Model save
Browse files- README.md +67 -0
- all_results.json +9 -0
- train_results.json +9 -0
- trainer_state.json +651 -0
README.md
ADDED
|
@@ -0,0 +1,67 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
base_model: google/gemma-7b
|
| 3 |
+
library_name: transformers
|
| 4 |
+
model_name: gemma-7b-silvia_cpo-basic_capibara-5e-5-025-v151
|
| 5 |
+
tags:
|
| 6 |
+
- generated_from_trainer
|
| 7 |
+
- trl
|
| 8 |
+
- orpo
|
| 9 |
+
licence: license
|
| 10 |
+
---
|
| 11 |
+
|
| 12 |
+
# Model Card for gemma-7b-silvia_cpo-basic_capibara-5e-5-025-v151
|
| 13 |
+
|
| 14 |
+
This model is a fine-tuned version of [google/gemma-7b](https://huggingface.co/google/gemma-7b).
|
| 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="silviasapora/gemma-7b-silvia_cpo-basic_capibara-5e-5-025-v151", 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/silvias/huggingface/runs/mw5gj0gs)
|
| 31 |
+
|
| 32 |
+
|
| 33 |
+
This model was trained with ORPO, a method introduced in [ORPO: Monolithic Preference Optimization without Reference Model](https://huggingface.co/papers/2403.07691).
|
| 34 |
+
|
| 35 |
+
### Framework versions
|
| 36 |
+
|
| 37 |
+
- TRL: 0.16.0
|
| 38 |
+
- Transformers: 4.50.3
|
| 39 |
+
- Pytorch: 2.5.1
|
| 40 |
+
- Datasets: 3.3.2
|
| 41 |
+
- Tokenizers: 0.21.0
|
| 42 |
+
|
| 43 |
+
## Citations
|
| 44 |
+
|
| 45 |
+
Cite ORPO as:
|
| 46 |
+
|
| 47 |
+
```bibtex
|
| 48 |
+
@article{hong2024orpo,
|
| 49 |
+
title = {{ORPO: Monolithic Preference Optimization without Reference Model}},
|
| 50 |
+
author = {Jiwoo Hong and Noah Lee and James Thorne},
|
| 51 |
+
year = 2024,
|
| 52 |
+
eprint = {arXiv:2403.07691}
|
| 53 |
+
}
|
| 54 |
+
```
|
| 55 |
+
|
| 56 |
+
Cite TRL as:
|
| 57 |
+
|
| 58 |
+
```bibtex
|
| 59 |
+
@misc{vonwerra2022trl,
|
| 60 |
+
title = {{TRL: Transformer Reinforcement Learning}},
|
| 61 |
+
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},
|
| 62 |
+
year = 2020,
|
| 63 |
+
journal = {GitHub repository},
|
| 64 |
+
publisher = {GitHub},
|
| 65 |
+
howpublished = {\url{https://github.com/huggingface/trl}}
|
| 66 |
+
}
|
| 67 |
+
```
|
all_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.9814696485623005,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 62.26987236738205,
|
| 5 |
+
"train_runtime": 5592.1515,
|
| 6 |
+
"train_samples": 3130,
|
| 7 |
+
"train_samples_per_second": 2.239,
|
| 8 |
+
"train_steps_per_second": 0.034
|
| 9 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.9814696485623005,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 62.26987236738205,
|
| 5 |
+
"train_runtime": 5592.1515,
|
| 6 |
+
"train_samples": 3130,
|
| 7 |
+
"train_samples_per_second": 2.239,
|
| 8 |
+
"train_steps_per_second": 0.034
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,651 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_global_step": null,
|
| 3 |
+
"best_metric": null,
|
| 4 |
+
"best_model_checkpoint": null,
|
| 5 |
+
"epoch": 3.9814696485623005,
|
| 6 |
+
"eval_steps": 500,
|
| 7 |
+
"global_step": 192,
|
| 8 |
+
"is_hyper_param_search": false,
|
| 9 |
+
"is_local_process_zero": true,
|
| 10 |
+
"is_world_process_zero": true,
|
| 11 |
+
"log_history": [
|
| 12 |
+
{
|
| 13 |
+
"epoch": 0.10223642172523961,
|
| 14 |
+
"grad_norm": 827.7908935546875,
|
| 15 |
+
"learning_rate": 1.25e-05,
|
| 16 |
+
"log_odds_chosen": -4.91914176940918,
|
| 17 |
+
"log_odds_ratio": -7.912337303161621,
|
| 18 |
+
"logps/chosen": -17.909399032592773,
|
| 19 |
+
"logps/rejected": -12.991429328918457,
|
| 20 |
+
"loss": 404.4927,
|
| 21 |
+
"nll_loss": 10.663026809692383,
|
| 22 |
+
"rewards/accuracies": 0.40312498807907104,
|
| 23 |
+
"rewards/chosen": -4.477349758148193,
|
| 24 |
+
"rewards/margins": -1.2294929027557373,
|
| 25 |
+
"rewards/rejected": -3.2478573322296143,
|
| 26 |
+
"step": 5
|
| 27 |
+
},
|
| 28 |
+
{
|
| 29 |
+
"epoch": 0.20447284345047922,
|
| 30 |
+
"grad_norm": 538.9002685546875,
|
| 31 |
+
"learning_rate": 2.5e-05,
|
| 32 |
+
"log_odds_chosen": -1.99321711063385,
|
| 33 |
+
"log_odds_ratio": -5.780060291290283,
|
| 34 |
+
"logps/chosen": -15.17625617980957,
|
| 35 |
+
"logps/rejected": -13.184392929077148,
|
| 36 |
+
"loss": 342.9983,
|
| 37 |
+
"nll_loss": 9.274579048156738,
|
| 38 |
+
"rewards/accuracies": 0.4156250059604645,
|
| 39 |
+
"rewards/chosen": -3.7940640449523926,
|
| 40 |
+
"rewards/margins": -0.4979652762413025,
|
| 41 |
+
"rewards/rejected": -3.296098232269287,
|
| 42 |
+
"step": 10
|
| 43 |
+
},
|
| 44 |
+
{
|
| 45 |
+
"epoch": 0.30670926517571884,
|
| 46 |
+
"grad_norm": 1268.311279296875,
|
| 47 |
+
"learning_rate": 3.7500000000000003e-05,
|
| 48 |
+
"log_odds_chosen": -1.9389375448226929,
|
| 49 |
+
"log_odds_ratio": -5.051381587982178,
|
| 50 |
+
"logps/chosen": -12.37875747680664,
|
| 51 |
+
"logps/rejected": -10.439730644226074,
|
| 52 |
+
"loss": 271.8747,
|
| 53 |
+
"nll_loss": 7.234231472015381,
|
| 54 |
+
"rewards/accuracies": 0.484375,
|
| 55 |
+
"rewards/chosen": -3.09468936920166,
|
| 56 |
+
"rewards/margins": -0.4847566485404968,
|
| 57 |
+
"rewards/rejected": -2.6099326610565186,
|
| 58 |
+
"step": 15
|
| 59 |
+
},
|
| 60 |
+
{
|
| 61 |
+
"epoch": 0.40894568690095845,
|
| 62 |
+
"grad_norm": 500.36431884765625,
|
| 63 |
+
"learning_rate": 5e-05,
|
| 64 |
+
"log_odds_chosen": -0.3404741883277893,
|
| 65 |
+
"log_odds_ratio": -1.416075348854065,
|
| 66 |
+
"logps/chosen": -3.7283692359924316,
|
| 67 |
+
"logps/rejected": -3.374943494796753,
|
| 68 |
+
"loss": 108.1039,
|
| 69 |
+
"nll_loss": 3.0280256271362305,
|
| 70 |
+
"rewards/accuracies": 0.528124988079071,
|
| 71 |
+
"rewards/chosen": -0.9320923089981079,
|
| 72 |
+
"rewards/margins": -0.08835643529891968,
|
| 73 |
+
"rewards/rejected": -0.8437358736991882,
|
| 74 |
+
"step": 20
|
| 75 |
+
},
|
| 76 |
+
{
|
| 77 |
+
"epoch": 0.5111821086261981,
|
| 78 |
+
"grad_norm": 416.5755615234375,
|
| 79 |
+
"learning_rate": 4.989581840408562e-05,
|
| 80 |
+
"log_odds_chosen": 0.25642240047454834,
|
| 81 |
+
"log_odds_ratio": -0.655372142791748,
|
| 82 |
+
"logps/chosen": -1.5129865407943726,
|
| 83 |
+
"logps/rejected": -1.724565863609314,
|
| 84 |
+
"loss": 61.4784,
|
| 85 |
+
"nll_loss": 1.7581512928009033,
|
| 86 |
+
"rewards/accuracies": 0.643750011920929,
|
| 87 |
+
"rewards/chosen": -0.37824663519859314,
|
| 88 |
+
"rewards/margins": 0.052894867956638336,
|
| 89 |
+
"rewards/rejected": -0.4311414659023285,
|
| 90 |
+
"step": 25
|
| 91 |
+
},
|
| 92 |
+
{
|
| 93 |
+
"epoch": 0.6134185303514377,
|
| 94 |
+
"grad_norm": 166.52938842773438,
|
| 95 |
+
"learning_rate": 4.9584141920736656e-05,
|
| 96 |
+
"log_odds_chosen": 0.1395658254623413,
|
| 97 |
+
"log_odds_ratio": -0.7378019094467163,
|
| 98 |
+
"logps/chosen": -1.480513334274292,
|
| 99 |
+
"logps/rejected": -1.599832534790039,
|
| 100 |
+
"loss": 59.0905,
|
| 101 |
+
"nll_loss": 1.6694732904434204,
|
| 102 |
+
"rewards/accuracies": 0.612500011920929,
|
| 103 |
+
"rewards/chosen": -0.370128333568573,
|
| 104 |
+
"rewards/margins": 0.02982982061803341,
|
| 105 |
+
"rewards/rejected": -0.39995813369750977,
|
| 106 |
+
"step": 30
|
| 107 |
+
},
|
| 108 |
+
{
|
| 109 |
+
"epoch": 0.7156549520766773,
|
| 110 |
+
"grad_norm": 81.22210693359375,
|
| 111 |
+
"learning_rate": 4.9067568226228646e-05,
|
| 112 |
+
"log_odds_chosen": 0.23929011821746826,
|
| 113 |
+
"log_odds_ratio": -0.6636127233505249,
|
| 114 |
+
"logps/chosen": -1.3320153951644897,
|
| 115 |
+
"logps/rejected": -1.5101702213287354,
|
| 116 |
+
"loss": 52.0935,
|
| 117 |
+
"nll_loss": 1.4632833003997803,
|
| 118 |
+
"rewards/accuracies": 0.6343749761581421,
|
| 119 |
+
"rewards/chosen": -0.33300384879112244,
|
| 120 |
+
"rewards/margins": 0.04453865438699722,
|
| 121 |
+
"rewards/rejected": -0.37754255533218384,
|
| 122 |
+
"step": 35
|
| 123 |
+
},
|
| 124 |
+
{
|
| 125 |
+
"epoch": 0.8178913738019169,
|
| 126 |
+
"grad_norm": 102.73954010009766,
|
| 127 |
+
"learning_rate": 4.83504027183137e-05,
|
| 128 |
+
"log_odds_chosen": 0.3248418867588043,
|
| 129 |
+
"log_odds_ratio": -0.6250384449958801,
|
| 130 |
+
"logps/chosen": -1.249929666519165,
|
| 131 |
+
"logps/rejected": -1.4835537672042847,
|
| 132 |
+
"loss": 50.6161,
|
| 133 |
+
"nll_loss": 1.4245882034301758,
|
| 134 |
+
"rewards/accuracies": 0.684374988079071,
|
| 135 |
+
"rewards/chosen": -0.31248241662979126,
|
| 136 |
+
"rewards/margins": 0.05840602517127991,
|
| 137 |
+
"rewards/rejected": -0.37088844180107117,
|
| 138 |
+
"step": 40
|
| 139 |
+
},
|
| 140 |
+
{
|
| 141 |
+
"epoch": 0.9201277955271565,
|
| 142 |
+
"grad_norm": 103.2927474975586,
|
| 143 |
+
"learning_rate": 4.743862263276376e-05,
|
| 144 |
+
"log_odds_chosen": 0.30287498235702515,
|
| 145 |
+
"log_odds_ratio": -0.6419548988342285,
|
| 146 |
+
"logps/chosen": -1.2506297826766968,
|
| 147 |
+
"logps/rejected": -1.4725720882415771,
|
| 148 |
+
"loss": 49.8597,
|
| 149 |
+
"nll_loss": 1.3988924026489258,
|
| 150 |
+
"rewards/accuracies": 0.6781250238418579,
|
| 151 |
+
"rewards/chosen": -0.3126574456691742,
|
| 152 |
+
"rewards/margins": 0.0554855652153492,
|
| 153 |
+
"rewards/rejected": -0.3681430220603943,
|
| 154 |
+
"step": 45
|
| 155 |
+
},
|
| 156 |
+
{
|
| 157 |
+
"epoch": 1.040894568690096,
|
| 158 |
+
"grad_norm": 152.4938201904297,
|
| 159 |
+
"learning_rate": 4.6339827225933665e-05,
|
| 160 |
+
"log_odds_chosen": 0.39640626311302185,
|
| 161 |
+
"log_odds_ratio": -0.6115574836730957,
|
| 162 |
+
"logps/chosen": -1.1695352792739868,
|
| 163 |
+
"logps/rejected": -1.471874713897705,
|
| 164 |
+
"loss": 54.7336,
|
| 165 |
+
"nll_loss": 1.3199635744094849,
|
| 166 |
+
"rewards/accuracies": 0.6935483813285828,
|
| 167 |
+
"rewards/chosen": -0.2923838198184967,
|
| 168 |
+
"rewards/margins": 0.07558488100767136,
|
| 169 |
+
"rewards/rejected": -0.36796867847442627,
|
| 170 |
+
"step": 50
|
| 171 |
+
},
|
| 172 |
+
{
|
| 173 |
+
"epoch": 1.1431309904153355,
|
| 174 |
+
"grad_norm": 89.43785858154297,
|
| 175 |
+
"learning_rate": 4.506317443854877e-05,
|
| 176 |
+
"log_odds_chosen": 0.5089691877365112,
|
| 177 |
+
"log_odds_ratio": -0.5568146705627441,
|
| 178 |
+
"logps/chosen": -1.0600959062576294,
|
| 179 |
+
"logps/rejected": -1.4196648597717285,
|
| 180 |
+
"loss": 43.6744,
|
| 181 |
+
"nll_loss": 1.2228777408599854,
|
| 182 |
+
"rewards/accuracies": 0.75,
|
| 183 |
+
"rewards/chosen": -0.26502397656440735,
|
| 184 |
+
"rewards/margins": 0.08989225327968597,
|
| 185 |
+
"rewards/rejected": -0.35491621494293213,
|
| 186 |
+
"step": 55
|
| 187 |
+
},
|
| 188 |
+
{
|
| 189 |
+
"epoch": 1.2453674121405751,
|
| 190 |
+
"grad_norm": 87.75115203857422,
|
| 191 |
+
"learning_rate": 4.361930456859455e-05,
|
| 192 |
+
"log_odds_chosen": 0.7825835943222046,
|
| 193 |
+
"log_odds_ratio": -0.47958317399024963,
|
| 194 |
+
"logps/chosen": -0.9152556657791138,
|
| 195 |
+
"logps/rejected": -1.431404948234558,
|
| 196 |
+
"loss": 38.5085,
|
| 197 |
+
"nll_loss": 1.072222352027893,
|
| 198 |
+
"rewards/accuracies": 0.8031250238418579,
|
| 199 |
+
"rewards/chosen": -0.22881391644477844,
|
| 200 |
+
"rewards/margins": 0.12903733551502228,
|
| 201 |
+
"rewards/rejected": -0.3578512370586395,
|
| 202 |
+
"step": 60
|
| 203 |
+
},
|
| 204 |
+
{
|
| 205 |
+
"epoch": 1.3476038338658147,
|
| 206 |
+
"grad_norm": 184.32460021972656,
|
| 207 |
+
"learning_rate": 4.202025158945856e-05,
|
| 208 |
+
"log_odds_chosen": 0.6628133654594421,
|
| 209 |
+
"log_odds_ratio": -0.5079701542854309,
|
| 210 |
+
"logps/chosen": -0.9785585403442383,
|
| 211 |
+
"logps/rejected": -1.419032335281372,
|
| 212 |
+
"loss": 40.1897,
|
| 213 |
+
"nll_loss": 1.1202460527420044,
|
| 214 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 215 |
+
"rewards/chosen": -0.24463963508605957,
|
| 216 |
+
"rewards/margins": 0.11011841148138046,
|
| 217 |
+
"rewards/rejected": -0.354758083820343,
|
| 218 |
+
"step": 65
|
| 219 |
+
},
|
| 220 |
+
{
|
| 221 |
+
"epoch": 1.4498402555910543,
|
| 222 |
+
"grad_norm": 270.92669677734375,
|
| 223 |
+
"learning_rate": 4.0279342852446234e-05,
|
| 224 |
+
"log_odds_chosen": 0.6627331972122192,
|
| 225 |
+
"log_odds_ratio": -0.529289722442627,
|
| 226 |
+
"logps/chosen": -0.9792484045028687,
|
| 227 |
+
"logps/rejected": -1.4207617044448853,
|
| 228 |
+
"loss": 39.7387,
|
| 229 |
+
"nll_loss": 1.1052920818328857,
|
| 230 |
+
"rewards/accuracies": 0.768750011920929,
|
| 231 |
+
"rewards/chosen": -0.24481210112571716,
|
| 232 |
+
"rewards/margins": 0.11037830263376236,
|
| 233 |
+
"rewards/rejected": -0.3551904261112213,
|
| 234 |
+
"step": 70
|
| 235 |
+
},
|
| 236 |
+
{
|
| 237 |
+
"epoch": 1.552076677316294,
|
| 238 |
+
"grad_norm": 75.13140869140625,
|
| 239 |
+
"learning_rate": 3.8411088009602646e-05,
|
| 240 |
+
"log_odds_chosen": 0.7180580496788025,
|
| 241 |
+
"log_odds_ratio": -0.49540311098098755,
|
| 242 |
+
"logps/chosen": -0.9218428730964661,
|
| 243 |
+
"logps/rejected": -1.3873802423477173,
|
| 244 |
+
"loss": 38.359,
|
| 245 |
+
"nll_loss": 1.0657460689544678,
|
| 246 |
+
"rewards/accuracies": 0.793749988079071,
|
| 247 |
+
"rewards/chosen": -0.23046071827411652,
|
| 248 |
+
"rewards/margins": 0.11638432741165161,
|
| 249 |
+
"rewards/rejected": -0.3468450605869293,
|
| 250 |
+
"step": 75
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"epoch": 1.6543130990415336,
|
| 254 |
+
"grad_norm": 76.11426544189453,
|
| 255 |
+
"learning_rate": 3.6431058082615964e-05,
|
| 256 |
+
"log_odds_chosen": 0.6318256258964539,
|
| 257 |
+
"log_odds_ratio": -0.5325360298156738,
|
| 258 |
+
"logps/chosen": -0.922771155834198,
|
| 259 |
+
"logps/rejected": -1.320297360420227,
|
| 260 |
+
"loss": 38.8799,
|
| 261 |
+
"nll_loss": 1.0614503622055054,
|
| 262 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 263 |
+
"rewards/chosen": -0.2306927889585495,
|
| 264 |
+
"rewards/margins": 0.09938153624534607,
|
| 265 |
+
"rewards/rejected": -0.33007434010505676,
|
| 266 |
+
"step": 80
|
| 267 |
+
},
|
| 268 |
+
{
|
| 269 |
+
"epoch": 1.7565495207667732,
|
| 270 |
+
"grad_norm": 87.10155487060547,
|
| 271 |
+
"learning_rate": 3.435575568570633e-05,
|
| 272 |
+
"log_odds_chosen": 0.7393198013305664,
|
| 273 |
+
"log_odds_ratio": -0.5058027505874634,
|
| 274 |
+
"logps/chosen": -0.8932069540023804,
|
| 275 |
+
"logps/rejected": -1.3613816499710083,
|
| 276 |
+
"loss": 38.9466,
|
| 277 |
+
"nll_loss": 1.0541436672210693,
|
| 278 |
+
"rewards/accuracies": 0.765625,
|
| 279 |
+
"rewards/chosen": -0.2233017385005951,
|
| 280 |
+
"rewards/margins": 0.117043636739254,
|
| 281 |
+
"rewards/rejected": -0.3403454124927521,
|
| 282 |
+
"step": 85
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"epoch": 1.8587859424920128,
|
| 286 |
+
"grad_norm": 116.91316223144531,
|
| 287 |
+
"learning_rate": 3.220247748413094e-05,
|
| 288 |
+
"log_odds_chosen": 0.7619136571884155,
|
| 289 |
+
"log_odds_ratio": -0.5002596974372864,
|
| 290 |
+
"logps/chosen": -0.9479422569274902,
|
| 291 |
+
"logps/rejected": -1.4273831844329834,
|
| 292 |
+
"loss": 38.4031,
|
| 293 |
+
"nll_loss": 1.0556285381317139,
|
| 294 |
+
"rewards/accuracies": 0.778124988079071,
|
| 295 |
+
"rewards/chosen": -0.23698556423187256,
|
| 296 |
+
"rewards/margins": 0.1198602169752121,
|
| 297 |
+
"rewards/rejected": -0.35684579610824585,
|
| 298 |
+
"step": 90
|
| 299 |
+
},
|
| 300 |
+
{
|
| 301 |
+
"epoch": 1.9610223642172524,
|
| 302 |
+
"grad_norm": 103.88111877441406,
|
| 303 |
+
"learning_rate": 2.9989170034648823e-05,
|
| 304 |
+
"log_odds_chosen": 0.6734960079193115,
|
| 305 |
+
"log_odds_ratio": -0.5330161452293396,
|
| 306 |
+
"logps/chosen": -0.8996769189834595,
|
| 307 |
+
"logps/rejected": -1.3381158113479614,
|
| 308 |
+
"loss": 38.0074,
|
| 309 |
+
"nll_loss": 1.03584623336792,
|
| 310 |
+
"rewards/accuracies": 0.7718750238418579,
|
| 311 |
+
"rewards/chosen": -0.22491922974586487,
|
| 312 |
+
"rewards/margins": 0.1096097081899643,
|
| 313 |
+
"rewards/rejected": -0.33452895283699036,
|
| 314 |
+
"step": 95
|
| 315 |
+
},
|
| 316 |
+
{
|
| 317 |
+
"epoch": 2.081789137380192,
|
| 318 |
+
"grad_norm": 93.64875030517578,
|
| 319 |
+
"learning_rate": 2.7734280209446865e-05,
|
| 320 |
+
"log_odds_chosen": 1.0058292150497437,
|
| 321 |
+
"log_odds_ratio": -0.42482826113700867,
|
| 322 |
+
"logps/chosen": -0.7912481427192688,
|
| 323 |
+
"logps/rejected": -1.4134339094161987,
|
| 324 |
+
"loss": 39.7125,
|
| 325 |
+
"nll_loss": 0.9234015345573425,
|
| 326 |
+
"rewards/accuracies": 0.8602150678634644,
|
| 327 |
+
"rewards/chosen": -0.1978120356798172,
|
| 328 |
+
"rewards/margins": 0.15554644167423248,
|
| 329 |
+
"rewards/rejected": -0.3533584773540497,
|
| 330 |
+
"step": 100
|
| 331 |
+
},
|
| 332 |
+
{
|
| 333 |
+
"epoch": 2.1840255591054314,
|
| 334 |
+
"grad_norm": 54.99567794799805,
|
| 335 |
+
"learning_rate": 2.5456601450173125e-05,
|
| 336 |
+
"log_odds_chosen": 1.354565978050232,
|
| 337 |
+
"log_odds_ratio": -0.32823148369789124,
|
| 338 |
+
"logps/chosen": -0.7432836890220642,
|
| 339 |
+
"logps/rejected": -1.5525643825531006,
|
| 340 |
+
"loss": 31.9032,
|
| 341 |
+
"nll_loss": 0.8803617358207703,
|
| 342 |
+
"rewards/accuracies": 0.903124988079071,
|
| 343 |
+
"rewards/chosen": -0.18582092225551605,
|
| 344 |
+
"rewards/margins": 0.2023201882839203,
|
| 345 |
+
"rewards/rejected": -0.38814109563827515,
|
| 346 |
+
"step": 105
|
| 347 |
+
},
|
| 348 |
+
{
|
| 349 |
+
"epoch": 2.286261980830671,
|
| 350 |
+
"grad_norm": 79.57364654541016,
|
| 351 |
+
"learning_rate": 2.3175117133477313e-05,
|
| 352 |
+
"log_odds_chosen": 1.357545018196106,
|
| 353 |
+
"log_odds_ratio": -0.339216947555542,
|
| 354 |
+
"logps/chosen": -0.7088838815689087,
|
| 355 |
+
"logps/rejected": -1.5366592407226562,
|
| 356 |
+
"loss": 31.0042,
|
| 357 |
+
"nll_loss": 0.8524471521377563,
|
| 358 |
+
"rewards/accuracies": 0.909375011920929,
|
| 359 |
+
"rewards/chosen": -0.17722097039222717,
|
| 360 |
+
"rewards/margins": 0.2069438397884369,
|
| 361 |
+
"rewards/rejected": -0.38416481018066406,
|
| 362 |
+
"step": 110
|
| 363 |
+
},
|
| 364 |
+
{
|
| 365 |
+
"epoch": 2.3884984025559106,
|
| 366 |
+
"grad_norm": 69.78687286376953,
|
| 367 |
+
"learning_rate": 2.0908842353532803e-05,
|
| 368 |
+
"log_odds_chosen": 1.4747217893600464,
|
| 369 |
+
"log_odds_ratio": -0.3158169388771057,
|
| 370 |
+
"logps/chosen": -0.6909688711166382,
|
| 371 |
+
"logps/rejected": -1.5954649448394775,
|
| 372 |
+
"loss": 30.9753,
|
| 373 |
+
"nll_loss": 0.8266602754592896,
|
| 374 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 375 |
+
"rewards/chosen": -0.17274221777915955,
|
| 376 |
+
"rewards/margins": 0.22612404823303223,
|
| 377 |
+
"rewards/rejected": -0.3988662362098694,
|
| 378 |
+
"step": 115
|
| 379 |
+
},
|
| 380 |
+
{
|
| 381 |
+
"epoch": 2.4907348242811502,
|
| 382 |
+
"grad_norm": 42.07189178466797,
|
| 383 |
+
"learning_rate": 1.867666544020798e-05,
|
| 384 |
+
"log_odds_chosen": 1.5207545757293701,
|
| 385 |
+
"log_odds_ratio": -0.30957141518592834,
|
| 386 |
+
"logps/chosen": -0.7246582508087158,
|
| 387 |
+
"logps/rejected": -1.694667100906372,
|
| 388 |
+
"loss": 31.3816,
|
| 389 |
+
"nll_loss": 0.8578312993049622,
|
| 390 |
+
"rewards/accuracies": 0.90625,
|
| 391 |
+
"rewards/chosen": -0.18116456270217896,
|
| 392 |
+
"rewards/margins": 0.24250221252441406,
|
| 393 |
+
"rewards/rejected": -0.423666775226593,
|
| 394 |
+
"step": 120
|
| 395 |
+
},
|
| 396 |
+
{
|
| 397 |
+
"epoch": 2.59297124600639,
|
| 398 |
+
"grad_norm": 79.26730346679688,
|
| 399 |
+
"learning_rate": 1.6497190533758348e-05,
|
| 400 |
+
"log_odds_chosen": 1.571445107460022,
|
| 401 |
+
"log_odds_ratio": -0.2881600260734558,
|
| 402 |
+
"logps/chosen": -0.7087723016738892,
|
| 403 |
+
"logps/rejected": -1.6907808780670166,
|
| 404 |
+
"loss": 32.1674,
|
| 405 |
+
"nll_loss": 0.8415018320083618,
|
| 406 |
+
"rewards/accuracies": 0.9281250238418579,
|
| 407 |
+
"rewards/chosen": -0.1771930754184723,
|
| 408 |
+
"rewards/margins": 0.24550211429595947,
|
| 409 |
+
"rewards/rejected": -0.42269521951675415,
|
| 410 |
+
"step": 125
|
| 411 |
+
},
|
| 412 |
+
{
|
| 413 |
+
"epoch": 2.6952076677316295,
|
| 414 |
+
"grad_norm": 48.99087142944336,
|
| 415 |
+
"learning_rate": 1.4388582528104628e-05,
|
| 416 |
+
"log_odds_chosen": 1.4256895780563354,
|
| 417 |
+
"log_odds_ratio": -0.3456164300441742,
|
| 418 |
+
"logps/chosen": -0.6957945823669434,
|
| 419 |
+
"logps/rejected": -1.5854990482330322,
|
| 420 |
+
"loss": 30.7822,
|
| 421 |
+
"nll_loss": 0.8388240933418274,
|
| 422 |
+
"rewards/accuracies": 0.878125011920929,
|
| 423 |
+
"rewards/chosen": -0.17394864559173584,
|
| 424 |
+
"rewards/margins": 0.22242608666419983,
|
| 425 |
+
"rewards/rejected": -0.39637476205825806,
|
| 426 |
+
"step": 130
|
| 427 |
+
},
|
| 428 |
+
{
|
| 429 |
+
"epoch": 2.797444089456869,
|
| 430 |
+
"grad_norm": 45.743141174316406,
|
| 431 |
+
"learning_rate": 1.2368415675021768e-05,
|
| 432 |
+
"log_odds_chosen": 1.6803510189056396,
|
| 433 |
+
"log_odds_ratio": -0.29377657175064087,
|
| 434 |
+
"logps/chosen": -0.6751910448074341,
|
| 435 |
+
"logps/rejected": -1.7030166387557983,
|
| 436 |
+
"loss": 29.4657,
|
| 437 |
+
"nll_loss": 0.8158040046691895,
|
| 438 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 439 |
+
"rewards/chosen": -0.16879776120185852,
|
| 440 |
+
"rewards/margins": 0.2569563686847687,
|
| 441 |
+
"rewards/rejected": -0.4257541596889496,
|
| 442 |
+
"step": 135
|
| 443 |
+
},
|
| 444 |
+
{
|
| 445 |
+
"epoch": 2.8996805111821087,
|
| 446 |
+
"grad_norm": 45.214969635009766,
|
| 447 |
+
"learning_rate": 1.0453527111051184e-05,
|
| 448 |
+
"log_odds_chosen": 1.5122963190078735,
|
| 449 |
+
"log_odds_ratio": -0.2940905690193176,
|
| 450 |
+
"logps/chosen": -0.6969622373580933,
|
| 451 |
+
"logps/rejected": -1.6159965991973877,
|
| 452 |
+
"loss": 31.0349,
|
| 453 |
+
"nll_loss": 0.8314476013183594,
|
| 454 |
+
"rewards/accuracies": 0.934374988079071,
|
| 455 |
+
"rewards/chosen": -0.17424055933952332,
|
| 456 |
+
"rewards/margins": 0.22975853085517883,
|
| 457 |
+
"rewards/rejected": -0.4039991497993469,
|
| 458 |
+
"step": 140
|
| 459 |
+
},
|
| 460 |
+
{
|
| 461 |
+
"epoch": 3.020447284345048,
|
| 462 |
+
"grad_norm": 74.37804412841797,
|
| 463 |
+
"learning_rate": 8.659876527920277e-06,
|
| 464 |
+
"log_odds_chosen": 1.7624249458312988,
|
| 465 |
+
"log_odds_ratio": -0.2721032202243805,
|
| 466 |
+
"logps/chosen": -0.6561669707298279,
|
| 467 |
+
"logps/rejected": -1.740464687347412,
|
| 468 |
+
"loss": 34.4684,
|
| 469 |
+
"nll_loss": 0.7908505797386169,
|
| 470 |
+
"rewards/accuracies": 0.9408602118492126,
|
| 471 |
+
"rewards/chosen": -0.16404174268245697,
|
| 472 |
+
"rewards/margins": 0.27107441425323486,
|
| 473 |
+
"rewards/rejected": -0.435116171836853,
|
| 474 |
+
"step": 145
|
| 475 |
+
},
|
| 476 |
+
{
|
| 477 |
+
"epoch": 3.1226837060702874,
|
| 478 |
+
"grad_norm": 45.229400634765625,
|
| 479 |
+
"learning_rate": 7.002413156050108e-06,
|
| 480 |
+
"log_odds_chosen": 2.416283130645752,
|
| 481 |
+
"log_odds_ratio": -0.1788024604320526,
|
| 482 |
+
"logps/chosen": -0.5219112634658813,
|
| 483 |
+
"logps/rejected": -1.9957574605941772,
|
| 484 |
+
"loss": 27.9227,
|
| 485 |
+
"nll_loss": 0.6778407692909241,
|
| 486 |
+
"rewards/accuracies": 0.9781249761581421,
|
| 487 |
+
"rewards/chosen": -0.13047781586647034,
|
| 488 |
+
"rewards/margins": 0.3684615194797516,
|
| 489 |
+
"rewards/rejected": -0.4989393651485443,
|
| 490 |
+
"step": 150
|
| 491 |
+
},
|
| 492 |
+
{
|
| 493 |
+
"epoch": 3.224920127795527,
|
| 494 |
+
"grad_norm": 57.19989776611328,
|
| 495 |
+
"learning_rate": 5.4949511697807765e-06,
|
| 496 |
+
"log_odds_chosen": 2.6506989002227783,
|
| 497 |
+
"log_odds_ratio": -0.15020069479942322,
|
| 498 |
+
"logps/chosen": -0.5697158575057983,
|
| 499 |
+
"logps/rejected": -2.3274993896484375,
|
| 500 |
+
"loss": 27.0184,
|
| 501 |
+
"nll_loss": 0.6888783574104309,
|
| 502 |
+
"rewards/accuracies": 0.96875,
|
| 503 |
+
"rewards/chosen": -0.14242896437644958,
|
| 504 |
+
"rewards/margins": 0.439445823431015,
|
| 505 |
+
"rewards/rejected": -0.5818748474121094,
|
| 506 |
+
"step": 155
|
| 507 |
+
},
|
| 508 |
+
{
|
| 509 |
+
"epoch": 3.3271565495207667,
|
| 510 |
+
"grad_norm": 59.800697326660156,
|
| 511 |
+
"learning_rate": 4.150054552753055e-06,
|
| 512 |
+
"log_odds_chosen": 2.5286622047424316,
|
| 513 |
+
"log_odds_ratio": -0.19968460500240326,
|
| 514 |
+
"logps/chosen": -0.5568191409111023,
|
| 515 |
+
"logps/rejected": -2.1876730918884277,
|
| 516 |
+
"loss": 27.8054,
|
| 517 |
+
"nll_loss": 0.6883557438850403,
|
| 518 |
+
"rewards/accuracies": 0.9468749761581421,
|
| 519 |
+
"rewards/chosen": -0.13920478522777557,
|
| 520 |
+
"rewards/margins": 0.40771350264549255,
|
| 521 |
+
"rewards/rejected": -0.5469182729721069,
|
| 522 |
+
"step": 160
|
| 523 |
+
},
|
| 524 |
+
{
|
| 525 |
+
"epoch": 3.4293929712460063,
|
| 526 |
+
"grad_norm": 54.28583908081055,
|
| 527 |
+
"learning_rate": 2.978932383039093e-06,
|
| 528 |
+
"log_odds_chosen": 2.5673203468322754,
|
| 529 |
+
"log_odds_ratio": -0.18180711567401886,
|
| 530 |
+
"logps/chosen": -0.5963090062141418,
|
| 531 |
+
"logps/rejected": -2.295097589492798,
|
| 532 |
+
"loss": 26.8584,
|
| 533 |
+
"nll_loss": 0.737380862236023,
|
| 534 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 535 |
+
"rewards/chosen": -0.14907725155353546,
|
| 536 |
+
"rewards/margins": 0.4246971607208252,
|
| 537 |
+
"rewards/rejected": -0.5737743973731995,
|
| 538 |
+
"step": 165
|
| 539 |
+
},
|
| 540 |
+
{
|
| 541 |
+
"epoch": 3.531629392971246,
|
| 542 |
+
"grad_norm": 55.28144836425781,
|
| 543 |
+
"learning_rate": 1.9913454107710173e-06,
|
| 544 |
+
"log_odds_chosen": 2.435741901397705,
|
| 545 |
+
"log_odds_ratio": -0.20390479266643524,
|
| 546 |
+
"logps/chosen": -0.5668953061103821,
|
| 547 |
+
"logps/rejected": -2.1278016567230225,
|
| 548 |
+
"loss": 27.9555,
|
| 549 |
+
"nll_loss": 0.6951500177383423,
|
| 550 |
+
"rewards/accuracies": 0.9375,
|
| 551 |
+
"rewards/chosen": -0.14172382652759552,
|
| 552 |
+
"rewards/margins": 0.3902266025543213,
|
| 553 |
+
"rewards/rejected": -0.5319504141807556,
|
| 554 |
+
"step": 170
|
| 555 |
+
},
|
| 556 |
+
{
|
| 557 |
+
"epoch": 3.6338658146964855,
|
| 558 |
+
"grad_norm": 46.94998550415039,
|
| 559 |
+
"learning_rate": 1.1955247068988261e-06,
|
| 560 |
+
"log_odds_chosen": 2.725480556488037,
|
| 561 |
+
"log_odds_ratio": -0.15135148167610168,
|
| 562 |
+
"logps/chosen": -0.5273882150650024,
|
| 563 |
+
"logps/rejected": -2.2832539081573486,
|
| 564 |
+
"loss": 28.734,
|
| 565 |
+
"nll_loss": 0.6737874746322632,
|
| 566 |
+
"rewards/accuracies": 0.965624988079071,
|
| 567 |
+
"rewards/chosen": -0.1318470537662506,
|
| 568 |
+
"rewards/margins": 0.43896645307540894,
|
| 569 |
+
"rewards/rejected": -0.5708134770393372,
|
| 570 |
+
"step": 175
|
| 571 |
+
},
|
| 572 |
+
{
|
| 573 |
+
"epoch": 3.736102236421725,
|
| 574 |
+
"grad_norm": 66.62850189208984,
|
| 575 |
+
"learning_rate": 5.981030611018234e-07,
|
| 576 |
+
"log_odds_chosen": 2.5903310775756836,
|
| 577 |
+
"log_odds_ratio": -0.15854457020759583,
|
| 578 |
+
"logps/chosen": -0.5564672946929932,
|
| 579 |
+
"logps/rejected": -2.2155492305755615,
|
| 580 |
+
"loss": 25.0194,
|
| 581 |
+
"nll_loss": 0.684872031211853,
|
| 582 |
+
"rewards/accuracies": 0.9781249761581421,
|
| 583 |
+
"rewards/chosen": -0.1391168236732483,
|
| 584 |
+
"rewards/margins": 0.4147705137729645,
|
| 585 |
+
"rewards/rejected": -0.5538873076438904,
|
| 586 |
+
"step": 180
|
| 587 |
+
},
|
| 588 |
+
{
|
| 589 |
+
"epoch": 3.8383386581469647,
|
| 590 |
+
"grad_norm": 50.985652923583984,
|
| 591 |
+
"learning_rate": 2.0405970061943002e-07,
|
| 592 |
+
"log_odds_chosen": 2.7051525115966797,
|
| 593 |
+
"log_odds_ratio": -0.16894596815109253,
|
| 594 |
+
"logps/chosen": -0.543402910232544,
|
| 595 |
+
"logps/rejected": -2.2726523876190186,
|
| 596 |
+
"loss": 30.0762,
|
| 597 |
+
"nll_loss": 0.6836854219436646,
|
| 598 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 599 |
+
"rewards/chosen": -0.135850727558136,
|
| 600 |
+
"rewards/margins": 0.43231239914894104,
|
| 601 |
+
"rewards/rejected": -0.5681630969047546,
|
| 602 |
+
"step": 185
|
| 603 |
+
},
|
| 604 |
+
{
|
| 605 |
+
"epoch": 3.9405750798722043,
|
| 606 |
+
"grad_norm": 50.59651565551758,
|
| 607 |
+
"learning_rate": 1.6678790744015238e-08,
|
| 608 |
+
"log_odds_chosen": 2.563751697540283,
|
| 609 |
+
"log_odds_ratio": -0.16612394154071808,
|
| 610 |
+
"logps/chosen": -0.5317026972770691,
|
| 611 |
+
"logps/rejected": -2.1680614948272705,
|
| 612 |
+
"loss": 25.6433,
|
| 613 |
+
"nll_loss": 0.6562982797622681,
|
| 614 |
+
"rewards/accuracies": 0.965624988079071,
|
| 615 |
+
"rewards/chosen": -0.13292567431926727,
|
| 616 |
+
"rewards/margins": 0.4090896546840668,
|
| 617 |
+
"rewards/rejected": -0.5420153737068176,
|
| 618 |
+
"step": 190
|
| 619 |
+
},
|
| 620 |
+
{
|
| 621 |
+
"epoch": 3.9814696485623005,
|
| 622 |
+
"step": 192,
|
| 623 |
+
"total_flos": 0.0,
|
| 624 |
+
"train_loss": 62.26987236738205,
|
| 625 |
+
"train_runtime": 5592.1515,
|
| 626 |
+
"train_samples_per_second": 2.239,
|
| 627 |
+
"train_steps_per_second": 0.034
|
| 628 |
+
}
|
| 629 |
+
],
|
| 630 |
+
"logging_steps": 5,
|
| 631 |
+
"max_steps": 192,
|
| 632 |
+
"num_input_tokens_seen": 0,
|
| 633 |
+
"num_train_epochs": 4,
|
| 634 |
+
"save_steps": 100,
|
| 635 |
+
"stateful_callbacks": {
|
| 636 |
+
"TrainerControl": {
|
| 637 |
+
"args": {
|
| 638 |
+
"should_epoch_stop": false,
|
| 639 |
+
"should_evaluate": false,
|
| 640 |
+
"should_log": false,
|
| 641 |
+
"should_save": true,
|
| 642 |
+
"should_training_stop": true
|
| 643 |
+
},
|
| 644 |
+
"attributes": {}
|
| 645 |
+
}
|
| 646 |
+
},
|
| 647 |
+
"total_flos": 0.0,
|
| 648 |
+
"train_batch_size": 1,
|
| 649 |
+
"trial_name": null,
|
| 650 |
+
"trial_params": null
|
| 651 |
+
}
|