Commit
•
79eea64
1
Parent(s):
072ac06
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 +826 -0
README.md
ADDED
|
@@ -0,0 +1,68 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
base_model: alignment-handbook/zephyr-7b-sft-full
|
| 3 |
+
library_name: transformers
|
| 4 |
+
model_name: zephyr-7b-dpo-full-shuffle-bad-after
|
| 5 |
+
tags:
|
| 6 |
+
- generated_from_trainer
|
| 7 |
+
- trl
|
| 8 |
+
- dpo
|
| 9 |
+
licence: license
|
| 10 |
+
---
|
| 11 |
+
|
| 12 |
+
# Model Card for zephyr-7b-dpo-full-shuffle-bad-after
|
| 13 |
+
|
| 14 |
+
This model is a fine-tuned version of [alignment-handbook/zephyr-7b-sft-full](https://huggingface.co/alignment-handbook/zephyr-7b-sft-full).
|
| 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="bigheiniuJ/zephyr-7b-dpo-full-shuffle-bad-after", 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: 0.12.1
|
| 37 |
+
- Transformers: 4.46.3
|
| 38 |
+
- Pytorch: 2.5.1
|
| 39 |
+
- Datasets: 2.21.0
|
| 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": 1.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.251720839215123,
|
| 5 |
+
"train_runtime": 7290.7659,
|
| 6 |
+
"train_samples": 61135,
|
| 7 |
+
"train_samples_per_second": 8.385,
|
| 8 |
+
"train_steps_per_second": 0.066
|
| 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": 1.0,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.251720839215123,
|
| 5 |
+
"train_runtime": 7290.7659,
|
| 6 |
+
"train_samples": 61135,
|
| 7 |
+
"train_samples_per_second": 8.385,
|
| 8 |
+
"train_steps_per_second": 0.066
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,826 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 1.0,
|
| 5 |
+
"eval_steps": 100,
|
| 6 |
+
"global_step": 478,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.0020920502092050207,
|
| 13 |
+
"grad_norm": 10.133265544600611,
|
| 14 |
+
"learning_rate": 1.0416666666666666e-08,
|
| 15 |
+
"logits/chosen": -3.21875,
|
| 16 |
+
"logits/rejected": -3.21875,
|
| 17 |
+
"logps/chosen": -420.0,
|
| 18 |
+
"logps/rejected": -552.0,
|
| 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": 0.02092050209205021,
|
| 28 |
+
"grad_norm": 9.184965981316628,
|
| 29 |
+
"learning_rate": 1.0416666666666667e-07,
|
| 30 |
+
"logits/chosen": -3.265625,
|
| 31 |
+
"logits/rejected": -3.28125,
|
| 32 |
+
"logps/chosen": -484.0,
|
| 33 |
+
"logps/rejected": -516.0,
|
| 34 |
+
"loss": 0.6919,
|
| 35 |
+
"rewards/accuracies": 0.1111111119389534,
|
| 36 |
+
"rewards/chosen": 0.001251220703125,
|
| 37 |
+
"rewards/margins": 0.00055694580078125,
|
| 38 |
+
"rewards/rejected": 0.00069427490234375,
|
| 39 |
+
"step": 10
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.04184100418410042,
|
| 43 |
+
"grad_norm": 10.105256126223187,
|
| 44 |
+
"learning_rate": 2.0833333333333333e-07,
|
| 45 |
+
"logits/chosen": -3.21875,
|
| 46 |
+
"logits/rejected": -3.234375,
|
| 47 |
+
"logps/chosen": -520.0,
|
| 48 |
+
"logps/rejected": -540.0,
|
| 49 |
+
"loss": 0.6902,
|
| 50 |
+
"rewards/accuracies": 0.3125,
|
| 51 |
+
"rewards/chosen": 0.00799560546875,
|
| 52 |
+
"rewards/margins": 0.004119873046875,
|
| 53 |
+
"rewards/rejected": 0.003875732421875,
|
| 54 |
+
"step": 20
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.06276150627615062,
|
| 58 |
+
"grad_norm": 8.97965579353216,
|
| 59 |
+
"learning_rate": 3.1249999999999997e-07,
|
| 60 |
+
"logits/chosen": -3.28125,
|
| 61 |
+
"logits/rejected": -3.265625,
|
| 62 |
+
"logps/chosen": -468.0,
|
| 63 |
+
"logps/rejected": -516.0,
|
| 64 |
+
"loss": 0.6829,
|
| 65 |
+
"rewards/accuracies": 0.5375000238418579,
|
| 66 |
+
"rewards/chosen": 0.040771484375,
|
| 67 |
+
"rewards/margins": 0.0198974609375,
|
| 68 |
+
"rewards/rejected": 0.0208740234375,
|
| 69 |
+
"step": 30
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.08368200836820083,
|
| 73 |
+
"grad_norm": 9.227660720825863,
|
| 74 |
+
"learning_rate": 4.1666666666666667e-07,
|
| 75 |
+
"logits/chosen": -3.25,
|
| 76 |
+
"logits/rejected": -3.328125,
|
| 77 |
+
"logps/chosen": -506.0,
|
| 78 |
+
"logps/rejected": -520.0,
|
| 79 |
+
"loss": 0.6627,
|
| 80 |
+
"rewards/accuracies": 0.675000011920929,
|
| 81 |
+
"rewards/chosen": 0.09130859375,
|
| 82 |
+
"rewards/margins": 0.059814453125,
|
| 83 |
+
"rewards/rejected": 0.03173828125,
|
| 84 |
+
"step": 40
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.10460251046025104,
|
| 88 |
+
"grad_norm": 9.819087605132305,
|
| 89 |
+
"learning_rate": 4.999733114418725e-07,
|
| 90 |
+
"logits/chosen": -3.359375,
|
| 91 |
+
"logits/rejected": -3.390625,
|
| 92 |
+
"logps/chosen": -498.0,
|
| 93 |
+
"logps/rejected": -544.0,
|
| 94 |
+
"loss": 0.6301,
|
| 95 |
+
"rewards/accuracies": 0.737500011920929,
|
| 96 |
+
"rewards/chosen": 0.0257568359375,
|
| 97 |
+
"rewards/margins": 0.1376953125,
|
| 98 |
+
"rewards/rejected": -0.11181640625,
|
| 99 |
+
"step": 50
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"epoch": 0.12552301255230125,
|
| 103 |
+
"grad_norm": 12.048088131076954,
|
| 104 |
+
"learning_rate": 4.990398100856366e-07,
|
| 105 |
+
"logits/chosen": -3.53125,
|
| 106 |
+
"logits/rejected": -3.484375,
|
| 107 |
+
"logps/chosen": -568.0,
|
| 108 |
+
"logps/rejected": -604.0,
|
| 109 |
+
"loss": 0.5911,
|
| 110 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 111 |
+
"rewards/chosen": -0.236328125,
|
| 112 |
+
"rewards/margins": 0.267578125,
|
| 113 |
+
"rewards/rejected": -0.50390625,
|
| 114 |
+
"step": 60
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.14644351464435146,
|
| 118 |
+
"grad_norm": 14.528700694610064,
|
| 119 |
+
"learning_rate": 4.967775735898179e-07,
|
| 120 |
+
"logits/chosen": -3.703125,
|
| 121 |
+
"logits/rejected": -3.65625,
|
| 122 |
+
"logps/chosen": -576.0,
|
| 123 |
+
"logps/rejected": -660.0,
|
| 124 |
+
"loss": 0.54,
|
| 125 |
+
"rewards/accuracies": 0.78125,
|
| 126 |
+
"rewards/chosen": -0.6015625,
|
| 127 |
+
"rewards/margins": 0.478515625,
|
| 128 |
+
"rewards/rejected": -1.078125,
|
| 129 |
+
"step": 70
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"epoch": 0.16736401673640167,
|
| 133 |
+
"grad_norm": 14.89693527880136,
|
| 134 |
+
"learning_rate": 4.931986719649298e-07,
|
| 135 |
+
"logits/chosen": -3.78125,
|
| 136 |
+
"logits/rejected": -3.734375,
|
| 137 |
+
"logps/chosen": -680.0,
|
| 138 |
+
"logps/rejected": -792.0,
|
| 139 |
+
"loss": 0.468,
|
| 140 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 141 |
+
"rewards/chosen": -1.4921875,
|
| 142 |
+
"rewards/margins": 0.75,
|
| 143 |
+
"rewards/rejected": -2.234375,
|
| 144 |
+
"step": 80
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 0.18828451882845187,
|
| 148 |
+
"grad_norm": 24.010753148040607,
|
| 149 |
+
"learning_rate": 4.883222001996351e-07,
|
| 150 |
+
"logits/chosen": -3.71875,
|
| 151 |
+
"logits/rejected": -3.734375,
|
| 152 |
+
"logps/chosen": -704.0,
|
| 153 |
+
"logps/rejected": -880.0,
|
| 154 |
+
"loss": 0.3987,
|
| 155 |
+
"rewards/accuracies": 0.84375,
|
| 156 |
+
"rewards/chosen": -2.03125,
|
| 157 |
+
"rewards/margins": 1.3125,
|
| 158 |
+
"rewards/rejected": -3.34375,
|
| 159 |
+
"step": 90
|
| 160 |
+
},
|
| 161 |
+
{
|
| 162 |
+
"epoch": 0.20920502092050208,
|
| 163 |
+
"grad_norm": 20.988032239641004,
|
| 164 |
+
"learning_rate": 4.821741763807186e-07,
|
| 165 |
+
"logits/chosen": -3.71875,
|
| 166 |
+
"logits/rejected": -3.703125,
|
| 167 |
+
"logps/chosen": -732.0,
|
| 168 |
+
"logps/rejected": -912.0,
|
| 169 |
+
"loss": 0.3407,
|
| 170 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 171 |
+
"rewards/chosen": -2.46875,
|
| 172 |
+
"rewards/margins": 1.5859375,
|
| 173 |
+
"rewards/rejected": -4.0625,
|
| 174 |
+
"step": 100
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"epoch": 0.20920502092050208,
|
| 178 |
+
"eval_logits/chosen": -3.75,
|
| 179 |
+
"eval_logits/rejected": -3.75,
|
| 180 |
+
"eval_logps/chosen": -724.0,
|
| 181 |
+
"eval_logps/rejected": -912.0,
|
| 182 |
+
"eval_loss": 0.3028574287891388,
|
| 183 |
+
"eval_rewards/accuracies": 0.8828125,
|
| 184 |
+
"eval_rewards/chosen": -2.25,
|
| 185 |
+
"eval_rewards/margins": 1.75,
|
| 186 |
+
"eval_rewards/rejected": -4.0,
|
| 187 |
+
"eval_runtime": 89.7294,
|
| 188 |
+
"eval_samples_per_second": 22.289,
|
| 189 |
+
"eval_steps_per_second": 0.357,
|
| 190 |
+
"step": 100
|
| 191 |
+
},
|
| 192 |
+
{
|
| 193 |
+
"epoch": 0.2301255230125523,
|
| 194 |
+
"grad_norm": 19.53117460181654,
|
| 195 |
+
"learning_rate": 4.747874028753375e-07,
|
| 196 |
+
"logits/chosen": -3.703125,
|
| 197 |
+
"logits/rejected": -3.65625,
|
| 198 |
+
"logps/chosen": -756.0,
|
| 199 |
+
"logps/rejected": -1024.0,
|
| 200 |
+
"loss": 0.2921,
|
| 201 |
+
"rewards/accuracies": 0.862500011920929,
|
| 202 |
+
"rewards/chosen": -2.578125,
|
| 203 |
+
"rewards/margins": 2.390625,
|
| 204 |
+
"rewards/rejected": -4.96875,
|
| 205 |
+
"step": 110
|
| 206 |
+
},
|
| 207 |
+
{
|
| 208 |
+
"epoch": 0.2510460251046025,
|
| 209 |
+
"grad_norm": 26.848552029083983,
|
| 210 |
+
"learning_rate": 4.662012913161997e-07,
|
| 211 |
+
"logits/chosen": -3.765625,
|
| 212 |
+
"logits/rejected": -3.75,
|
| 213 |
+
"logps/chosen": -692.0,
|
| 214 |
+
"logps/rejected": -1040.0,
|
| 215 |
+
"loss": 0.2388,
|
| 216 |
+
"rewards/accuracies": 0.918749988079071,
|
| 217 |
+
"rewards/chosen": -2.15625,
|
| 218 |
+
"rewards/margins": 3.140625,
|
| 219 |
+
"rewards/rejected": -5.3125,
|
| 220 |
+
"step": 120
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"epoch": 0.2719665271966527,
|
| 224 |
+
"grad_norm": 49.46963494787925,
|
| 225 |
+
"learning_rate": 4.5646165232345103e-07,
|
| 226 |
+
"logits/chosen": -3.75,
|
| 227 |
+
"logits/rejected": -3.671875,
|
| 228 |
+
"logps/chosen": -820.0,
|
| 229 |
+
"logps/rejected": -1376.0,
|
| 230 |
+
"loss": 0.2143,
|
| 231 |
+
"rewards/accuracies": 0.8812500238418579,
|
| 232 |
+
"rewards/chosen": -3.265625,
|
| 233 |
+
"rewards/margins": 5.125,
|
| 234 |
+
"rewards/rejected": -8.375,
|
| 235 |
+
"step": 130
|
| 236 |
+
},
|
| 237 |
+
{
|
| 238 |
+
"epoch": 0.2928870292887029,
|
| 239 |
+
"grad_norm": 27.856656903315532,
|
| 240 |
+
"learning_rate": 4.456204510851956e-07,
|
| 241 |
+
"logits/chosen": -3.75,
|
| 242 |
+
"logits/rejected": -3.625,
|
| 243 |
+
"logps/chosen": -852.0,
|
| 244 |
+
"logps/rejected": -1416.0,
|
| 245 |
+
"loss": 0.196,
|
| 246 |
+
"rewards/accuracies": 0.875,
|
| 247 |
+
"rewards/chosen": -3.40625,
|
| 248 |
+
"rewards/margins": 5.25,
|
| 249 |
+
"rewards/rejected": -8.625,
|
| 250 |
+
"step": 140
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"epoch": 0.3138075313807531,
|
| 254 |
+
"grad_norm": 23.325737015073706,
|
| 255 |
+
"learning_rate": 4.337355301007335e-07,
|
| 256 |
+
"logits/chosen": -3.796875,
|
| 257 |
+
"logits/rejected": -3.71875,
|
| 258 |
+
"logps/chosen": -740.0,
|
| 259 |
+
"logps/rejected": -1360.0,
|
| 260 |
+
"loss": 0.1871,
|
| 261 |
+
"rewards/accuracies": 0.925000011920929,
|
| 262 |
+
"rewards/chosen": -2.453125,
|
| 263 |
+
"rewards/margins": 5.78125,
|
| 264 |
+
"rewards/rejected": -8.25,
|
| 265 |
+
"step": 150
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 0.33472803347280333,
|
| 269 |
+
"grad_norm": 40.59499239982402,
|
| 270 |
+
"learning_rate": 4.2087030056579986e-07,
|
| 271 |
+
"logits/chosen": -3.765625,
|
| 272 |
+
"logits/rejected": -3.65625,
|
| 273 |
+
"logps/chosen": -708.0,
|
| 274 |
+
"logps/rejected": -1296.0,
|
| 275 |
+
"loss": 0.1955,
|
| 276 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 277 |
+
"rewards/chosen": -2.453125,
|
| 278 |
+
"rewards/margins": 5.5,
|
| 279 |
+
"rewards/rejected": -7.9375,
|
| 280 |
+
"step": 160
|
| 281 |
+
},
|
| 282 |
+
{
|
| 283 |
+
"epoch": 0.35564853556485354,
|
| 284 |
+
"grad_norm": 32.891772625367125,
|
| 285 |
+
"learning_rate": 4.070934040463998e-07,
|
| 286 |
+
"logits/chosen": -3.703125,
|
| 287 |
+
"logits/rejected": -3.734375,
|
| 288 |
+
"logps/chosen": -716.0,
|
| 289 |
+
"logps/rejected": -1320.0,
|
| 290 |
+
"loss": 0.2043,
|
| 291 |
+
"rewards/accuracies": 0.893750011920929,
|
| 292 |
+
"rewards/chosen": -2.359375,
|
| 293 |
+
"rewards/margins": 5.375,
|
| 294 |
+
"rewards/rejected": -7.75,
|
| 295 |
+
"step": 170
|
| 296 |
+
},
|
| 297 |
+
{
|
| 298 |
+
"epoch": 0.37656903765690375,
|
| 299 |
+
"grad_norm": 21.73830270472661,
|
| 300 |
+
"learning_rate": 3.9247834624635404e-07,
|
| 301 |
+
"logits/chosen": -3.765625,
|
| 302 |
+
"logits/rejected": -3.71875,
|
| 303 |
+
"logps/chosen": -780.0,
|
| 304 |
+
"logps/rejected": -1376.0,
|
| 305 |
+
"loss": 0.1924,
|
| 306 |
+
"rewards/accuracies": 0.90625,
|
| 307 |
+
"rewards/chosen": -2.578125,
|
| 308 |
+
"rewards/margins": 5.84375,
|
| 309 |
+
"rewards/rejected": -8.4375,
|
| 310 |
+
"step": 180
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"epoch": 0.39748953974895396,
|
| 314 |
+
"grad_norm": 16.372098376222763,
|
| 315 |
+
"learning_rate": 3.7710310482256523e-07,
|
| 316 |
+
"logits/chosen": -3.8125,
|
| 317 |
+
"logits/rejected": -3.734375,
|
| 318 |
+
"logps/chosen": -696.0,
|
| 319 |
+
"logps/rejected": -1344.0,
|
| 320 |
+
"loss": 0.1693,
|
| 321 |
+
"rewards/accuracies": 0.893750011920929,
|
| 322 |
+
"rewards/chosen": -2.234375,
|
| 323 |
+
"rewards/margins": 5.96875,
|
| 324 |
+
"rewards/rejected": -8.1875,
|
| 325 |
+
"step": 190
|
| 326 |
+
},
|
| 327 |
+
{
|
| 328 |
+
"epoch": 0.41841004184100417,
|
| 329 |
+
"grad_norm": 41.66546161927783,
|
| 330 |
+
"learning_rate": 3.610497133404795e-07,
|
| 331 |
+
"logits/chosen": -3.703125,
|
| 332 |
+
"logits/rejected": -3.703125,
|
| 333 |
+
"logps/chosen": -844.0,
|
| 334 |
+
"logps/rejected": -1624.0,
|
| 335 |
+
"loss": 0.1574,
|
| 336 |
+
"rewards/accuracies": 0.893750011920929,
|
| 337 |
+
"rewards/chosen": -3.328125,
|
| 338 |
+
"rewards/margins": 7.625,
|
| 339 |
+
"rewards/rejected": -11.0,
|
| 340 |
+
"step": 200
|
| 341 |
+
},
|
| 342 |
+
{
|
| 343 |
+
"epoch": 0.41841004184100417,
|
| 344 |
+
"eval_logits/chosen": -3.765625,
|
| 345 |
+
"eval_logits/rejected": -3.71875,
|
| 346 |
+
"eval_logps/chosen": -780.0,
|
| 347 |
+
"eval_logps/rejected": -1536.0,
|
| 348 |
+
"eval_loss": 0.15842990577220917,
|
| 349 |
+
"eval_rewards/accuracies": 0.9296875,
|
| 350 |
+
"eval_rewards/chosen": -2.796875,
|
| 351 |
+
"eval_rewards/margins": 7.40625,
|
| 352 |
+
"eval_rewards/rejected": -10.1875,
|
| 353 |
+
"eval_runtime": 90.1471,
|
| 354 |
+
"eval_samples_per_second": 22.186,
|
| 355 |
+
"eval_steps_per_second": 0.355,
|
| 356 |
+
"step": 200
|
| 357 |
+
},
|
| 358 |
+
{
|
| 359 |
+
"epoch": 0.4393305439330544,
|
| 360 |
+
"grad_norm": 38.17786401364724,
|
| 361 |
+
"learning_rate": 3.4440382358952115e-07,
|
| 362 |
+
"logits/chosen": -3.71875,
|
| 363 |
+
"logits/rejected": -3.625,
|
| 364 |
+
"logps/chosen": -728.0,
|
| 365 |
+
"logps/rejected": -1584.0,
|
| 366 |
+
"loss": 0.169,
|
| 367 |
+
"rewards/accuracies": 0.925000011920929,
|
| 368 |
+
"rewards/chosen": -2.578125,
|
| 369 |
+
"rewards/margins": 8.125,
|
| 370 |
+
"rewards/rejected": -10.75,
|
| 371 |
+
"step": 210
|
| 372 |
+
},
|
| 373 |
+
{
|
| 374 |
+
"epoch": 0.4602510460251046,
|
| 375 |
+
"grad_norm": 28.36201506292477,
|
| 376 |
+
"learning_rate": 3.272542485937368e-07,
|
| 377 |
+
"logits/chosen": -3.734375,
|
| 378 |
+
"logits/rejected": -3.703125,
|
| 379 |
+
"logps/chosen": -732.0,
|
| 380 |
+
"logps/rejected": -1560.0,
|
| 381 |
+
"loss": 0.1756,
|
| 382 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 383 |
+
"rewards/chosen": -2.359375,
|
| 384 |
+
"rewards/margins": 8.0,
|
| 385 |
+
"rewards/rejected": -10.375,
|
| 386 |
+
"step": 220
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"epoch": 0.4811715481171548,
|
| 390 |
+
"grad_norm": 37.321244440982746,
|
| 391 |
+
"learning_rate": 3.096924887558854e-07,
|
| 392 |
+
"logits/chosen": -3.8125,
|
| 393 |
+
"logits/rejected": -3.703125,
|
| 394 |
+
"logps/chosen": -800.0,
|
| 395 |
+
"logps/rejected": -1656.0,
|
| 396 |
+
"loss": 0.1563,
|
| 397 |
+
"rewards/accuracies": 0.90625,
|
| 398 |
+
"rewards/chosen": -2.78125,
|
| 399 |
+
"rewards/margins": 7.84375,
|
| 400 |
+
"rewards/rejected": -10.625,
|
| 401 |
+
"step": 230
|
| 402 |
+
},
|
| 403 |
+
{
|
| 404 |
+
"epoch": 0.502092050209205,
|
| 405 |
+
"grad_norm": 29.25849606336902,
|
| 406 |
+
"learning_rate": 2.9181224366319943e-07,
|
| 407 |
+
"logits/chosen": -3.71875,
|
| 408 |
+
"logits/rejected": -3.578125,
|
| 409 |
+
"logps/chosen": -716.0,
|
| 410 |
+
"logps/rejected": -1440.0,
|
| 411 |
+
"loss": 0.1714,
|
| 412 |
+
"rewards/accuracies": 0.918749988079071,
|
| 413 |
+
"rewards/chosen": -2.328125,
|
| 414 |
+
"rewards/margins": 7.1875,
|
| 415 |
+
"rewards/rejected": -9.5625,
|
| 416 |
+
"step": 240
|
| 417 |
+
},
|
| 418 |
+
{
|
| 419 |
+
"epoch": 0.5230125523012552,
|
| 420 |
+
"grad_norm": 50.281378871196466,
|
| 421 |
+
"learning_rate": 2.7370891215954565e-07,
|
| 422 |
+
"logits/chosen": -3.734375,
|
| 423 |
+
"logits/rejected": -3.65625,
|
| 424 |
+
"logps/chosen": -760.0,
|
| 425 |
+
"logps/rejected": -1576.0,
|
| 426 |
+
"loss": 0.1625,
|
| 427 |
+
"rewards/accuracies": 0.949999988079071,
|
| 428 |
+
"rewards/chosen": -2.6875,
|
| 429 |
+
"rewards/margins": 7.8125,
|
| 430 |
+
"rewards/rejected": -10.5,
|
| 431 |
+
"step": 250
|
| 432 |
+
},
|
| 433 |
+
{
|
| 434 |
+
"epoch": 0.5439330543933054,
|
| 435 |
+
"grad_norm": 35.703635573374726,
|
| 436 |
+
"learning_rate": 2.55479083351317e-07,
|
| 437 |
+
"logits/chosen": -3.734375,
|
| 438 |
+
"logits/rejected": -3.640625,
|
| 439 |
+
"logps/chosen": -800.0,
|
| 440 |
+
"logps/rejected": -1600.0,
|
| 441 |
+
"loss": 0.1357,
|
| 442 |
+
"rewards/accuracies": 0.90625,
|
| 443 |
+
"rewards/chosen": -3.078125,
|
| 444 |
+
"rewards/margins": 7.625,
|
| 445 |
+
"rewards/rejected": -10.6875,
|
| 446 |
+
"step": 260
|
| 447 |
+
},
|
| 448 |
+
{
|
| 449 |
+
"epoch": 0.5648535564853556,
|
| 450 |
+
"grad_norm": 24.94230756271867,
|
| 451 |
+
"learning_rate": 2.3722002126275822e-07,
|
| 452 |
+
"logits/chosen": -3.65625,
|
| 453 |
+
"logits/rejected": -3.65625,
|
| 454 |
+
"logps/chosen": -804.0,
|
| 455 |
+
"logps/rejected": -1544.0,
|
| 456 |
+
"loss": 0.1701,
|
| 457 |
+
"rewards/accuracies": 0.90625,
|
| 458 |
+
"rewards/chosen": -3.296875,
|
| 459 |
+
"rewards/margins": 7.1875,
|
| 460 |
+
"rewards/rejected": -10.5,
|
| 461 |
+
"step": 270
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"epoch": 0.5857740585774058,
|
| 465 |
+
"grad_norm": 28.862452074602317,
|
| 466 |
+
"learning_rate": 2.19029145890313e-07,
|
| 467 |
+
"logits/chosen": -3.65625,
|
| 468 |
+
"logits/rejected": -3.578125,
|
| 469 |
+
"logps/chosen": -832.0,
|
| 470 |
+
"logps/rejected": -1520.0,
|
| 471 |
+
"loss": 0.1453,
|
| 472 |
+
"rewards/accuracies": 0.9437500238418579,
|
| 473 |
+
"rewards/chosen": -3.15625,
|
| 474 |
+
"rewards/margins": 6.625,
|
| 475 |
+
"rewards/rejected": -9.75,
|
| 476 |
+
"step": 280
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"epoch": 0.606694560669456,
|
| 480 |
+
"grad_norm": 33.07346655704146,
|
| 481 |
+
"learning_rate": 2.0100351342479216e-07,
|
| 482 |
+
"logits/chosen": -3.59375,
|
| 483 |
+
"logits/rejected": -3.5625,
|
| 484 |
+
"logps/chosen": -840.0,
|
| 485 |
+
"logps/rejected": -1592.0,
|
| 486 |
+
"loss": 0.1752,
|
| 487 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 488 |
+
"rewards/chosen": -3.4375,
|
| 489 |
+
"rewards/margins": 7.3125,
|
| 490 |
+
"rewards/rejected": -10.75,
|
| 491 |
+
"step": 290
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"epoch": 0.6276150627615062,
|
| 495 |
+
"grad_norm": 28.81151134715954,
|
| 496 |
+
"learning_rate": 1.8323929841460178e-07,
|
| 497 |
+
"logits/chosen": -3.625,
|
| 498 |
+
"logits/rejected": -3.578125,
|
| 499 |
+
"logps/chosen": -720.0,
|
| 500 |
+
"logps/rejected": -1408.0,
|
| 501 |
+
"loss": 0.1551,
|
| 502 |
+
"rewards/accuracies": 0.918749988079071,
|
| 503 |
+
"rewards/chosen": -2.359375,
|
| 504 |
+
"rewards/margins": 6.5,
|
| 505 |
+
"rewards/rejected": -8.875,
|
| 506 |
+
"step": 300
|
| 507 |
+
},
|
| 508 |
+
{
|
| 509 |
+
"epoch": 0.6276150627615062,
|
| 510 |
+
"eval_logits/chosen": -3.640625,
|
| 511 |
+
"eval_logits/rejected": -3.59375,
|
| 512 |
+
"eval_logps/chosen": -744.0,
|
| 513 |
+
"eval_logps/rejected": -1472.0,
|
| 514 |
+
"eval_loss": 0.13853245973587036,
|
| 515 |
+
"eval_rewards/accuracies": 0.93359375,
|
| 516 |
+
"eval_rewards/chosen": -2.453125,
|
| 517 |
+
"eval_rewards/margins": 7.15625,
|
| 518 |
+
"eval_rewards/rejected": -9.625,
|
| 519 |
+
"eval_runtime": 90.005,
|
| 520 |
+
"eval_samples_per_second": 22.221,
|
| 521 |
+
"eval_steps_per_second": 0.356,
|
| 522 |
+
"step": 300
|
| 523 |
+
},
|
| 524 |
+
{
|
| 525 |
+
"epoch": 0.6485355648535565,
|
| 526 |
+
"grad_norm": 39.90299619186908,
|
| 527 |
+
"learning_rate": 1.6583128063291573e-07,
|
| 528 |
+
"logits/chosen": -3.59375,
|
| 529 |
+
"logits/rejected": -3.53125,
|
| 530 |
+
"logps/chosen": -780.0,
|
| 531 |
+
"logps/rejected": -1672.0,
|
| 532 |
+
"loss": 0.1403,
|
| 533 |
+
"rewards/accuracies": 0.949999988079071,
|
| 534 |
+
"rewards/chosen": -2.921875,
|
| 535 |
+
"rewards/margins": 8.5,
|
| 536 |
+
"rewards/rejected": -11.4375,
|
| 537 |
+
"step": 310
|
| 538 |
+
},
|
| 539 |
+
{
|
| 540 |
+
"epoch": 0.6694560669456067,
|
| 541 |
+
"grad_norm": 30.707942119385958,
|
| 542 |
+
"learning_rate": 1.488723393865766e-07,
|
| 543 |
+
"logits/chosen": -3.609375,
|
| 544 |
+
"logits/rejected": -3.578125,
|
| 545 |
+
"logps/chosen": -864.0,
|
| 546 |
+
"logps/rejected": -1776.0,
|
| 547 |
+
"loss": 0.1436,
|
| 548 |
+
"rewards/accuracies": 0.9375,
|
| 549 |
+
"rewards/chosen": -3.296875,
|
| 550 |
+
"rewards/margins": 9.0,
|
| 551 |
+
"rewards/rejected": -12.3125,
|
| 552 |
+
"step": 320
|
| 553 |
+
},
|
| 554 |
+
{
|
| 555 |
+
"epoch": 0.6903765690376569,
|
| 556 |
+
"grad_norm": 22.500572393034663,
|
| 557 |
+
"learning_rate": 1.3245295796480788e-07,
|
| 558 |
+
"logits/chosen": -3.5625,
|
| 559 |
+
"logits/rejected": -3.484375,
|
| 560 |
+
"logps/chosen": -888.0,
|
| 561 |
+
"logps/rejected": -1808.0,
|
| 562 |
+
"loss": 0.1576,
|
| 563 |
+
"rewards/accuracies": 0.918749988079071,
|
| 564 |
+
"rewards/chosen": -3.828125,
|
| 565 |
+
"rewards/margins": 9.0,
|
| 566 |
+
"rewards/rejected": -12.8125,
|
| 567 |
+
"step": 330
|
| 568 |
+
},
|
| 569 |
+
{
|
| 570 |
+
"epoch": 0.7112970711297071,
|
| 571 |
+
"grad_norm": 27.937514221124733,
|
| 572 |
+
"learning_rate": 1.1666074087171627e-07,
|
| 573 |
+
"logits/chosen": -3.53125,
|
| 574 |
+
"logits/rejected": -3.46875,
|
| 575 |
+
"logps/chosen": -960.0,
|
| 576 |
+
"logps/rejected": -1920.0,
|
| 577 |
+
"loss": 0.1881,
|
| 578 |
+
"rewards/accuracies": 0.9437500238418579,
|
| 579 |
+
"rewards/chosen": -4.5625,
|
| 580 |
+
"rewards/margins": 9.3125,
|
| 581 |
+
"rewards/rejected": -13.875,
|
| 582 |
+
"step": 340
|
| 583 |
+
},
|
| 584 |
+
{
|
| 585 |
+
"epoch": 0.7322175732217573,
|
| 586 |
+
"grad_norm": 21.646791307337544,
|
| 587 |
+
"learning_rate": 1.0157994641835734e-07,
|
| 588 |
+
"logits/chosen": -3.609375,
|
| 589 |
+
"logits/rejected": -3.515625,
|
| 590 |
+
"logps/chosen": -808.0,
|
| 591 |
+
"logps/rejected": -1488.0,
|
| 592 |
+
"loss": 0.1677,
|
| 593 |
+
"rewards/accuracies": 0.893750011920929,
|
| 594 |
+
"rewards/chosen": -3.421875,
|
| 595 |
+
"rewards/margins": 6.4375,
|
| 596 |
+
"rewards/rejected": -9.875,
|
| 597 |
+
"step": 350
|
| 598 |
+
},
|
| 599 |
+
{
|
| 600 |
+
"epoch": 0.7531380753138075,
|
| 601 |
+
"grad_norm": 26.51641238698488,
|
| 602 |
+
"learning_rate": 8.729103716819111e-08,
|
| 603 |
+
"logits/chosen": -3.515625,
|
| 604 |
+
"logits/rejected": -3.4375,
|
| 605 |
+
"logps/chosen": -804.0,
|
| 606 |
+
"logps/rejected": -1520.0,
|
| 607 |
+
"loss": 0.154,
|
| 608 |
+
"rewards/accuracies": 0.9437500238418579,
|
| 609 |
+
"rewards/chosen": -2.9375,
|
| 610 |
+
"rewards/margins": 6.875,
|
| 611 |
+
"rewards/rejected": -9.8125,
|
| 612 |
+
"step": 360
|
| 613 |
+
},
|
| 614 |
+
{
|
| 615 |
+
"epoch": 0.7740585774058577,
|
| 616 |
+
"grad_norm": 32.61897967589747,
|
| 617 |
+
"learning_rate": 7.387025063449081e-08,
|
| 618 |
+
"logits/chosen": -3.59375,
|
| 619 |
+
"logits/rejected": -3.5,
|
| 620 |
+
"logps/chosen": -792.0,
|
| 621 |
+
"logps/rejected": -1480.0,
|
| 622 |
+
"loss": 0.1439,
|
| 623 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 624 |
+
"rewards/chosen": -3.28125,
|
| 625 |
+
"rewards/margins": 6.46875,
|
| 626 |
+
"rewards/rejected": -9.75,
|
| 627 |
+
"step": 370
|
| 628 |
+
},
|
| 629 |
+
{
|
| 630 |
+
"epoch": 0.7949790794979079,
|
| 631 |
+
"grad_norm": 24.073566416356307,
|
| 632 |
+
"learning_rate": 6.138919252022435e-08,
|
| 633 |
+
"logits/chosen": -3.59375,
|
| 634 |
+
"logits/rejected": -3.53125,
|
| 635 |
+
"logps/chosen": -852.0,
|
| 636 |
+
"logps/rejected": -1696.0,
|
| 637 |
+
"loss": 0.1474,
|
| 638 |
+
"rewards/accuracies": 0.9437500238418579,
|
| 639 |
+
"rewards/chosen": -3.5,
|
| 640 |
+
"rewards/margins": 7.875,
|
| 641 |
+
"rewards/rejected": -11.375,
|
| 642 |
+
"step": 380
|
| 643 |
+
},
|
| 644 |
+
{
|
| 645 |
+
"epoch": 0.8158995815899581,
|
| 646 |
+
"grad_norm": 22.024234097844488,
|
| 647 |
+
"learning_rate": 4.991445467064689e-08,
|
| 648 |
+
"logits/chosen": -3.625,
|
| 649 |
+
"logits/rejected": -3.515625,
|
| 650 |
+
"logps/chosen": -820.0,
|
| 651 |
+
"logps/rejected": -1672.0,
|
| 652 |
+
"loss": 0.155,
|
| 653 |
+
"rewards/accuracies": 0.956250011920929,
|
| 654 |
+
"rewards/chosen": -3.421875,
|
| 655 |
+
"rewards/margins": 7.90625,
|
| 656 |
+
"rewards/rejected": -11.375,
|
| 657 |
+
"step": 390
|
| 658 |
+
},
|
| 659 |
+
{
|
| 660 |
+
"epoch": 0.8368200836820083,
|
| 661 |
+
"grad_norm": 22.5208730955551,
|
| 662 |
+
"learning_rate": 3.9507259776993954e-08,
|
| 663 |
+
"logits/chosen": -3.53125,
|
| 664 |
+
"logits/rejected": -3.46875,
|
| 665 |
+
"logps/chosen": -816.0,
|
| 666 |
+
"logps/rejected": -1592.0,
|
| 667 |
+
"loss": 0.1511,
|
| 668 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 669 |
+
"rewards/chosen": -3.421875,
|
| 670 |
+
"rewards/margins": 7.3125,
|
| 671 |
+
"rewards/rejected": -10.6875,
|
| 672 |
+
"step": 400
|
| 673 |
+
},
|
| 674 |
+
{
|
| 675 |
+
"epoch": 0.8368200836820083,
|
| 676 |
+
"eval_logits/chosen": -3.5625,
|
| 677 |
+
"eval_logits/rejected": -3.5,
|
| 678 |
+
"eval_logps/chosen": -836.0,
|
| 679 |
+
"eval_logps/rejected": -1592.0,
|
| 680 |
+
"eval_loss": 0.13080458343029022,
|
| 681 |
+
"eval_rewards/accuracies": 0.94921875,
|
| 682 |
+
"eval_rewards/chosen": -3.34375,
|
| 683 |
+
"eval_rewards/margins": 7.46875,
|
| 684 |
+
"eval_rewards/rejected": -10.8125,
|
| 685 |
+
"eval_runtime": 90.1024,
|
| 686 |
+
"eval_samples_per_second": 22.197,
|
| 687 |
+
"eval_steps_per_second": 0.355,
|
| 688 |
+
"step": 400
|
| 689 |
+
},
|
| 690 |
+
{
|
| 691 |
+
"epoch": 0.8577405857740585,
|
| 692 |
+
"grad_norm": 33.05531795103649,
|
| 693 |
+
"learning_rate": 3.022313472693447e-08,
|
| 694 |
+
"logits/chosen": -3.609375,
|
| 695 |
+
"logits/rejected": -3.53125,
|
| 696 |
+
"logps/chosen": -836.0,
|
| 697 |
+
"logps/rejected": -1672.0,
|
| 698 |
+
"loss": 0.1302,
|
| 699 |
+
"rewards/accuracies": 0.9312499761581421,
|
| 700 |
+
"rewards/chosen": -3.5,
|
| 701 |
+
"rewards/margins": 7.78125,
|
| 702 |
+
"rewards/rejected": -11.25,
|
| 703 |
+
"step": 410
|
| 704 |
+
},
|
| 705 |
+
{
|
| 706 |
+
"epoch": 0.8786610878661087,
|
| 707 |
+
"grad_norm": 22.39150376227694,
|
| 708 |
+
"learning_rate": 2.2111614344599684e-08,
|
| 709 |
+
"logits/chosen": -3.515625,
|
| 710 |
+
"logits/rejected": -3.4375,
|
| 711 |
+
"logps/chosen": -820.0,
|
| 712 |
+
"logps/rejected": -1640.0,
|
| 713 |
+
"loss": 0.1296,
|
| 714 |
+
"rewards/accuracies": 0.925000011920929,
|
| 715 |
+
"rewards/chosen": -3.3125,
|
| 716 |
+
"rewards/margins": 7.78125,
|
| 717 |
+
"rewards/rejected": -11.0625,
|
| 718 |
+
"step": 420
|
| 719 |
+
},
|
| 720 |
+
{
|
| 721 |
+
"epoch": 0.899581589958159,
|
| 722 |
+
"grad_norm": 34.532509934361165,
|
| 723 |
+
"learning_rate": 1.521597710086439e-08,
|
| 724 |
+
"logits/chosen": -3.625,
|
| 725 |
+
"logits/rejected": -3.5,
|
| 726 |
+
"logps/chosen": -804.0,
|
| 727 |
+
"logps/rejected": -1648.0,
|
| 728 |
+
"loss": 0.138,
|
| 729 |
+
"rewards/accuracies": 0.925000011920929,
|
| 730 |
+
"rewards/chosen": -3.125,
|
| 731 |
+
"rewards/margins": 7.875,
|
| 732 |
+
"rewards/rejected": -11.0,
|
| 733 |
+
"step": 430
|
| 734 |
+
},
|
| 735 |
+
{
|
| 736 |
+
"epoch": 0.9205020920502092,
|
| 737 |
+
"grad_norm": 33.11707475610062,
|
| 738 |
+
"learning_rate": 9.57301420397924e-09,
|
| 739 |
+
"logits/chosen": -3.609375,
|
| 740 |
+
"logits/rejected": -3.5625,
|
| 741 |
+
"logps/chosen": -844.0,
|
| 742 |
+
"logps/rejected": -1632.0,
|
| 743 |
+
"loss": 0.1543,
|
| 744 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 745 |
+
"rewards/chosen": -3.46875,
|
| 746 |
+
"rewards/margins": 7.5625,
|
| 747 |
+
"rewards/rejected": -11.0625,
|
| 748 |
+
"step": 440
|
| 749 |
+
},
|
| 750 |
+
{
|
| 751 |
+
"epoch": 0.9414225941422594,
|
| 752 |
+
"grad_norm": 46.292590138732955,
|
| 753 |
+
"learning_rate": 5.212833302556258e-09,
|
| 754 |
+
"logits/chosen": -3.515625,
|
| 755 |
+
"logits/rejected": -3.5,
|
| 756 |
+
"logps/chosen": -824.0,
|
| 757 |
+
"logps/rejected": -1632.0,
|
| 758 |
+
"loss": 0.1439,
|
| 759 |
+
"rewards/accuracies": 0.9312499761581421,
|
| 760 |
+
"rewards/chosen": -3.53125,
|
| 761 |
+
"rewards/margins": 7.75,
|
| 762 |
+
"rewards/rejected": -11.25,
|
| 763 |
+
"step": 450
|
| 764 |
+
},
|
| 765 |
+
{
|
| 766 |
+
"epoch": 0.9623430962343096,
|
| 767 |
+
"grad_norm": 13.872523653141515,
|
| 768 |
+
"learning_rate": 2.158697848236607e-09,
|
| 769 |
+
"logits/chosen": -3.546875,
|
| 770 |
+
"logits/rejected": -3.46875,
|
| 771 |
+
"logps/chosen": -864.0,
|
| 772 |
+
"logps/rejected": -1792.0,
|
| 773 |
+
"loss": 0.1534,
|
| 774 |
+
"rewards/accuracies": 0.9437500238418579,
|
| 775 |
+
"rewards/chosen": -3.34375,
|
| 776 |
+
"rewards/margins": 9.0,
|
| 777 |
+
"rewards/rejected": -12.3125,
|
| 778 |
+
"step": 460
|
| 779 |
+
},
|
| 780 |
+
{
|
| 781 |
+
"epoch": 0.9832635983263598,
|
| 782 |
+
"grad_norm": 40.836340138155315,
|
| 783 |
+
"learning_rate": 4.269029751107489e-10,
|
| 784 |
+
"logits/chosen": -3.609375,
|
| 785 |
+
"logits/rejected": -3.46875,
|
| 786 |
+
"logps/chosen": -784.0,
|
| 787 |
+
"logps/rejected": -1576.0,
|
| 788 |
+
"loss": 0.1607,
|
| 789 |
+
"rewards/accuracies": 0.90625,
|
| 790 |
+
"rewards/chosen": -3.234375,
|
| 791 |
+
"rewards/margins": 7.625,
|
| 792 |
+
"rewards/rejected": -10.875,
|
| 793 |
+
"step": 470
|
| 794 |
+
},
|
| 795 |
+
{
|
| 796 |
+
"epoch": 1.0,
|
| 797 |
+
"step": 478,
|
| 798 |
+
"total_flos": 0.0,
|
| 799 |
+
"train_loss": 0.251720839215123,
|
| 800 |
+
"train_runtime": 7290.7659,
|
| 801 |
+
"train_samples_per_second": 8.385,
|
| 802 |
+
"train_steps_per_second": 0.066
|
| 803 |
+
}
|
| 804 |
+
],
|
| 805 |
+
"logging_steps": 10,
|
| 806 |
+
"max_steps": 478,
|
| 807 |
+
"num_input_tokens_seen": 0,
|
| 808 |
+
"num_train_epochs": 1,
|
| 809 |
+
"save_steps": 100,
|
| 810 |
+
"stateful_callbacks": {
|
| 811 |
+
"TrainerControl": {
|
| 812 |
+
"args": {
|
| 813 |
+
"should_epoch_stop": false,
|
| 814 |
+
"should_evaluate": false,
|
| 815 |
+
"should_log": false,
|
| 816 |
+
"should_save": true,
|
| 817 |
+
"should_training_stop": true
|
| 818 |
+
},
|
| 819 |
+
"attributes": {}
|
| 820 |
+
}
|
| 821 |
+
},
|
| 822 |
+
"total_flos": 0.0,
|
| 823 |
+
"train_batch_size": 8,
|
| 824 |
+
"trial_name": null,
|
| 825 |
+
"trial_params": null
|
| 826 |
+
}
|