Model save
Browse files- README.md +68 -0
- all_results.json +9 -0
- generation_config.json +10 -0
- train_results.json +9 -0
- trainer_state.json +795 -0
README.md
ADDED
|
@@ -0,0 +1,68 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
base_model: AmberYifan/llama2-7b-sft-ultrachat-safeRLHF
|
| 3 |
+
library_name: transformers
|
| 4 |
+
model_name: Llama-2-7b-sft-spin-10k
|
| 5 |
+
tags:
|
| 6 |
+
- generated_from_trainer
|
| 7 |
+
- trl
|
| 8 |
+
- dpo
|
| 9 |
+
licence: license
|
| 10 |
+
---
|
| 11 |
+
|
| 12 |
+
# Model Card for Llama-2-7b-sft-spin-10k
|
| 13 |
+
|
| 14 |
+
This model is a fine-tuned version of [AmberYifan/llama2-7b-sft-ultrachat-safeRLHF](https://huggingface.co/AmberYifan/llama2-7b-sft-ultrachat-safeRLHF).
|
| 15 |
+
It has been trained using [TRL](https://github.com/huggingface/trl).
|
| 16 |
+
|
| 17 |
+
## Quick start
|
| 18 |
+
|
| 19 |
+
```python
|
| 20 |
+
from transformers import pipeline
|
| 21 |
+
|
| 22 |
+
question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
|
| 23 |
+
generator = pipeline("text-generation", model="AmberYifan/Llama-2-7b-sft-spin-10k", device="cuda")
|
| 24 |
+
output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
|
| 25 |
+
print(output["generated_text"])
|
| 26 |
+
```
|
| 27 |
+
|
| 28 |
+
## Training procedure
|
| 29 |
+
|
| 30 |
+
[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/yifanwang/huggingface/runs/dfh8g84y)
|
| 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: 0.12.0
|
| 37 |
+
- Transformers: 4.46.3
|
| 38 |
+
- Pytorch: 2.1.2+cu121
|
| 39 |
+
- Datasets: 3.1.0
|
| 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": 2.9904153354632586,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.036010837255037814,
|
| 5 |
+
"train_runtime": 4360.4368,
|
| 6 |
+
"train_samples": 10000,
|
| 7 |
+
"train_samples_per_second": 6.88,
|
| 8 |
+
"train_steps_per_second": 0.107
|
| 9 |
+
}
|
generation_config.json
ADDED
|
@@ -0,0 +1,10 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bos_token_id": 1,
|
| 3 |
+
"do_sample": true,
|
| 4 |
+
"eos_token_id": 2,
|
| 5 |
+
"max_length": 4096,
|
| 6 |
+
"pad_token_id": 0,
|
| 7 |
+
"temperature": 0.6,
|
| 8 |
+
"top_p": 0.9,
|
| 9 |
+
"transformers_version": "4.46.3"
|
| 10 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 2.9904153354632586,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.036010837255037814,
|
| 5 |
+
"train_runtime": 4360.4368,
|
| 6 |
+
"train_samples": 10000,
|
| 7 |
+
"train_samples_per_second": 6.88,
|
| 8 |
+
"train_steps_per_second": 0.107
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,795 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 2.9904153354632586,
|
| 5 |
+
"eval_steps": 500,
|
| 6 |
+
"global_step": 468,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.006389776357827476,
|
| 13 |
+
"grad_norm": 29.368945511033267,
|
| 14 |
+
"learning_rate": 1.0638297872340425e-08,
|
| 15 |
+
"logits/chosen": -1.53125,
|
| 16 |
+
"logits/rejected": -1.28125,
|
| 17 |
+
"logps/chosen": -145.0,
|
| 18 |
+
"logps/rejected": -130.0,
|
| 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": 0.06389776357827476,
|
| 28 |
+
"grad_norm": 28.719594921405,
|
| 29 |
+
"learning_rate": 1.0638297872340425e-07,
|
| 30 |
+
"logits/chosen": -1.546875,
|
| 31 |
+
"logits/rejected": -1.34375,
|
| 32 |
+
"logps/chosen": -145.0,
|
| 33 |
+
"logps/rejected": -129.0,
|
| 34 |
+
"loss": 0.687,
|
| 35 |
+
"rewards/accuracies": 0.3055555522441864,
|
| 36 |
+
"rewards/chosen": 0.002960205078125,
|
| 37 |
+
"rewards/margins": 0.0189208984375,
|
| 38 |
+
"rewards/rejected": -0.0159912109375,
|
| 39 |
+
"step": 10
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.12779552715654952,
|
| 43 |
+
"grad_norm": 22.494308989969756,
|
| 44 |
+
"learning_rate": 2.127659574468085e-07,
|
| 45 |
+
"logits/chosen": -1.5546875,
|
| 46 |
+
"logits/rejected": -1.390625,
|
| 47 |
+
"logps/chosen": -146.0,
|
| 48 |
+
"logps/rejected": -134.0,
|
| 49 |
+
"loss": 0.5984,
|
| 50 |
+
"rewards/accuracies": 0.925000011920929,
|
| 51 |
+
"rewards/chosen": 0.00970458984375,
|
| 52 |
+
"rewards/margins": 0.203125,
|
| 53 |
+
"rewards/rejected": -0.193359375,
|
| 54 |
+
"step": 20
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.19169329073482427,
|
| 58 |
+
"grad_norm": 9.449890990720535,
|
| 59 |
+
"learning_rate": 3.1914893617021275e-07,
|
| 60 |
+
"logits/chosen": -1.546875,
|
| 61 |
+
"logits/rejected": -1.296875,
|
| 62 |
+
"logps/chosen": -140.0,
|
| 63 |
+
"logps/rejected": -136.0,
|
| 64 |
+
"loss": 0.3229,
|
| 65 |
+
"rewards/accuracies": 1.0,
|
| 66 |
+
"rewards/chosen": 0.025634765625,
|
| 67 |
+
"rewards/margins": 1.0625,
|
| 68 |
+
"rewards/rejected": -1.0390625,
|
| 69 |
+
"step": 30
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.25559105431309903,
|
| 73 |
+
"grad_norm": 0.38707031057010677,
|
| 74 |
+
"learning_rate": 4.25531914893617e-07,
|
| 75 |
+
"logits/chosen": -1.515625,
|
| 76 |
+
"logits/rejected": -1.375,
|
| 77 |
+
"logps/chosen": -149.0,
|
| 78 |
+
"logps/rejected": -169.0,
|
| 79 |
+
"loss": 0.0577,
|
| 80 |
+
"rewards/accuracies": 1.0,
|
| 81 |
+
"rewards/chosen": -0.2578125,
|
| 82 |
+
"rewards/margins": 3.671875,
|
| 83 |
+
"rewards/rejected": -3.9375,
|
| 84 |
+
"step": 40
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.3194888178913738,
|
| 88 |
+
"grad_norm": 0.20241899573369684,
|
| 89 |
+
"learning_rate": 4.96437054631829e-07,
|
| 90 |
+
"logits/chosen": -1.515625,
|
| 91 |
+
"logits/rejected": -1.359375,
|
| 92 |
+
"logps/chosen": -147.0,
|
| 93 |
+
"logps/rejected": -203.0,
|
| 94 |
+
"loss": 0.0033,
|
| 95 |
+
"rewards/accuracies": 1.0,
|
| 96 |
+
"rewards/chosen": -0.66015625,
|
| 97 |
+
"rewards/margins": 6.875,
|
| 98 |
+
"rewards/rejected": -7.53125,
|
| 99 |
+
"step": 50
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"epoch": 0.38338658146964855,
|
| 103 |
+
"grad_norm": 0.3039285596197402,
|
| 104 |
+
"learning_rate": 4.845605700712589e-07,
|
| 105 |
+
"logits/chosen": -1.5390625,
|
| 106 |
+
"logits/rejected": -1.3984375,
|
| 107 |
+
"logps/chosen": -153.0,
|
| 108 |
+
"logps/rejected": -221.0,
|
| 109 |
+
"loss": 0.0025,
|
| 110 |
+
"rewards/accuracies": 1.0,
|
| 111 |
+
"rewards/chosen": -0.7265625,
|
| 112 |
+
"rewards/margins": 8.0,
|
| 113 |
+
"rewards/rejected": -8.75,
|
| 114 |
+
"step": 60
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.4472843450479233,
|
| 118 |
+
"grad_norm": 0.362275931660039,
|
| 119 |
+
"learning_rate": 4.7268408551068883e-07,
|
| 120 |
+
"logits/chosen": -1.53125,
|
| 121 |
+
"logits/rejected": -1.4296875,
|
| 122 |
+
"logps/chosen": -151.0,
|
| 123 |
+
"logps/rejected": -222.0,
|
| 124 |
+
"loss": 0.0018,
|
| 125 |
+
"rewards/accuracies": 1.0,
|
| 126 |
+
"rewards/chosen": -0.8515625,
|
| 127 |
+
"rewards/margins": 8.375,
|
| 128 |
+
"rewards/rejected": -9.25,
|
| 129 |
+
"step": 70
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"epoch": 0.5111821086261981,
|
| 133 |
+
"grad_norm": 0.2868857164394034,
|
| 134 |
+
"learning_rate": 4.6080760095011875e-07,
|
| 135 |
+
"logits/chosen": -1.5625,
|
| 136 |
+
"logits/rejected": -1.4375,
|
| 137 |
+
"logps/chosen": -154.0,
|
| 138 |
+
"logps/rejected": -231.0,
|
| 139 |
+
"loss": 0.0009,
|
| 140 |
+
"rewards/accuracies": 1.0,
|
| 141 |
+
"rewards/chosen": -0.734375,
|
| 142 |
+
"rewards/margins": 9.3125,
|
| 143 |
+
"rewards/rejected": -10.0625,
|
| 144 |
+
"step": 80
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 0.5750798722044729,
|
| 148 |
+
"grad_norm": 0.15240372298676685,
|
| 149 |
+
"learning_rate": 4.4893111638954866e-07,
|
| 150 |
+
"logits/chosen": -1.546875,
|
| 151 |
+
"logits/rejected": -1.421875,
|
| 152 |
+
"logps/chosen": -148.0,
|
| 153 |
+
"logps/rejected": -230.0,
|
| 154 |
+
"loss": 0.0007,
|
| 155 |
+
"rewards/accuracies": 1.0,
|
| 156 |
+
"rewards/chosen": -0.73046875,
|
| 157 |
+
"rewards/margins": 9.5,
|
| 158 |
+
"rewards/rejected": -10.25,
|
| 159 |
+
"step": 90
|
| 160 |
+
},
|
| 161 |
+
{
|
| 162 |
+
"epoch": 0.6389776357827476,
|
| 163 |
+
"grad_norm": 0.02392673493900165,
|
| 164 |
+
"learning_rate": 4.3705463182897863e-07,
|
| 165 |
+
"logits/chosen": -1.5546875,
|
| 166 |
+
"logits/rejected": -1.3828125,
|
| 167 |
+
"logps/chosen": -154.0,
|
| 168 |
+
"logps/rejected": -235.0,
|
| 169 |
+
"loss": 0.0007,
|
| 170 |
+
"rewards/accuracies": 1.0,
|
| 171 |
+
"rewards/chosen": -0.7578125,
|
| 172 |
+
"rewards/margins": 9.75,
|
| 173 |
+
"rewards/rejected": -10.5625,
|
| 174 |
+
"step": 100
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"epoch": 0.7028753993610224,
|
| 178 |
+
"grad_norm": 0.09175324835066521,
|
| 179 |
+
"learning_rate": 4.251781472684085e-07,
|
| 180 |
+
"logits/chosen": -1.546875,
|
| 181 |
+
"logits/rejected": -1.421875,
|
| 182 |
+
"logps/chosen": -150.0,
|
| 183 |
+
"logps/rejected": -234.0,
|
| 184 |
+
"loss": 0.0011,
|
| 185 |
+
"rewards/accuracies": 1.0,
|
| 186 |
+
"rewards/chosen": -0.83203125,
|
| 187 |
+
"rewards/margins": 10.0625,
|
| 188 |
+
"rewards/rejected": -10.875,
|
| 189 |
+
"step": 110
|
| 190 |
+
},
|
| 191 |
+
{
|
| 192 |
+
"epoch": 0.7667731629392971,
|
| 193 |
+
"grad_norm": 0.06184587725303002,
|
| 194 |
+
"learning_rate": 4.1330166270783846e-07,
|
| 195 |
+
"logits/chosen": -1.546875,
|
| 196 |
+
"logits/rejected": -1.40625,
|
| 197 |
+
"logps/chosen": -145.0,
|
| 198 |
+
"logps/rejected": -240.0,
|
| 199 |
+
"loss": 0.0006,
|
| 200 |
+
"rewards/accuracies": 1.0,
|
| 201 |
+
"rewards/chosen": -0.89453125,
|
| 202 |
+
"rewards/margins": 10.375,
|
| 203 |
+
"rewards/rejected": -11.25,
|
| 204 |
+
"step": 120
|
| 205 |
+
},
|
| 206 |
+
{
|
| 207 |
+
"epoch": 0.8306709265175719,
|
| 208 |
+
"grad_norm": 0.024649453162613934,
|
| 209 |
+
"learning_rate": 4.0142517814726837e-07,
|
| 210 |
+
"logits/chosen": -1.5546875,
|
| 211 |
+
"logits/rejected": -1.4296875,
|
| 212 |
+
"logps/chosen": -150.0,
|
| 213 |
+
"logps/rejected": -242.0,
|
| 214 |
+
"loss": 0.0002,
|
| 215 |
+
"rewards/accuracies": 1.0,
|
| 216 |
+
"rewards/chosen": -0.859375,
|
| 217 |
+
"rewards/margins": 10.625,
|
| 218 |
+
"rewards/rejected": -11.5,
|
| 219 |
+
"step": 130
|
| 220 |
+
},
|
| 221 |
+
{
|
| 222 |
+
"epoch": 0.8945686900958466,
|
| 223 |
+
"grad_norm": 0.022941440244503526,
|
| 224 |
+
"learning_rate": 3.8954869358669834e-07,
|
| 225 |
+
"logits/chosen": -1.5703125,
|
| 226 |
+
"logits/rejected": -1.4453125,
|
| 227 |
+
"logps/chosen": -171.0,
|
| 228 |
+
"logps/rejected": -242.0,
|
| 229 |
+
"loss": 0.0005,
|
| 230 |
+
"rewards/accuracies": 1.0,
|
| 231 |
+
"rewards/chosen": -1.046875,
|
| 232 |
+
"rewards/margins": 10.375,
|
| 233 |
+
"rewards/rejected": -11.4375,
|
| 234 |
+
"step": 140
|
| 235 |
+
},
|
| 236 |
+
{
|
| 237 |
+
"epoch": 0.9584664536741214,
|
| 238 |
+
"grad_norm": 0.2502811167205942,
|
| 239 |
+
"learning_rate": 3.7767220902612825e-07,
|
| 240 |
+
"logits/chosen": -1.546875,
|
| 241 |
+
"logits/rejected": -1.4765625,
|
| 242 |
+
"logps/chosen": -160.0,
|
| 243 |
+
"logps/rejected": -250.0,
|
| 244 |
+
"loss": 0.0004,
|
| 245 |
+
"rewards/accuracies": 1.0,
|
| 246 |
+
"rewards/chosen": -0.984375,
|
| 247 |
+
"rewards/margins": 10.8125,
|
| 248 |
+
"rewards/rejected": -11.75,
|
| 249 |
+
"step": 150
|
| 250 |
+
},
|
| 251 |
+
{
|
| 252 |
+
"epoch": 0.9968051118210862,
|
| 253 |
+
"eval_logits/chosen": -1.4921875,
|
| 254 |
+
"eval_logits/rejected": -1.46875,
|
| 255 |
+
"eval_logps/chosen": -148.0,
|
| 256 |
+
"eval_logps/rejected": -250.0,
|
| 257 |
+
"eval_loss": 7.716297841398045e-05,
|
| 258 |
+
"eval_rewards/accuracies": 1.0,
|
| 259 |
+
"eval_rewards/chosen": -0.90625,
|
| 260 |
+
"eval_rewards/margins": 10.9375,
|
| 261 |
+
"eval_rewards/rejected": -11.8125,
|
| 262 |
+
"eval_runtime": 11.6151,
|
| 263 |
+
"eval_samples_per_second": 17.219,
|
| 264 |
+
"eval_steps_per_second": 0.603,
|
| 265 |
+
"step": 156
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 1.0223642172523961,
|
| 269 |
+
"grad_norm": 0.05686403061946404,
|
| 270 |
+
"learning_rate": 3.6579572446555817e-07,
|
| 271 |
+
"logits/chosen": -1.5859375,
|
| 272 |
+
"logits/rejected": -1.453125,
|
| 273 |
+
"logps/chosen": -174.0,
|
| 274 |
+
"logps/rejected": -249.0,
|
| 275 |
+
"loss": 0.0005,
|
| 276 |
+
"rewards/accuracies": 1.0,
|
| 277 |
+
"rewards/chosen": -0.95703125,
|
| 278 |
+
"rewards/margins": 10.875,
|
| 279 |
+
"rewards/rejected": -11.8125,
|
| 280 |
+
"step": 160
|
| 281 |
+
},
|
| 282 |
+
{
|
| 283 |
+
"epoch": 1.0862619808306708,
|
| 284 |
+
"grad_norm": 0.025647512662174864,
|
| 285 |
+
"learning_rate": 3.5391923990498813e-07,
|
| 286 |
+
"logits/chosen": -1.53125,
|
| 287 |
+
"logits/rejected": -1.4765625,
|
| 288 |
+
"logps/chosen": -151.0,
|
| 289 |
+
"logps/rejected": -253.0,
|
| 290 |
+
"loss": 0.0002,
|
| 291 |
+
"rewards/accuracies": 1.0,
|
| 292 |
+
"rewards/chosen": -0.890625,
|
| 293 |
+
"rewards/margins": 11.4375,
|
| 294 |
+
"rewards/rejected": -12.3125,
|
| 295 |
+
"step": 170
|
| 296 |
+
},
|
| 297 |
+
{
|
| 298 |
+
"epoch": 1.1501597444089458,
|
| 299 |
+
"grad_norm": 0.10600901838990953,
|
| 300 |
+
"learning_rate": 3.42042755344418e-07,
|
| 301 |
+
"logits/chosen": -1.5546875,
|
| 302 |
+
"logits/rejected": -1.46875,
|
| 303 |
+
"logps/chosen": -160.0,
|
| 304 |
+
"logps/rejected": -249.0,
|
| 305 |
+
"loss": 0.0003,
|
| 306 |
+
"rewards/accuracies": 1.0,
|
| 307 |
+
"rewards/chosen": -0.94140625,
|
| 308 |
+
"rewards/margins": 11.25,
|
| 309 |
+
"rewards/rejected": -12.25,
|
| 310 |
+
"step": 180
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"epoch": 1.2140575079872205,
|
| 314 |
+
"grad_norm": 0.13119830498432972,
|
| 315 |
+
"learning_rate": 3.3016627078384796e-07,
|
| 316 |
+
"logits/chosen": -1.53125,
|
| 317 |
+
"logits/rejected": -1.484375,
|
| 318 |
+
"logps/chosen": -146.0,
|
| 319 |
+
"logps/rejected": -251.0,
|
| 320 |
+
"loss": 0.0002,
|
| 321 |
+
"rewards/accuracies": 1.0,
|
| 322 |
+
"rewards/chosen": -0.88671875,
|
| 323 |
+
"rewards/margins": 11.4375,
|
| 324 |
+
"rewards/rejected": -12.3125,
|
| 325 |
+
"step": 190
|
| 326 |
+
},
|
| 327 |
+
{
|
| 328 |
+
"epoch": 1.2779552715654952,
|
| 329 |
+
"grad_norm": 0.030239526412913494,
|
| 330 |
+
"learning_rate": 3.182897862232779e-07,
|
| 331 |
+
"logits/chosen": -1.5390625,
|
| 332 |
+
"logits/rejected": -1.4609375,
|
| 333 |
+
"logps/chosen": -134.0,
|
| 334 |
+
"logps/rejected": -252.0,
|
| 335 |
+
"loss": 0.0003,
|
| 336 |
+
"rewards/accuracies": 1.0,
|
| 337 |
+
"rewards/chosen": -0.9453125,
|
| 338 |
+
"rewards/margins": 11.6875,
|
| 339 |
+
"rewards/rejected": -12.625,
|
| 340 |
+
"step": 200
|
| 341 |
+
},
|
| 342 |
+
{
|
| 343 |
+
"epoch": 1.34185303514377,
|
| 344 |
+
"grad_norm": 0.0080485091857848,
|
| 345 |
+
"learning_rate": 3.0641330166270784e-07,
|
| 346 |
+
"logits/chosen": -1.5859375,
|
| 347 |
+
"logits/rejected": -1.5078125,
|
| 348 |
+
"logps/chosen": -154.0,
|
| 349 |
+
"logps/rejected": -260.0,
|
| 350 |
+
"loss": 0.0002,
|
| 351 |
+
"rewards/accuracies": 1.0,
|
| 352 |
+
"rewards/chosen": -0.97265625,
|
| 353 |
+
"rewards/margins": 11.8125,
|
| 354 |
+
"rewards/rejected": -12.8125,
|
| 355 |
+
"step": 210
|
| 356 |
+
},
|
| 357 |
+
{
|
| 358 |
+
"epoch": 1.4057507987220448,
|
| 359 |
+
"grad_norm": 0.01503760669759659,
|
| 360 |
+
"learning_rate": 2.9453681710213776e-07,
|
| 361 |
+
"logits/chosen": -1.59375,
|
| 362 |
+
"logits/rejected": -1.5234375,
|
| 363 |
+
"logps/chosen": -154.0,
|
| 364 |
+
"logps/rejected": -256.0,
|
| 365 |
+
"loss": 0.0009,
|
| 366 |
+
"rewards/accuracies": 1.0,
|
| 367 |
+
"rewards/chosen": -0.890625,
|
| 368 |
+
"rewards/margins": 11.8125,
|
| 369 |
+
"rewards/rejected": -12.6875,
|
| 370 |
+
"step": 220
|
| 371 |
+
},
|
| 372 |
+
{
|
| 373 |
+
"epoch": 1.4696485623003195,
|
| 374 |
+
"grad_norm": 0.01745595116822664,
|
| 375 |
+
"learning_rate": 2.8266033254156767e-07,
|
| 376 |
+
"logits/chosen": -1.5703125,
|
| 377 |
+
"logits/rejected": -1.5078125,
|
| 378 |
+
"logps/chosen": -152.0,
|
| 379 |
+
"logps/rejected": -256.0,
|
| 380 |
+
"loss": 0.0004,
|
| 381 |
+
"rewards/accuracies": 1.0,
|
| 382 |
+
"rewards/chosen": -0.73828125,
|
| 383 |
+
"rewards/margins": 12.0,
|
| 384 |
+
"rewards/rejected": -12.75,
|
| 385 |
+
"step": 230
|
| 386 |
+
},
|
| 387 |
+
{
|
| 388 |
+
"epoch": 1.5335463258785942,
|
| 389 |
+
"grad_norm": 0.010639490905132864,
|
| 390 |
+
"learning_rate": 2.7078384798099764e-07,
|
| 391 |
+
"logits/chosen": -1.6015625,
|
| 392 |
+
"logits/rejected": -1.5,
|
| 393 |
+
"logps/chosen": -152.0,
|
| 394 |
+
"logps/rejected": -262.0,
|
| 395 |
+
"loss": 0.0001,
|
| 396 |
+
"rewards/accuracies": 1.0,
|
| 397 |
+
"rewards/chosen": -0.8359375,
|
| 398 |
+
"rewards/margins": 12.125,
|
| 399 |
+
"rewards/rejected": -13.0,
|
| 400 |
+
"step": 240
|
| 401 |
+
},
|
| 402 |
+
{
|
| 403 |
+
"epoch": 1.5974440894568689,
|
| 404 |
+
"grad_norm": 0.021325021450464145,
|
| 405 |
+
"learning_rate": 2.589073634204275e-07,
|
| 406 |
+
"logits/chosen": -1.53125,
|
| 407 |
+
"logits/rejected": -1.5,
|
| 408 |
+
"logps/chosen": -153.0,
|
| 409 |
+
"logps/rejected": -256.0,
|
| 410 |
+
"loss": 0.0005,
|
| 411 |
+
"rewards/accuracies": 1.0,
|
| 412 |
+
"rewards/chosen": -0.80859375,
|
| 413 |
+
"rewards/margins": 11.6875,
|
| 414 |
+
"rewards/rejected": -12.5,
|
| 415 |
+
"step": 250
|
| 416 |
+
},
|
| 417 |
+
{
|
| 418 |
+
"epoch": 1.6613418530351438,
|
| 419 |
+
"grad_norm": 0.03402727752531963,
|
| 420 |
+
"learning_rate": 2.4703087885985747e-07,
|
| 421 |
+
"logits/chosen": -1.546875,
|
| 422 |
+
"logits/rejected": -1.5234375,
|
| 423 |
+
"logps/chosen": -143.0,
|
| 424 |
+
"logps/rejected": -255.0,
|
| 425 |
+
"loss": 0.0002,
|
| 426 |
+
"rewards/accuracies": 1.0,
|
| 427 |
+
"rewards/chosen": -0.7265625,
|
| 428 |
+
"rewards/margins": 11.6875,
|
| 429 |
+
"rewards/rejected": -12.4375,
|
| 430 |
+
"step": 260
|
| 431 |
+
},
|
| 432 |
+
{
|
| 433 |
+
"epoch": 1.7252396166134185,
|
| 434 |
+
"grad_norm": 0.01735332750787422,
|
| 435 |
+
"learning_rate": 2.351543942992874e-07,
|
| 436 |
+
"logits/chosen": -1.6171875,
|
| 437 |
+
"logits/rejected": -1.5078125,
|
| 438 |
+
"logps/chosen": -165.0,
|
| 439 |
+
"logps/rejected": -260.0,
|
| 440 |
+
"loss": 0.0001,
|
| 441 |
+
"rewards/accuracies": 1.0,
|
| 442 |
+
"rewards/chosen": -0.80859375,
|
| 443 |
+
"rewards/margins": 12.1875,
|
| 444 |
+
"rewards/rejected": -13.0,
|
| 445 |
+
"step": 270
|
| 446 |
+
},
|
| 447 |
+
{
|
| 448 |
+
"epoch": 1.7891373801916934,
|
| 449 |
+
"grad_norm": 0.02817904426646656,
|
| 450 |
+
"learning_rate": 2.2327790973871732e-07,
|
| 451 |
+
"logits/chosen": -1.5703125,
|
| 452 |
+
"logits/rejected": -1.5078125,
|
| 453 |
+
"logps/chosen": -158.0,
|
| 454 |
+
"logps/rejected": -266.0,
|
| 455 |
+
"loss": 0.0001,
|
| 456 |
+
"rewards/accuracies": 1.0,
|
| 457 |
+
"rewards/chosen": -0.7734375,
|
| 458 |
+
"rewards/margins": 12.375,
|
| 459 |
+
"rewards/rejected": -13.125,
|
| 460 |
+
"step": 280
|
| 461 |
+
},
|
| 462 |
+
{
|
| 463 |
+
"epoch": 1.8530351437699681,
|
| 464 |
+
"grad_norm": 0.026646622883463183,
|
| 465 |
+
"learning_rate": 2.1140142517814726e-07,
|
| 466 |
+
"logits/chosen": -1.5859375,
|
| 467 |
+
"logits/rejected": -1.46875,
|
| 468 |
+
"logps/chosen": -154.0,
|
| 469 |
+
"logps/rejected": -258.0,
|
| 470 |
+
"loss": 0.0007,
|
| 471 |
+
"rewards/accuracies": 1.0,
|
| 472 |
+
"rewards/chosen": -0.90234375,
|
| 473 |
+
"rewards/margins": 11.9375,
|
| 474 |
+
"rewards/rejected": -12.875,
|
| 475 |
+
"step": 290
|
| 476 |
+
},
|
| 477 |
+
{
|
| 478 |
+
"epoch": 1.9169329073482428,
|
| 479 |
+
"grad_norm": 0.005464018856443713,
|
| 480 |
+
"learning_rate": 1.9952494061757718e-07,
|
| 481 |
+
"logits/chosen": -1.5546875,
|
| 482 |
+
"logits/rejected": -1.4765625,
|
| 483 |
+
"logps/chosen": -155.0,
|
| 484 |
+
"logps/rejected": -268.0,
|
| 485 |
+
"loss": 0.0001,
|
| 486 |
+
"rewards/accuracies": 1.0,
|
| 487 |
+
"rewards/chosen": -1.296875,
|
| 488 |
+
"rewards/margins": 12.3125,
|
| 489 |
+
"rewards/rejected": -13.625,
|
| 490 |
+
"step": 300
|
| 491 |
+
},
|
| 492 |
+
{
|
| 493 |
+
"epoch": 1.9808306709265175,
|
| 494 |
+
"grad_norm": 0.01607250092110866,
|
| 495 |
+
"learning_rate": 1.876484560570071e-07,
|
| 496 |
+
"logits/chosen": -1.5546875,
|
| 497 |
+
"logits/rejected": -1.4765625,
|
| 498 |
+
"logps/chosen": -152.0,
|
| 499 |
+
"logps/rejected": -268.0,
|
| 500 |
+
"loss": 0.0001,
|
| 501 |
+
"rewards/accuracies": 1.0,
|
| 502 |
+
"rewards/chosen": -1.34375,
|
| 503 |
+
"rewards/margins": 12.5625,
|
| 504 |
+
"rewards/rejected": -13.875,
|
| 505 |
+
"step": 310
|
| 506 |
+
},
|
| 507 |
+
{
|
| 508 |
+
"epoch": 2.0,
|
| 509 |
+
"eval_logits/chosen": -1.5078125,
|
| 510 |
+
"eval_logits/rejected": -1.515625,
|
| 511 |
+
"eval_logps/chosen": -151.0,
|
| 512 |
+
"eval_logps/rejected": -266.0,
|
| 513 |
+
"eval_loss": 1.9190907551092096e-05,
|
| 514 |
+
"eval_rewards/accuracies": 1.0,
|
| 515 |
+
"eval_rewards/chosen": -1.1640625,
|
| 516 |
+
"eval_rewards/margins": 12.3125,
|
| 517 |
+
"eval_rewards/rejected": -13.4375,
|
| 518 |
+
"eval_runtime": 11.5575,
|
| 519 |
+
"eval_samples_per_second": 17.305,
|
| 520 |
+
"eval_steps_per_second": 0.606,
|
| 521 |
+
"step": 313
|
| 522 |
+
},
|
| 523 |
+
{
|
| 524 |
+
"epoch": 2.0447284345047922,
|
| 525 |
+
"grad_norm": 0.01361688937176726,
|
| 526 |
+
"learning_rate": 1.7577197149643706e-07,
|
| 527 |
+
"logits/chosen": -1.578125,
|
| 528 |
+
"logits/rejected": -1.53125,
|
| 529 |
+
"logps/chosen": -157.0,
|
| 530 |
+
"logps/rejected": -270.0,
|
| 531 |
+
"loss": 0.0,
|
| 532 |
+
"rewards/accuracies": 1.0,
|
| 533 |
+
"rewards/chosen": -1.203125,
|
| 534 |
+
"rewards/margins": 12.6875,
|
| 535 |
+
"rewards/rejected": -13.9375,
|
| 536 |
+
"step": 320
|
| 537 |
+
},
|
| 538 |
+
{
|
| 539 |
+
"epoch": 2.108626198083067,
|
| 540 |
+
"grad_norm": 0.05503143614977478,
|
| 541 |
+
"learning_rate": 1.6389548693586697e-07,
|
| 542 |
+
"logits/chosen": -1.5859375,
|
| 543 |
+
"logits/rejected": -1.5,
|
| 544 |
+
"logps/chosen": -162.0,
|
| 545 |
+
"logps/rejected": -266.0,
|
| 546 |
+
"loss": 0.0002,
|
| 547 |
+
"rewards/accuracies": 1.0,
|
| 548 |
+
"rewards/chosen": -1.1796875,
|
| 549 |
+
"rewards/margins": 12.375,
|
| 550 |
+
"rewards/rejected": -13.5625,
|
| 551 |
+
"step": 330
|
| 552 |
+
},
|
| 553 |
+
{
|
| 554 |
+
"epoch": 2.1725239616613417,
|
| 555 |
+
"grad_norm": 0.010269985783571838,
|
| 556 |
+
"learning_rate": 1.520190023752969e-07,
|
| 557 |
+
"logits/chosen": -1.53125,
|
| 558 |
+
"logits/rejected": -1.5234375,
|
| 559 |
+
"logps/chosen": -152.0,
|
| 560 |
+
"logps/rejected": -268.0,
|
| 561 |
+
"loss": 0.0001,
|
| 562 |
+
"rewards/accuracies": 1.0,
|
| 563 |
+
"rewards/chosen": -1.078125,
|
| 564 |
+
"rewards/margins": 12.75,
|
| 565 |
+
"rewards/rejected": -13.8125,
|
| 566 |
+
"step": 340
|
| 567 |
+
},
|
| 568 |
+
{
|
| 569 |
+
"epoch": 2.236421725239617,
|
| 570 |
+
"grad_norm": 0.0683919347457657,
|
| 571 |
+
"learning_rate": 1.4014251781472683e-07,
|
| 572 |
+
"logits/chosen": -1.5625,
|
| 573 |
+
"logits/rejected": -1.4921875,
|
| 574 |
+
"logps/chosen": -156.0,
|
| 575 |
+
"logps/rejected": -266.0,
|
| 576 |
+
"loss": 0.0001,
|
| 577 |
+
"rewards/accuracies": 1.0,
|
| 578 |
+
"rewards/chosen": -1.09375,
|
| 579 |
+
"rewards/margins": 12.75,
|
| 580 |
+
"rewards/rejected": -13.8125,
|
| 581 |
+
"step": 350
|
| 582 |
+
},
|
| 583 |
+
{
|
| 584 |
+
"epoch": 2.3003194888178915,
|
| 585 |
+
"grad_norm": 0.015461126488895941,
|
| 586 |
+
"learning_rate": 1.2826603325415677e-07,
|
| 587 |
+
"logits/chosen": -1.5546875,
|
| 588 |
+
"logits/rejected": -1.546875,
|
| 589 |
+
"logps/chosen": -146.0,
|
| 590 |
+
"logps/rejected": -266.0,
|
| 591 |
+
"loss": 0.0001,
|
| 592 |
+
"rewards/accuracies": 1.0,
|
| 593 |
+
"rewards/chosen": -1.1015625,
|
| 594 |
+
"rewards/margins": 12.625,
|
| 595 |
+
"rewards/rejected": -13.75,
|
| 596 |
+
"step": 360
|
| 597 |
+
},
|
| 598 |
+
{
|
| 599 |
+
"epoch": 2.364217252396166,
|
| 600 |
+
"grad_norm": 0.004666909345169454,
|
| 601 |
+
"learning_rate": 1.163895486935867e-07,
|
| 602 |
+
"logits/chosen": -1.5390625,
|
| 603 |
+
"logits/rejected": -1.4921875,
|
| 604 |
+
"logps/chosen": -160.0,
|
| 605 |
+
"logps/rejected": -270.0,
|
| 606 |
+
"loss": 0.0002,
|
| 607 |
+
"rewards/accuracies": 1.0,
|
| 608 |
+
"rewards/chosen": -1.1328125,
|
| 609 |
+
"rewards/margins": 12.8125,
|
| 610 |
+
"rewards/rejected": -13.9375,
|
| 611 |
+
"step": 370
|
| 612 |
+
},
|
| 613 |
+
{
|
| 614 |
+
"epoch": 2.428115015974441,
|
| 615 |
+
"grad_norm": 0.021768817881518163,
|
| 616 |
+
"learning_rate": 1.0451306413301662e-07,
|
| 617 |
+
"logits/chosen": -1.5859375,
|
| 618 |
+
"logits/rejected": -1.5078125,
|
| 619 |
+
"logps/chosen": -157.0,
|
| 620 |
+
"logps/rejected": -266.0,
|
| 621 |
+
"loss": 0.0001,
|
| 622 |
+
"rewards/accuracies": 1.0,
|
| 623 |
+
"rewards/chosen": -1.1484375,
|
| 624 |
+
"rewards/margins": 12.8125,
|
| 625 |
+
"rewards/rejected": -14.0,
|
| 626 |
+
"step": 380
|
| 627 |
+
},
|
| 628 |
+
{
|
| 629 |
+
"epoch": 2.4920127795527156,
|
| 630 |
+
"grad_norm": 0.006712968496189408,
|
| 631 |
+
"learning_rate": 9.263657957244655e-08,
|
| 632 |
+
"logits/chosen": -1.578125,
|
| 633 |
+
"logits/rejected": -1.4765625,
|
| 634 |
+
"logps/chosen": -150.0,
|
| 635 |
+
"logps/rejected": -272.0,
|
| 636 |
+
"loss": 0.0002,
|
| 637 |
+
"rewards/accuracies": 1.0,
|
| 638 |
+
"rewards/chosen": -1.046875,
|
| 639 |
+
"rewards/margins": 13.125,
|
| 640 |
+
"rewards/rejected": -14.125,
|
| 641 |
+
"step": 390
|
| 642 |
+
},
|
| 643 |
+
{
|
| 644 |
+
"epoch": 2.5559105431309903,
|
| 645 |
+
"grad_norm": 0.03157363956222121,
|
| 646 |
+
"learning_rate": 8.076009501187649e-08,
|
| 647 |
+
"logits/chosen": -1.578125,
|
| 648 |
+
"logits/rejected": -1.515625,
|
| 649 |
+
"logps/chosen": -160.0,
|
| 650 |
+
"logps/rejected": -268.0,
|
| 651 |
+
"loss": 0.0001,
|
| 652 |
+
"rewards/accuracies": 1.0,
|
| 653 |
+
"rewards/chosen": -1.0625,
|
| 654 |
+
"rewards/margins": 13.0625,
|
| 655 |
+
"rewards/rejected": -14.125,
|
| 656 |
+
"step": 400
|
| 657 |
+
},
|
| 658 |
+
{
|
| 659 |
+
"epoch": 2.619808306709265,
|
| 660 |
+
"grad_norm": 0.012508580641767728,
|
| 661 |
+
"learning_rate": 6.88836104513064e-08,
|
| 662 |
+
"logits/chosen": -1.5390625,
|
| 663 |
+
"logits/rejected": -1.5,
|
| 664 |
+
"logps/chosen": -151.0,
|
| 665 |
+
"logps/rejected": -268.0,
|
| 666 |
+
"loss": 0.0001,
|
| 667 |
+
"rewards/accuracies": 1.0,
|
| 668 |
+
"rewards/chosen": -1.03125,
|
| 669 |
+
"rewards/margins": 12.9375,
|
| 670 |
+
"rewards/rejected": -13.9375,
|
| 671 |
+
"step": 410
|
| 672 |
+
},
|
| 673 |
+
{
|
| 674 |
+
"epoch": 2.68370607028754,
|
| 675 |
+
"grad_norm": 0.014858628369684351,
|
| 676 |
+
"learning_rate": 5.700712589073634e-08,
|
| 677 |
+
"logits/chosen": -1.5546875,
|
| 678 |
+
"logits/rejected": -1.515625,
|
| 679 |
+
"logps/chosen": -159.0,
|
| 680 |
+
"logps/rejected": -268.0,
|
| 681 |
+
"loss": 0.0001,
|
| 682 |
+
"rewards/accuracies": 1.0,
|
| 683 |
+
"rewards/chosen": -1.171875,
|
| 684 |
+
"rewards/margins": 12.8125,
|
| 685 |
+
"rewards/rejected": -14.0,
|
| 686 |
+
"step": 420
|
| 687 |
+
},
|
| 688 |
+
{
|
| 689 |
+
"epoch": 2.747603833865815,
|
| 690 |
+
"grad_norm": 0.00119033851952325,
|
| 691 |
+
"learning_rate": 4.5130641330166267e-08,
|
| 692 |
+
"logits/chosen": -1.59375,
|
| 693 |
+
"logits/rejected": -1.46875,
|
| 694 |
+
"logps/chosen": -165.0,
|
| 695 |
+
"logps/rejected": -270.0,
|
| 696 |
+
"loss": 0.0001,
|
| 697 |
+
"rewards/accuracies": 1.0,
|
| 698 |
+
"rewards/chosen": -1.0703125,
|
| 699 |
+
"rewards/margins": 13.25,
|
| 700 |
+
"rewards/rejected": -14.3125,
|
| 701 |
+
"step": 430
|
| 702 |
+
},
|
| 703 |
+
{
|
| 704 |
+
"epoch": 2.8115015974440896,
|
| 705 |
+
"grad_norm": 0.01000781639917085,
|
| 706 |
+
"learning_rate": 3.32541567695962e-08,
|
| 707 |
+
"logits/chosen": -1.5390625,
|
| 708 |
+
"logits/rejected": -1.5234375,
|
| 709 |
+
"logps/chosen": -142.0,
|
| 710 |
+
"logps/rejected": -270.0,
|
| 711 |
+
"loss": 0.0001,
|
| 712 |
+
"rewards/accuracies": 1.0,
|
| 713 |
+
"rewards/chosen": -1.0234375,
|
| 714 |
+
"rewards/margins": 13.375,
|
| 715 |
+
"rewards/rejected": -14.375,
|
| 716 |
+
"step": 440
|
| 717 |
+
},
|
| 718 |
+
{
|
| 719 |
+
"epoch": 2.8753993610223643,
|
| 720 |
+
"grad_norm": 0.032333252949445446,
|
| 721 |
+
"learning_rate": 2.1377672209026125e-08,
|
| 722 |
+
"logits/chosen": -1.5546875,
|
| 723 |
+
"logits/rejected": -1.5078125,
|
| 724 |
+
"logps/chosen": -159.0,
|
| 725 |
+
"logps/rejected": -270.0,
|
| 726 |
+
"loss": 0.0,
|
| 727 |
+
"rewards/accuracies": 1.0,
|
| 728 |
+
"rewards/chosen": -1.078125,
|
| 729 |
+
"rewards/margins": 13.3125,
|
| 730 |
+
"rewards/rejected": -14.375,
|
| 731 |
+
"step": 450
|
| 732 |
+
},
|
| 733 |
+
{
|
| 734 |
+
"epoch": 2.939297124600639,
|
| 735 |
+
"grad_norm": 0.0060308968638626,
|
| 736 |
+
"learning_rate": 9.501187648456057e-09,
|
| 737 |
+
"logits/chosen": -1.578125,
|
| 738 |
+
"logits/rejected": -1.4921875,
|
| 739 |
+
"logps/chosen": -153.0,
|
| 740 |
+
"logps/rejected": -270.0,
|
| 741 |
+
"loss": 0.0001,
|
| 742 |
+
"rewards/accuracies": 1.0,
|
| 743 |
+
"rewards/chosen": -1.0390625,
|
| 744 |
+
"rewards/margins": 13.3125,
|
| 745 |
+
"rewards/rejected": -14.375,
|
| 746 |
+
"step": 460
|
| 747 |
+
},
|
| 748 |
+
{
|
| 749 |
+
"epoch": 2.9904153354632586,
|
| 750 |
+
"eval_logits/chosen": -1.5078125,
|
| 751 |
+
"eval_logits/rejected": -1.5234375,
|
| 752 |
+
"eval_logps/chosen": -150.0,
|
| 753 |
+
"eval_logps/rejected": -270.0,
|
| 754 |
+
"eval_loss": 1.3302266779646743e-05,
|
| 755 |
+
"eval_rewards/accuracies": 1.0,
|
| 756 |
+
"eval_rewards/chosen": -1.09375,
|
| 757 |
+
"eval_rewards/margins": 12.8125,
|
| 758 |
+
"eval_rewards/rejected": -13.9375,
|
| 759 |
+
"eval_runtime": 14.4196,
|
| 760 |
+
"eval_samples_per_second": 13.87,
|
| 761 |
+
"eval_steps_per_second": 0.485,
|
| 762 |
+
"step": 468
|
| 763 |
+
},
|
| 764 |
+
{
|
| 765 |
+
"epoch": 2.9904153354632586,
|
| 766 |
+
"step": 468,
|
| 767 |
+
"total_flos": 0.0,
|
| 768 |
+
"train_loss": 0.036010837255037814,
|
| 769 |
+
"train_runtime": 4360.4368,
|
| 770 |
+
"train_samples_per_second": 6.88,
|
| 771 |
+
"train_steps_per_second": 0.107
|
| 772 |
+
}
|
| 773 |
+
],
|
| 774 |
+
"logging_steps": 10,
|
| 775 |
+
"max_steps": 468,
|
| 776 |
+
"num_input_tokens_seen": 0,
|
| 777 |
+
"num_train_epochs": 3,
|
| 778 |
+
"save_steps": 500,
|
| 779 |
+
"stateful_callbacks": {
|
| 780 |
+
"TrainerControl": {
|
| 781 |
+
"args": {
|
| 782 |
+
"should_epoch_stop": false,
|
| 783 |
+
"should_evaluate": false,
|
| 784 |
+
"should_log": false,
|
| 785 |
+
"should_save": true,
|
| 786 |
+
"should_training_stop": true
|
| 787 |
+
},
|
| 788 |
+
"attributes": {}
|
| 789 |
+
}
|
| 790 |
+
},
|
| 791 |
+
"total_flos": 0.0,
|
| 792 |
+
"train_batch_size": 8,
|
| 793 |
+
"trial_name": null,
|
| 794 |
+
"trial_params": null
|
| 795 |
+
}
|