+
"epoch": 1.9808306709265175,
|
| 959 |
+
"grad_norm": 44.97604050874076,
|
| 960 |
+
"learning_rate": 1.8875739644970412e-07,
|
| 961 |
+
"logits/chosen": -3.953125,
|
| 962 |
+
"logits/rejected": -3.953125,
|
| 963 |
+
"logps/chosen": -156.0,
|
| 964 |
+
"logps/rejected": -174.0,
|
| 965 |
+
"loss": 0.5448,
|
| 966 |
+
"rewards/accuracies": 0.22499999403953552,
|
| 967 |
+
"rewards/chosen": -3.71875,
|
| 968 |
+
"rewards/margins": 1.5625,
|
| 969 |
+
"rewards/rejected": -5.28125,
|
| 970 |
+
"step": 620
|
| 971 |
+
},
|
| 972 |
+
{
|
| 973 |
+
"epoch": 2.0,
|
| 974 |
+
"eval_logits/chosen": -4.0,
|
| 975 |
+
"eval_logits/rejected": -4.0,
|
| 976 |
+
"eval_logps/chosen": -139.0,
|
| 977 |
+
"eval_logps/rejected": -146.0,
|
| 978 |
+
"eval_loss": 0.6721875071525574,
|
| 979 |
+
"eval_rewards/accuracies": 0.1785714328289032,
|
| 980 |
+
"eval_rewards/chosen": -3.5625,
|
| 981 |
+
"eval_rewards/margins": 0.365234375,
|
| 982 |
+
"eval_rewards/rejected": -3.9375,
|
| 983 |
+
"eval_runtime": 12.0754,
|
| 984 |
+
"eval_samples_per_second": 16.563,
|
| 985 |
+
"eval_steps_per_second": 0.58,
|
| 986 |
+
"step": 626
|
| 987 |
+
},
|
| 988 |
+
{
|
| 989 |
+
"epoch": 2.012779552715655,
|
| 990 |
+
"grad_norm": 5.387854507366258,
|
| 991 |
+
"learning_rate": 1.8284023668639053e-07,
|
| 992 |
+
"logits/chosen": -4.0625,
|
| 993 |
+
"logits/rejected": -4.0625,
|
| 994 |
+
"logps/chosen": -161.0,
|
| 995 |
+
"logps/rejected": -187.0,
|
| 996 |
+
"loss": 0.5223,
|
| 997 |
+
"rewards/accuracies": 0.2874999940395355,
|
| 998 |
+
"rewards/chosen": -3.515625,
|
| 999 |
+
"rewards/margins": 2.3125,
|
| 1000 |
+
"rewards/rejected": -5.8125,
|
| 1001 |
+
"step": 630
|
| 1002 |
+
},
|
| 1003 |
+
{
|
| 1004 |
+
"epoch": 2.0447284345047922,
|
| 1005 |
+
"grad_norm": 1.2556332999951558,
|
| 1006 |
+
"learning_rate": 1.7692307692307693e-07,
|
| 1007 |
+
"logits/chosen": -4.03125,
|
| 1008 |
+
"logits/rejected": -4.0625,
|
| 1009 |
+
"logps/chosen": -147.0,
|
| 1010 |
+
"logps/rejected": -174.0,
|
| 1011 |
+
"loss": 0.4742,
|
| 1012 |
+
"rewards/accuracies": 0.32499998807907104,
|
| 1013 |
+
"rewards/chosen": -2.625,
|
| 1014 |
+
"rewards/margins": 2.703125,
|
| 1015 |
+
"rewards/rejected": -5.3125,
|
| 1016 |
+
"step": 640
|
| 1017 |
+
},
|
| 1018 |
+
{
|
| 1019 |
+
"epoch": 2.07667731629393,
|
| 1020 |
+
"grad_norm": 1.0901290893409048,
|
| 1021 |
+
"learning_rate": 1.710059171597633e-07,
|
| 1022 |
+
"logits/chosen": -4.03125,
|
| 1023 |
+
"logits/rejected": -4.03125,
|
| 1024 |
+
"logps/chosen": -158.0,
|
| 1025 |
+
"logps/rejected": -181.0,
|
| 1026 |
+
"loss": 0.4933,
|
| 1027 |
+
"rewards/accuracies": 0.3125,
|
| 1028 |
+
"rewards/chosen": -3.734375,
|
| 1029 |
+
"rewards/margins": 2.5625,
|
| 1030 |
+
"rewards/rejected": -6.3125,
|
| 1031 |
+
"step": 650
|
| 1032 |
+
},
|
| 1033 |
+
{
|
| 1034 |
+
"epoch": 2.108626198083067,
|
| 1035 |
+
"grad_norm": 0.7709648536735118,
|
| 1036 |
+
"learning_rate": 1.650887573964497e-07,
|
| 1037 |
+
"logits/chosen": -3.96875,
|
| 1038 |
+
"logits/rejected": -4.0625,
|
| 1039 |
+
"logps/chosen": -145.0,
|
| 1040 |
+
"logps/rejected": -198.0,
|
| 1041 |
+
"loss": 0.4835,
|
| 1042 |
+
"rewards/accuracies": 0.4000000059604645,
|
| 1043 |
+
"rewards/chosen": -2.984375,
|
| 1044 |
+
"rewards/margins": 4.15625,
|
| 1045 |
+
"rewards/rejected": -7.15625,
|
| 1046 |
+
"step": 660
|
| 1047 |
+
},
|
| 1048 |
+
{
|
| 1049 |
+
"epoch": 2.1405750798722045,
|
| 1050 |
+
"grad_norm": 1.4684150448771174,
|
| 1051 |
+
"learning_rate": 1.591715976331361e-07,
|
| 1052 |
+
"logits/chosen": -3.96875,
|
| 1053 |
+
"logits/rejected": -4.03125,
|
| 1054 |
+
"logps/chosen": -157.0,
|
| 1055 |
+
"logps/rejected": -195.0,
|
| 1056 |
+
"loss": 0.4913,
|
| 1057 |
+
"rewards/accuracies": 0.2874999940395355,
|
| 1058 |
+
"rewards/chosen": -4.0625,
|
| 1059 |
+
"rewards/margins": 3.09375,
|
| 1060 |
+
"rewards/rejected": -7.125,
|
| 1061 |
+
"step": 670
|
| 1062 |
+
},
|
| 1063 |
+
{
|
| 1064 |
+
"epoch": 2.1725239616613417,
|
| 1065 |
+
"grad_norm": 0.15445202520650775,
|
| 1066 |
+
"learning_rate": 1.5325443786982248e-07,
|
| 1067 |
+
"logits/chosen": -3.96875,
|
| 1068 |
+
"logits/rejected": -3.96875,
|
| 1069 |
+
"logps/chosen": -155.0,
|
| 1070 |
+
"logps/rejected": -190.0,
|
| 1071 |
+
"loss": 0.5065,
|
| 1072 |
+
"rewards/accuracies": 0.26249998807907104,
|
| 1073 |
+
"rewards/chosen": -3.734375,
|
| 1074 |
+
"rewards/margins": 2.84375,
|
| 1075 |
+
"rewards/rejected": -6.59375,
|
| 1076 |
+
"step": 680
|
| 1077 |
+
},
|
| 1078 |
+
{
|
| 1079 |
+
"epoch": 2.2044728434504792,
|
| 1080 |
+
"grad_norm": 0.1682080177164145,
|
| 1081 |
+
"learning_rate": 1.4733727810650885e-07,
|
| 1082 |
+
"logits/chosen": -3.90625,
|
| 1083 |
+
"logits/rejected": -3.921875,
|
| 1084 |
+
"logps/chosen": -161.0,
|
| 1085 |
+
"logps/rejected": -190.0,
|
| 1086 |
+
"loss": 0.5062,
|
| 1087 |
+
"rewards/accuracies": 0.26249998807907104,
|
| 1088 |
+
"rewards/chosen": -4.375,
|
| 1089 |
+
"rewards/margins": 2.515625,
|
| 1090 |
+
"rewards/rejected": -6.90625,
|
| 1091 |
+
"step": 690
|
| 1092 |
+
},
|
| 1093 |
+
{
|
| 1094 |
+
"epoch": 2.236421725239617,
|
| 1095 |
+
"grad_norm": 0.48752210028572696,
|
| 1096 |
+
"learning_rate": 1.4142011834319526e-07,
|
| 1097 |
+
"logits/chosen": -3.96875,
|
| 1098 |
+
"logits/rejected": -4.0,
|
| 1099 |
+
"logps/chosen": -167.0,
|
| 1100 |
+
"logps/rejected": -201.0,
|
| 1101 |
+
"loss": 0.4656,
|
| 1102 |
+
"rewards/accuracies": 0.2874999940395355,
|
| 1103 |
+
"rewards/chosen": -4.03125,
|
| 1104 |
+
"rewards/margins": 2.6875,
|
| 1105 |
+
"rewards/rejected": -6.71875,
|
| 1106 |
+
"step": 700
|
| 1107 |
+
},
|
| 1108 |
+
{
|
| 1109 |
+
"epoch": 2.268370607028754,
|
| 1110 |
+
"grad_norm": 0.3925155000101438,
|
| 1111 |
+
"learning_rate": 1.3550295857988164e-07,
|
| 1112 |
+
"logits/chosen": -3.9375,
|
| 1113 |
+
"logits/rejected": -4.0,
|
| 1114 |
+
"logps/chosen": -168.0,
|
| 1115 |
+
"logps/rejected": -214.0,
|
| 1116 |
+
"loss": 0.4607,
|
| 1117 |
+
"rewards/accuracies": 0.2750000059604645,
|
| 1118 |
+
"rewards/chosen": -4.5,
|
| 1119 |
+
"rewards/margins": 3.25,
|
| 1120 |
+
"rewards/rejected": -7.75,
|
| 1121 |
+
"step": 710
|
| 1122 |
+
},
|
| 1123 |
+
{
|
| 1124 |
+
"epoch": 2.3003194888178915,
|
| 1125 |
+
"grad_norm": 0.4199125611952186,
|
| 1126 |
+
"learning_rate": 1.2958579881656802e-07,
|
| 1127 |
+
"logits/chosen": -3.953125,
|
| 1128 |
+
"logits/rejected": -3.96875,
|
| 1129 |
+
"logps/chosen": -172.0,
|
| 1130 |
+
"logps/rejected": -219.0,
|
| 1131 |
+
"loss": 0.504,
|
| 1132 |
+
"rewards/accuracies": 0.3125,
|
| 1133 |
+
"rewards/chosen": -4.46875,
|
| 1134 |
+
"rewards/margins": 3.53125,
|
| 1135 |
+
"rewards/rejected": -8.0,
|
| 1136 |
+
"step": 720
|
| 1137 |
+
},
|
| 1138 |
+
{
|
| 1139 |
+
"epoch": 2.3322683706070286,
|
| 1140 |
+
"grad_norm": 0.18962721290307222,
|
| 1141 |
+
"learning_rate": 1.2366863905325443e-07,
|
| 1142 |
+
"logits/chosen": -3.921875,
|
| 1143 |
+
"logits/rejected": -4.03125,
|
| 1144 |
+
"logps/chosen": -167.0,
|
| 1145 |
+
"logps/rejected": -209.0,
|
| 1146 |
+
"loss": 0.4868,
|
| 1147 |
+
"rewards/accuracies": 0.2874999940395355,
|
| 1148 |
+
"rewards/chosen": -5.09375,
|
| 1149 |
+
"rewards/margins": 3.046875,
|
| 1150 |
+
"rewards/rejected": -8.125,
|
| 1151 |
+
"step": 730
|
| 1152 |
+
},
|
| 1153 |
+
{
|
| 1154 |
+
"epoch": 2.364217252396166,
|
| 1155 |
+
"grad_norm": 0.15594202879866378,
|
| 1156 |
+
"learning_rate": 1.1775147928994082e-07,
|
| 1157 |
+
"logits/chosen": -3.953125,
|
| 1158 |
+
"logits/rejected": -3.90625,
|
| 1159 |
+
"logps/chosen": -178.0,
|
| 1160 |
+
"logps/rejected": -198.0,
|
| 1161 |
+
"loss": 0.5083,
|
| 1162 |
+
"rewards/accuracies": 0.25,
|
| 1163 |
+
"rewards/chosen": -5.21875,
|
| 1164 |
+
"rewards/margins": 2.546875,
|
| 1165 |
+
"rewards/rejected": -7.75,
|
| 1166 |
+
"step": 740
|
| 1167 |
+
},
|
| 1168 |
+
{
|
| 1169 |
+
"epoch": 2.3961661341853033,
|
| 1170 |
+
"grad_norm": 2.995272322650428,
|
| 1171 |
+
"learning_rate": 1.1183431952662721e-07,
|
| 1172 |
+
"logits/chosen": -3.921875,
|
| 1173 |
+
"logits/rejected": -3.875,
|
| 1174 |
+
"logps/chosen": -181.0,
|
| 1175 |
+
"logps/rejected": -218.0,
|
| 1176 |
+
"loss": 0.4611,
|
| 1177 |
+
"rewards/accuracies": 0.32499998807907104,
|
| 1178 |
+
"rewards/chosen": -5.09375,
|
| 1179 |
+
"rewards/margins": 2.984375,
|
| 1180 |
+
"rewards/rejected": -8.125,
|
| 1181 |
+
"step": 750
|
| 1182 |
+
},
|
| 1183 |
+
{
|
| 1184 |
+
"epoch": 2.428115015974441,
|
| 1185 |
+
"grad_norm": 0.18498092374200545,
|
| 1186 |
+
"learning_rate": 1.059171597633136e-07,
|
| 1187 |
+
"logits/chosen": -3.890625,
|
| 1188 |
+
"logits/rejected": -3.890625,
|
| 1189 |
+
"logps/chosen": -169.0,
|
| 1190 |
+
"logps/rejected": -186.0,
|
| 1191 |
+
"loss": 0.5076,
|
| 1192 |
+
"rewards/accuracies": 0.17499999701976776,
|
| 1193 |
+
"rewards/chosen": -5.28125,
|
| 1194 |
+
"rewards/margins": 1.953125,
|
| 1195 |
+
"rewards/rejected": -7.21875,
|
| 1196 |
+
"step": 760
|
| 1197 |
+
},
|
| 1198 |
+
{
|
| 1199 |
+
"epoch": 2.460063897763578,
|
| 1200 |
+
"grad_norm": 0.08884257760247284,
|
| 1201 |
+
"learning_rate": 1e-07,
|
| 1202 |
+
"logits/chosen": -3.859375,
|
| 1203 |
+
"logits/rejected": -3.921875,
|
| 1204 |
+
"logps/chosen": -145.0,
|
| 1205 |
+
"logps/rejected": -198.0,
|
| 1206 |
+
"loss": 0.4692,
|
| 1207 |
+
"rewards/accuracies": 0.2874999940395355,
|
| 1208 |
+
"rewards/chosen": -4.28125,
|
| 1209 |
+
"rewards/margins": 3.234375,
|
| 1210 |
+
"rewards/rejected": -7.53125,
|
| 1211 |
+
"step": 770
|
| 1212 |
+
},
|
| 1213 |
+
{
|
| 1214 |
+
"epoch": 2.4920127795527156,
|
| 1215 |
+
"grad_norm": 2.04148740612421,
|
| 1216 |
+
"learning_rate": 9.408284023668639e-08,
|
| 1217 |
+
"logits/chosen": -3.921875,
|
| 1218 |
+
"logits/rejected": -3.953125,
|
| 1219 |
+
"logps/chosen": -158.0,
|
| 1220 |
+
"logps/rejected": -210.0,
|
| 1221 |
+
"loss": 0.4633,
|
| 1222 |
+
"rewards/accuracies": 0.4124999940395355,
|
| 1223 |
+
"rewards/chosen": -3.90625,
|
| 1224 |
+
"rewards/margins": 4.21875,
|
| 1225 |
+
"rewards/rejected": -8.125,
|
| 1226 |
+
"step": 780
|
| 1227 |
+
},
|
| 1228 |
+
{
|
| 1229 |
+
"epoch": 2.523961661341853,
|
| 1230 |
+
"grad_norm": 1.9100967302386418,
|
| 1231 |
+
"learning_rate": 8.816568047337278e-08,
|
| 1232 |
+
"logits/chosen": -3.984375,
|
| 1233 |
+
"logits/rejected": -4.0,
|
| 1234 |
+
"logps/chosen": -167.0,
|
| 1235 |
+
"logps/rejected": -200.0,
|
| 1236 |
+
"loss": 0.5019,
|
| 1237 |
+
"rewards/accuracies": 0.25,
|
| 1238 |
+
"rewards/chosen": -5.125,
|
| 1239 |
+
"rewards/margins": 2.4375,
|
| 1240 |
+
"rewards/rejected": -7.5625,
|
| 1241 |
+
"step": 790
|
| 1242 |
+
},
|
| 1243 |
+
{
|
| 1244 |
+
"epoch": 2.5559105431309903,
|
| 1245 |
+
"grad_norm": 0.42642836462978445,
|
| 1246 |
+
"learning_rate": 8.224852071005916e-08,
|
| 1247 |
+
"logits/chosen": -3.859375,
|
| 1248 |
+
"logits/rejected": -3.921875,
|
| 1249 |
+
"logps/chosen": -184.0,
|
| 1250 |
+
"logps/rejected": -225.0,
|
| 1251 |
+
"loss": 0.4952,
|
| 1252 |
+
"rewards/accuracies": 0.26249998807907104,
|
| 1253 |
+
"rewards/chosen": -5.125,
|
| 1254 |
+
"rewards/margins": 3.078125,
|
| 1255 |
+
"rewards/rejected": -8.1875,
|
| 1256 |
+
"step": 800
|
| 1257 |
+
},
|
| 1258 |
+
{
|
| 1259 |
+
"epoch": 2.587859424920128,
|
| 1260 |
+
"grad_norm": 0.1611701890949659,
|
| 1261 |
+
"learning_rate": 7.633136094674555e-08,
|
| 1262 |
+
"logits/chosen": -3.875,
|
| 1263 |
+
"logits/rejected": -3.90625,
|
| 1264 |
+
"logps/chosen": -168.0,
|
| 1265 |
+
"logps/rejected": -202.0,
|
| 1266 |
+
"loss": 0.4953,
|
| 1267 |
+
"rewards/accuracies": 0.2874999940395355,
|
| 1268 |
+
"rewards/chosen": -5.40625,
|
| 1269 |
+
"rewards/margins": 3.234375,
|
| 1270 |
+
"rewards/rejected": -8.625,
|
| 1271 |
+
"step": 810
|
| 1272 |
+
},
|
| 1273 |
+
{
|
| 1274 |
+
"epoch": 2.619808306709265,
|
| 1275 |
+
"grad_norm": 0.15355437619982337,
|
| 1276 |
+
"learning_rate": 7.041420118343195e-08,
|
| 1277 |
+
"logits/chosen": -3.875,
|
| 1278 |
+
"logits/rejected": -3.953125,
|
| 1279 |
+
"logps/chosen": -182.0,
|
| 1280 |
+
"logps/rejected": -216.0,
|
| 1281 |
+
"loss": 0.502,
|
| 1282 |
+
"rewards/accuracies": 0.21250000596046448,
|
| 1283 |
+
"rewards/chosen": -5.8125,
|
| 1284 |
+
"rewards/margins": 2.515625,
|
| 1285 |
+
"rewards/rejected": -8.3125,
|
| 1286 |
+
"step": 820
|
| 1287 |
+
},
|
| 1288 |
+
{
|
| 1289 |
+
"epoch": 2.6517571884984026,
|
| 1290 |
+
"grad_norm": 0.23679653291952513,
|
| 1291 |
+
"learning_rate": 6.449704142011835e-08,
|
| 1292 |
+
"logits/chosen": -3.90625,
|
| 1293 |
+
"logits/rejected": -3.953125,
|
| 1294 |
+
"logps/chosen": -175.0,
|
| 1295 |
+
"logps/rejected": -210.0,
|
| 1296 |
+
"loss": 0.5132,
|
| 1297 |
+
"rewards/accuracies": 0.23749999701976776,
|
| 1298 |
+
"rewards/chosen": -4.9375,
|
| 1299 |
+
"rewards/margins": 2.8125,
|
| 1300 |
+
"rewards/rejected": -7.75,
|
| 1301 |
+
"step": 830
|
| 1302 |
+
},
|
| 1303 |
+
{
|
| 1304 |
+
"epoch": 2.68370607028754,
|
| 1305 |
+
"grad_norm": 1.1527288279228376,
|
| 1306 |
+
"learning_rate": 5.857988165680473e-08,
|
| 1307 |
+
"logits/chosen": -3.84375,
|
| 1308 |
+
"logits/rejected": -3.859375,
|
| 1309 |
+
"logps/chosen": -166.0,
|
| 1310 |
+
"logps/rejected": -205.0,
|
| 1311 |
+
"loss": 0.4931,
|
| 1312 |
+
"rewards/accuracies": 0.30000001192092896,
|
| 1313 |
+
"rewards/chosen": -4.9375,
|
| 1314 |
+
"rewards/margins": 2.9375,
|
| 1315 |
+
"rewards/rejected": -7.875,
|
| 1316 |
+
"step": 840
|
| 1317 |
+
},
|
| 1318 |
+
{
|
| 1319 |
+
"epoch": 2.7156549520766773,
|
| 1320 |
+
"grad_norm": 0.16029055506415082,
|
| 1321 |
+
"learning_rate": 5.266272189349112e-08,
|
| 1322 |
+
"logits/chosen": -3.859375,
|
| 1323 |
+
"logits/rejected": -3.890625,
|
| 1324 |
+
"logps/chosen": -170.0,
|
| 1325 |
+
"logps/rejected": -222.0,
|
| 1326 |
+
"loss": 0.4838,
|
| 1327 |
+
"rewards/accuracies": 0.3375000059604645,
|
| 1328 |
+
"rewards/chosen": -4.625,
|
| 1329 |
+
"rewards/margins": 4.0,
|
| 1330 |
+
"rewards/rejected": -8.625,
|
| 1331 |
+
"step": 850
|
| 1332 |
+
},
|
| 1333 |
+
{
|
| 1334 |
+
"epoch": 2.747603833865815,
|
| 1335 |
+
"grad_norm": 0.20428026877833508,
|
| 1336 |
+
"learning_rate": 4.674556213017751e-08,
|
| 1337 |
+
"logits/chosen": -3.875,
|
| 1338 |
+
"logits/rejected": -3.890625,
|
| 1339 |
+
"logps/chosen": -175.0,
|
| 1340 |
+
"logps/rejected": -218.0,
|
| 1341 |
+
"loss": 0.4984,
|
| 1342 |
+
"rewards/accuracies": 0.30000001192092896,
|
| 1343 |
+
"rewards/chosen": -5.375,
|
| 1344 |
+
"rewards/margins": 3.21875,
|
| 1345 |
+
"rewards/rejected": -8.625,
|
| 1346 |
+
"step": 860
|
| 1347 |
+
},
|
| 1348 |
+
{
|
| 1349 |
+
"epoch": 2.779552715654952,
|
| 1350 |
+
"grad_norm": 0.3728752675780405,
|
| 1351 |
+
"learning_rate": 4.082840236686391e-08,
|
| 1352 |
+
"logits/chosen": -3.953125,
|
| 1353 |
+
"logits/rejected": -3.953125,
|
| 1354 |
+
"logps/chosen": -189.0,
|
| 1355 |
+
"logps/rejected": -225.0,
|
| 1356 |
+
"loss": 0.4737,
|
| 1357 |
+
"rewards/accuracies": 0.22499999403953552,
|
| 1358 |
+
"rewards/chosen": -5.9375,
|
| 1359 |
+
"rewards/margins": 2.59375,
|
| 1360 |
+
"rewards/rejected": -8.5625,
|
| 1361 |
+
"step": 870
|
| 1362 |
+
},
|
| 1363 |
+
{
|
| 1364 |
+
"epoch": 2.8115015974440896,
|
| 1365 |
+
"grad_norm": 0.20010578216519595,
|
| 1366 |
+
"learning_rate": 3.4911242603550294e-08,
|
| 1367 |
+
"logits/chosen": -3.859375,
|
| 1368 |
+
"logits/rejected": -3.890625,
|
| 1369 |
+
"logps/chosen": -165.0,
|
| 1370 |
+
"logps/rejected": -185.0,
|
| 1371 |
+
"loss": 0.5236,
|
| 1372 |
+
"rewards/accuracies": 0.17499999701976776,
|
| 1373 |
+
"rewards/chosen": -5.5625,
|
| 1374 |
+
"rewards/margins": 1.5625,
|
| 1375 |
+
"rewards/rejected": -7.125,
|
| 1376 |
+
"step": 880
|
| 1377 |
+
},
|
| 1378 |
+
{
|
| 1379 |
+
"epoch": 2.8434504792332267,
|
| 1380 |
+
"grad_norm": 0.24869309029073208,
|
| 1381 |
+
"learning_rate": 2.8994082840236687e-08,
|
| 1382 |
+
"logits/chosen": -3.859375,
|
| 1383 |
+
"logits/rejected": -3.875,
|
| 1384 |
+
"logps/chosen": -180.0,
|
| 1385 |
+
"logps/rejected": -203.0,
|
| 1386 |
+
"loss": 0.4868,
|
| 1387 |
+
"rewards/accuracies": 0.2874999940395355,
|
| 1388 |
+
"rewards/chosen": -5.0,
|
| 1389 |
+
"rewards/margins": 2.65625,
|
| 1390 |
+
"rewards/rejected": -7.65625,
|
| 1391 |
+
"step": 890
|
| 1392 |
+
},
|
| 1393 |
+
{
|
| 1394 |
+
"epoch": 2.8753993610223643,
|
| 1395 |
+
"grad_norm": 0.1537738848665123,
|
| 1396 |
+
"learning_rate": 2.3076923076923076e-08,
|
| 1397 |
+
"logits/chosen": -3.84375,
|
| 1398 |
+
"logits/rejected": -3.875,
|
| 1399 |
+
"logps/chosen": -175.0,
|
| 1400 |
+
"logps/rejected": -224.0,
|
| 1401 |
+
"loss": 0.5018,
|
| 1402 |
+
"rewards/accuracies": 0.3499999940395355,
|
| 1403 |
+
"rewards/chosen": -5.34375,
|
| 1404 |
+
"rewards/margins": 4.0,
|
| 1405 |
+
"rewards/rejected": -9.3125,
|
| 1406 |
+
"step": 900
|
| 1407 |
+
},
|
| 1408 |
+
{
|
| 1409 |
+
"epoch": 2.9073482428115014,
|
| 1410 |
+
"grad_norm": 0.32035635694293296,
|
| 1411 |
+
"learning_rate": 1.7159763313609465e-08,
|
| 1412 |
+
"logits/chosen": -3.921875,
|
| 1413 |
+
"logits/rejected": -3.953125,
|
| 1414 |
+
"logps/chosen": -196.0,
|
| 1415 |
+
"logps/rejected": -229.0,
|
| 1416 |
+
"loss": 0.5212,
|
| 1417 |
+
"rewards/accuracies": 0.22499999403953552,
|
| 1418 |
+
"rewards/chosen": -6.03125,
|
| 1419 |
+
"rewards/margins": 2.421875,
|
| 1420 |
+
"rewards/rejected": -8.4375,
|
| 1421 |
+
"step": 910
|
| 1422 |
+
},
|
| 1423 |
+
{
|
| 1424 |
+
"epoch": 2.939297124600639,
|
| 1425 |
+
"grad_norm": 0.407062337314559,
|
| 1426 |
+
"learning_rate": 1.1242603550295858e-08,
|
| 1427 |
+
"logits/chosen": -3.875,
|
| 1428 |
+
"logits/rejected": -3.875,
|
| 1429 |
+
"logps/chosen": -172.0,
|
| 1430 |
+
"logps/rejected": -202.0,
|
| 1431 |
+
"loss": 0.4832,
|
| 1432 |
+
"rewards/accuracies": 0.3125,
|
| 1433 |
+
"rewards/chosen": -5.40625,
|
| 1434 |
+
"rewards/margins": 3.046875,
|
| 1435 |
+
"rewards/rejected": -8.4375,
|
| 1436 |
+
"step": 920
|
| 1437 |
+
},
|
| 1438 |
+
{
|
| 1439 |
+
"epoch": 2.9712460063897765,
|
| 1440 |
+
"grad_norm": 0.20061865266174939,
|
| 1441 |
+
"learning_rate": 5.325443786982248e-09,
|
| 1442 |
+
"logits/chosen": -3.828125,
|
| 1443 |
+
"logits/rejected": -3.8125,
|
| 1444 |
+
"logps/chosen": -173.0,
|
| 1445 |
+
"logps/rejected": -205.0,
|
| 1446 |
+
"loss": 0.4932,
|
| 1447 |
+
"rewards/accuracies": 0.30000001192092896,
|
| 1448 |
+
"rewards/chosen": -5.125,
|
| 1449 |
+
"rewards/margins": 3.28125,
|
| 1450 |
+
"rewards/rejected": -8.4375,
|
| 1451 |
+
"step": 930
|
| 1452 |
+
},
|
| 1453 |
+
{
|
| 1454 |
+
"epoch": 3.0,
|
| 1455 |
+
"eval_logits/chosen": -3.84375,
|
| 1456 |
+
"eval_logits/rejected": -3.859375,
|
| 1457 |
+
"eval_logps/chosen": -167.0,
|
| 1458 |
+
"eval_logps/rejected": -175.0,
|
| 1459 |
+
"eval_loss": 0.7137500047683716,
|
| 1460 |
+
"eval_rewards/accuracies": 0.1607142835855484,
|
| 1461 |
+
"eval_rewards/chosen": -6.375,
|
| 1462 |
+
"eval_rewards/margins": 0.53125,
|
| 1463 |
+
"eval_rewards/rejected": -6.90625,
|
| 1464 |
+
"eval_runtime": 14.9916,
|
| 1465 |
+
"eval_samples_per_second": 13.341,
|
| 1466 |
+
"eval_steps_per_second": 0.467,
|
| 1467 |
+
"step": 939
|
| 1468 |
+
},
|
| 1469 |
+
{
|
| 1470 |
+
"epoch": 3.0,
|
| 1471 |
+
"step": 939,
|
| 1472 |
+
"total_flos": 0.0,
|
| 1473 |
+
"train_loss": 0.5581134729111157,
|
| 1474 |
+
"train_runtime": 5075.7144,
|
| 1475 |
+
"train_samples_per_second": 5.91,
|
| 1476 |
+
"train_steps_per_second": 0.185
|
| 1477 |
+
}
|
| 1478 |
+
],
|
| 1479 |
+
"logging_steps": 10,
|
| 1480 |
+
"max_steps": 939,
|
| 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 |
+
}
|