Model save
Browse files- README.md +67 -0
- all_results.json +9 -0
- runs/Jan28_02-13-00_b4d476355ce0/events.out.tfevents.1738030393.b4d476355ce0.128156.0 +2 -2
- train_results.json +9 -0
- trainer_state.json +1050 -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-basic-5e-5-v4
|
| 5 |
+
tags:
|
| 6 |
+
- generated_from_trainer
|
| 7 |
+
- trl
|
| 8 |
+
- orpo
|
| 9 |
+
licence: license
|
| 10 |
+
---
|
| 11 |
+
|
| 12 |
+
# Model Card for gemma-7b-silvia-basic-5e-5-v4
|
| 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-basic-5e-5-v4", 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/jxlh26x2)
|
| 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: 0.13.0
|
| 38 |
+
- Transformers: 4.48.1
|
| 39 |
+
- Pytorch: 2.5.1
|
| 40 |
+
- Datasets: 3.2.0
|
| 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": 2.9765925925925925,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": -61.895723918127636,
|
| 5 |
+
"train_runtime": 9291.6158,
|
| 6 |
+
"train_samples": 6750,
|
| 7 |
+
"train_samples_per_second": 2.179,
|
| 8 |
+
"train_steps_per_second": 0.034
|
| 9 |
+
}
|
runs/Jan28_02-13-00_b4d476355ce0/events.out.tfevents.1738030393.b4d476355ce0.128156.0
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:20318eb5dd60dbcaf0a1eea062e47c19c33f57f691dccc4936353efb3c983ff7
|
| 3 |
+
size 51930
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 2.9765925925925925,
|
| 3 |
+
"total_flos": 0.0,
|
| 4 |
+
"train_loss": -61.895723918127636,
|
| 5 |
+
"train_runtime": 9291.6158,
|
| 6 |
+
"train_samples": 6750,
|
| 7 |
+
"train_samples_per_second": 2.179,
|
| 8 |
+
"train_steps_per_second": 0.034
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,1050 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 2.9765925925925925,
|
| 5 |
+
"eval_steps": 500,
|
| 6 |
+
"global_step": 315,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.047407407407407405,
|
| 13 |
+
"grad_norm": 1309.7264404296875,
|
| 14 |
+
"learning_rate": 7.8125e-06,
|
| 15 |
+
"log_odds_chosen": -1.0092629194259644,
|
| 16 |
+
"log_odds_ratio": -10.180293083190918,
|
| 17 |
+
"logps/chosen": -21.700639724731445,
|
| 18 |
+
"logps/rejected": -20.691608428955078,
|
| 19 |
+
"loss": 710.1368,
|
| 20 |
+
"nll_loss": 9.67548656463623,
|
| 21 |
+
"rewards/accuracies": 0.484375,
|
| 22 |
+
"rewards/chosen": -10.850319862365723,
|
| 23 |
+
"rewards/margins": -0.5045148134231567,
|
| 24 |
+
"rewards/rejected": -10.345804214477539,
|
| 25 |
+
"step": 5
|
| 26 |
+
},
|
| 27 |
+
{
|
| 28 |
+
"epoch": 0.09481481481481481,
|
| 29 |
+
"grad_norm": 1233.22119140625,
|
| 30 |
+
"learning_rate": 1.5625e-05,
|
| 31 |
+
"log_odds_chosen": -2.799683094024658,
|
| 32 |
+
"log_odds_ratio": -10.153043746948242,
|
| 33 |
+
"logps/chosen": -20.964075088500977,
|
| 34 |
+
"logps/rejected": -18.164752960205078,
|
| 35 |
+
"loss": 714.4481,
|
| 36 |
+
"nll_loss": 8.683107376098633,
|
| 37 |
+
"rewards/accuracies": 0.44062501192092896,
|
| 38 |
+
"rewards/chosen": -10.482037544250488,
|
| 39 |
+
"rewards/margins": -1.3996615409851074,
|
| 40 |
+
"rewards/rejected": -9.082376480102539,
|
| 41 |
+
"step": 10
|
| 42 |
+
},
|
| 43 |
+
{
|
| 44 |
+
"epoch": 0.14222222222222222,
|
| 45 |
+
"grad_norm": 664.7484741210938,
|
| 46 |
+
"learning_rate": 2.34375e-05,
|
| 47 |
+
"log_odds_chosen": -1.274759292602539,
|
| 48 |
+
"log_odds_ratio": -10.432046890258789,
|
| 49 |
+
"logps/chosen": -20.83938980102539,
|
| 50 |
+
"logps/rejected": -19.563129425048828,
|
| 51 |
+
"loss": 686.9441,
|
| 52 |
+
"nll_loss": 8.075501441955566,
|
| 53 |
+
"rewards/accuracies": 0.515625,
|
| 54 |
+
"rewards/chosen": -10.419694900512695,
|
| 55 |
+
"rewards/margins": -0.638128936290741,
|
| 56 |
+
"rewards/rejected": -9.781564712524414,
|
| 57 |
+
"step": 15
|
| 58 |
+
},
|
| 59 |
+
{
|
| 60 |
+
"epoch": 0.18962962962962962,
|
| 61 |
+
"grad_norm": 5581.04345703125,
|
| 62 |
+
"learning_rate": 3.125e-05,
|
| 63 |
+
"log_odds_chosen": -3.752788543701172,
|
| 64 |
+
"log_odds_ratio": -8.436421394348145,
|
| 65 |
+
"logps/chosen": -15.843017578125,
|
| 66 |
+
"logps/rejected": -12.092793464660645,
|
| 67 |
+
"loss": 565.9146,
|
| 68 |
+
"nll_loss": 6.499529838562012,
|
| 69 |
+
"rewards/accuracies": 0.4375,
|
| 70 |
+
"rewards/chosen": -7.9215087890625,
|
| 71 |
+
"rewards/margins": -1.8751119375228882,
|
| 72 |
+
"rewards/rejected": -6.046396732330322,
|
| 73 |
+
"step": 20
|
| 74 |
+
},
|
| 75 |
+
{
|
| 76 |
+
"epoch": 0.23703703703703705,
|
| 77 |
+
"grad_norm": 271.88677978515625,
|
| 78 |
+
"learning_rate": 3.90625e-05,
|
| 79 |
+
"log_odds_chosen": -0.09555625915527344,
|
| 80 |
+
"log_odds_ratio": -1.5934648513793945,
|
| 81 |
+
"logps/chosen": -3.681891918182373,
|
| 82 |
+
"logps/rejected": -3.571626663208008,
|
| 83 |
+
"loss": 120.1892,
|
| 84 |
+
"nll_loss": 3.53655743598938,
|
| 85 |
+
"rewards/accuracies": 0.5562499761581421,
|
| 86 |
+
"rewards/chosen": -1.8409459590911865,
|
| 87 |
+
"rewards/margins": -0.055132925510406494,
|
| 88 |
+
"rewards/rejected": -1.785813331604004,
|
| 89 |
+
"step": 25
|
| 90 |
+
},
|
| 91 |
+
{
|
| 92 |
+
"epoch": 0.28444444444444444,
|
| 93 |
+
"grad_norm": 225.77731323242188,
|
| 94 |
+
"learning_rate": 4.6875e-05,
|
| 95 |
+
"log_odds_chosen": 0.05429550260305405,
|
| 96 |
+
"log_odds_ratio": -0.8730447888374329,
|
| 97 |
+
"logps/chosen": -1.855459213256836,
|
| 98 |
+
"logps/rejected": -1.9073245525360107,
|
| 99 |
+
"loss": 58.9805,
|
| 100 |
+
"nll_loss": 2.5173518657684326,
|
| 101 |
+
"rewards/accuracies": 0.5249999761581421,
|
| 102 |
+
"rewards/chosen": -0.927729606628418,
|
| 103 |
+
"rewards/margins": 0.025932852178812027,
|
| 104 |
+
"rewards/rejected": -0.9536622762680054,
|
| 105 |
+
"step": 30
|
| 106 |
+
},
|
| 107 |
+
{
|
| 108 |
+
"epoch": 0.33185185185185184,
|
| 109 |
+
"grad_norm": 330.9626770019531,
|
| 110 |
+
"learning_rate": 4.998613757348784e-05,
|
| 111 |
+
"log_odds_chosen": 0.21338006854057312,
|
| 112 |
+
"log_odds_ratio": -0.8034284710884094,
|
| 113 |
+
"logps/chosen": -1.7032153606414795,
|
| 114 |
+
"logps/rejected": -1.8913615942001343,
|
| 115 |
+
"loss": 52.0844,
|
| 116 |
+
"nll_loss": 2.3274359703063965,
|
| 117 |
+
"rewards/accuracies": 0.518750011920929,
|
| 118 |
+
"rewards/chosen": -0.8516076803207397,
|
| 119 |
+
"rewards/margins": 0.09407330304384232,
|
| 120 |
+
"rewards/rejected": -0.9456807971000671,
|
| 121 |
+
"step": 35
|
| 122 |
+
},
|
| 123 |
+
{
|
| 124 |
+
"epoch": 0.37925925925925924,
|
| 125 |
+
"grad_norm": 175.3074493408203,
|
| 126 |
+
"learning_rate": 4.990147841143462e-05,
|
| 127 |
+
"log_odds_chosen": 0.31794872879981995,
|
| 128 |
+
"log_odds_ratio": -0.7034906148910522,
|
| 129 |
+
"logps/chosen": -1.5063345432281494,
|
| 130 |
+
"logps/rejected": -1.7724323272705078,
|
| 131 |
+
"loss": 45.107,
|
| 132 |
+
"nll_loss": 2.3017280101776123,
|
| 133 |
+
"rewards/accuracies": 0.590624988079071,
|
| 134 |
+
"rewards/chosen": -0.7531672716140747,
|
| 135 |
+
"rewards/margins": 0.1330489218235016,
|
| 136 |
+
"rewards/rejected": -0.8862161636352539,
|
| 137 |
+
"step": 40
|
| 138 |
+
},
|
| 139 |
+
{
|
| 140 |
+
"epoch": 0.4266666666666667,
|
| 141 |
+
"grad_norm": 148.9108123779297,
|
| 142 |
+
"learning_rate": 4.97401218720448e-05,
|
| 143 |
+
"log_odds_chosen": 0.32661986351013184,
|
| 144 |
+
"log_odds_ratio": -0.695238471031189,
|
| 145 |
+
"logps/chosen": -1.4421089887619019,
|
| 146 |
+
"logps/rejected": -1.732181191444397,
|
| 147 |
+
"loss": 42.7725,
|
| 148 |
+
"nll_loss": 2.121166944503784,
|
| 149 |
+
"rewards/accuracies": 0.6031249761581421,
|
| 150 |
+
"rewards/chosen": -0.7210544943809509,
|
| 151 |
+
"rewards/margins": 0.14503604173660278,
|
| 152 |
+
"rewards/rejected": -0.8660905957221985,
|
| 153 |
+
"step": 45
|
| 154 |
+
},
|
| 155 |
+
{
|
| 156 |
+
"epoch": 0.4740740740740741,
|
| 157 |
+
"grad_norm": 168.20802307128906,
|
| 158 |
+
"learning_rate": 4.9502564938797946e-05,
|
| 159 |
+
"log_odds_chosen": 0.41916948556900024,
|
| 160 |
+
"log_odds_ratio": -0.6709465980529785,
|
| 161 |
+
"logps/chosen": -1.3388197422027588,
|
| 162 |
+
"logps/rejected": -1.7024009227752686,
|
| 163 |
+
"loss": 38.517,
|
| 164 |
+
"nll_loss": 2.0670645236968994,
|
| 165 |
+
"rewards/accuracies": 0.625,
|
| 166 |
+
"rewards/chosen": -0.6694098711013794,
|
| 167 |
+
"rewards/margins": 0.18179059028625488,
|
| 168 |
+
"rewards/rejected": -0.8512004613876343,
|
| 169 |
+
"step": 50
|
| 170 |
+
},
|
| 171 |
+
{
|
| 172 |
+
"epoch": 0.5214814814814814,
|
| 173 |
+
"grad_norm": 1289.98828125,
|
| 174 |
+
"learning_rate": 4.918953929490768e-05,
|
| 175 |
+
"log_odds_chosen": 1.8812118768692017,
|
| 176 |
+
"log_odds_ratio": -0.7291940450668335,
|
| 177 |
+
"logps/chosen": -1.6740214824676514,
|
| 178 |
+
"logps/rejected": -3.4779555797576904,
|
| 179 |
+
"loss": 26.985,
|
| 180 |
+
"nll_loss": 2.3563015460968018,
|
| 181 |
+
"rewards/accuracies": 0.625,
|
| 182 |
+
"rewards/chosen": -0.8370107412338257,
|
| 183 |
+
"rewards/margins": 0.9019671678543091,
|
| 184 |
+
"rewards/rejected": -1.7389777898788452,
|
| 185 |
+
"step": 55
|
| 186 |
+
},
|
| 187 |
+
{
|
| 188 |
+
"epoch": 0.5688888888888889,
|
| 189 |
+
"grad_norm": 846.8933715820312,
|
| 190 |
+
"learning_rate": 4.88020090697132e-05,
|
| 191 |
+
"log_odds_chosen": 5.034865379333496,
|
| 192 |
+
"log_odds_ratio": -0.7896591424942017,
|
| 193 |
+
"logps/chosen": -1.8219846487045288,
|
| 194 |
+
"logps/rejected": -6.7804718017578125,
|
| 195 |
+
"loss": -18.5626,
|
| 196 |
+
"nll_loss": 2.4730238914489746,
|
| 197 |
+
"rewards/accuracies": 0.653124988079071,
|
| 198 |
+
"rewards/chosen": -0.9109923243522644,
|
| 199 |
+
"rewards/margins": 2.479243755340576,
|
| 200 |
+
"rewards/rejected": -3.3902359008789062,
|
| 201 |
+
"step": 60
|
| 202 |
+
},
|
| 203 |
+
{
|
| 204 |
+
"epoch": 0.6162962962962963,
|
| 205 |
+
"grad_norm": 1050.5416259765625,
|
| 206 |
+
"learning_rate": 4.834116786912897e-05,
|
| 207 |
+
"log_odds_chosen": 7.387457370758057,
|
| 208 |
+
"log_odds_ratio": -2.9080045223236084,
|
| 209 |
+
"logps/chosen": -5.713994026184082,
|
| 210 |
+
"logps/rejected": -13.041893005371094,
|
| 211 |
+
"loss": 67.5015,
|
| 212 |
+
"nll_loss": 5.906378746032715,
|
| 213 |
+
"rewards/accuracies": 0.6312500238418579,
|
| 214 |
+
"rewards/chosen": -2.856997013092041,
|
| 215 |
+
"rewards/margins": 3.6639492511749268,
|
| 216 |
+
"rewards/rejected": -6.520946502685547,
|
| 217 |
+
"step": 65
|
| 218 |
+
},
|
| 219 |
+
{
|
| 220 |
+
"epoch": 0.6637037037037037,
|
| 221 |
+
"grad_norm": 14261.298828125,
|
| 222 |
+
"learning_rate": 4.7808435099299045e-05,
|
| 223 |
+
"log_odds_chosen": 14.964570999145508,
|
| 224 |
+
"log_odds_ratio": -1.2665787935256958,
|
| 225 |
+
"logps/chosen": -2.980807304382324,
|
| 226 |
+
"logps/rejected": -17.873931884765625,
|
| 227 |
+
"loss": -140.3602,
|
| 228 |
+
"nll_loss": 3.395730495452881,
|
| 229 |
+
"rewards/accuracies": 0.6656249761581421,
|
| 230 |
+
"rewards/chosen": -1.490403652191162,
|
| 231 |
+
"rewards/margins": 7.446562767028809,
|
| 232 |
+
"rewards/rejected": -8.936965942382812,
|
| 233 |
+
"step": 70
|
| 234 |
+
},
|
| 235 |
+
{
|
| 236 |
+
"epoch": 0.7111111111111111,
|
| 237 |
+
"grad_norm": 3282.506591796875,
|
| 238 |
+
"learning_rate": 4.720545159477922e-05,
|
| 239 |
+
"log_odds_chosen": 11.172109603881836,
|
| 240 |
+
"log_odds_ratio": -1.5487353801727295,
|
| 241 |
+
"logps/chosen": -2.9426088333129883,
|
| 242 |
+
"logps/rejected": -14.05163860321045,
|
| 243 |
+
"loss": -81.785,
|
| 244 |
+
"nll_loss": 3.614391326904297,
|
| 245 |
+
"rewards/accuracies": 0.643750011920929,
|
| 246 |
+
"rewards/chosen": -1.4713044166564941,
|
| 247 |
+
"rewards/margins": 5.554513931274414,
|
| 248 |
+
"rewards/rejected": -7.025819301605225,
|
| 249 |
+
"step": 75
|
| 250 |
+
},
|
| 251 |
+
{
|
| 252 |
+
"epoch": 0.7585185185185185,
|
| 253 |
+
"grad_norm": 22710.70703125,
|
| 254 |
+
"learning_rate": 4.653407456471222e-05,
|
| 255 |
+
"log_odds_chosen": 14.214700698852539,
|
| 256 |
+
"log_odds_ratio": -3.5922465324401855,
|
| 257 |
+
"logps/chosen": -5.844705104827881,
|
| 258 |
+
"logps/rejected": -20.002582550048828,
|
| 259 |
+
"loss": -37.6675,
|
| 260 |
+
"nll_loss": 6.309117794036865,
|
| 261 |
+
"rewards/accuracies": 0.59375,
|
| 262 |
+
"rewards/chosen": -2.9223525524139404,
|
| 263 |
+
"rewards/margins": 7.078938961029053,
|
| 264 |
+
"rewards/rejected": -10.001291275024414,
|
| 265 |
+
"step": 80
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"epoch": 0.8059259259259259,
|
| 269 |
+
"grad_norm": 4206.48681640625,
|
| 270 |
+
"learning_rate": 4.579637187256222e-05,
|
| 271 |
+
"log_odds_chosen": 11.754661560058594,
|
| 272 |
+
"log_odds_ratio": -2.568394184112549,
|
| 273 |
+
"logps/chosen": -5.506086826324463,
|
| 274 |
+
"logps/rejected": -17.211210250854492,
|
| 275 |
+
"loss": -9.382,
|
| 276 |
+
"nll_loss": 5.795596122741699,
|
| 277 |
+
"rewards/accuracies": 0.6156250238418579,
|
| 278 |
+
"rewards/chosen": -2.7530434131622314,
|
| 279 |
+
"rewards/margins": 5.852560997009277,
|
| 280 |
+
"rewards/rejected": -8.605605125427246,
|
| 281 |
+
"step": 85
|
| 282 |
+
},
|
| 283 |
+
{
|
| 284 |
+
"epoch": 0.8533333333333334,
|
| 285 |
+
"grad_norm": 2450.24169921875,
|
| 286 |
+
"learning_rate": 4.499461566702685e-05,
|
| 287 |
+
"log_odds_chosen": 10.724493980407715,
|
| 288 |
+
"log_odds_ratio": -0.7071830034255981,
|
| 289 |
+
"logps/chosen": -1.4572128057479858,
|
| 290 |
+
"logps/rejected": -12.113273620605469,
|
| 291 |
+
"loss": -122.1039,
|
| 292 |
+
"nll_loss": 2.4590225219726562,
|
| 293 |
+
"rewards/accuracies": 0.6156250238418579,
|
| 294 |
+
"rewards/chosen": -0.7286064028739929,
|
| 295 |
+
"rewards/margins": 5.328030109405518,
|
| 296 |
+
"rewards/rejected": -6.056636810302734,
|
| 297 |
+
"step": 90
|
| 298 |
+
},
|
| 299 |
+
{
|
| 300 |
+
"epoch": 0.9007407407407407,
|
| 301 |
+
"grad_norm": 2608.7001953125,
|
| 302 |
+
"learning_rate": 4.413127538374411e-05,
|
| 303 |
+
"log_odds_chosen": 16.54291534423828,
|
| 304 |
+
"log_odds_ratio": -3.4148666858673096,
|
| 305 |
+
"logps/chosen": -4.583926200866699,
|
| 306 |
+
"logps/rejected": -21.081348419189453,
|
| 307 |
+
"loss": -115.7481,
|
| 308 |
+
"nll_loss": 4.339598655700684,
|
| 309 |
+
"rewards/accuracies": 0.5874999761581421,
|
| 310 |
+
"rewards/chosen": -2.2919631004333496,
|
| 311 |
+
"rewards/margins": 8.248711585998535,
|
| 312 |
+
"rewards/rejected": -10.540674209594727,
|
| 313 |
+
"step": 95
|
| 314 |
+
},
|
| 315 |
+
{
|
| 316 |
+
"epoch": 0.9481481481481482,
|
| 317 |
+
"grad_norm": 1113.737548828125,
|
| 318 |
+
"learning_rate": 4.320901013934887e-05,
|
| 319 |
+
"log_odds_chosen": 20.849918365478516,
|
| 320 |
+
"log_odds_ratio": -2.3039841651916504,
|
| 321 |
+
"logps/chosen": -3.297405242919922,
|
| 322 |
+
"logps/rejected": -24.073766708374023,
|
| 323 |
+
"loss": -224.7481,
|
| 324 |
+
"nll_loss": 4.907585144042969,
|
| 325 |
+
"rewards/accuracies": 0.6187499761581421,
|
| 326 |
+
"rewards/chosen": -1.648702621459961,
|
| 327 |
+
"rewards/margins": 10.38818073272705,
|
| 328 |
+
"rewards/rejected": -12.036883354187012,
|
| 329 |
+
"step": 100
|
| 330 |
+
},
|
| 331 |
+
{
|
| 332 |
+
"epoch": 0.9955555555555555,
|
| 333 |
+
"grad_norm": 564.52294921875,
|
| 334 |
+
"learning_rate": 4.223066054130568e-05,
|
| 335 |
+
"log_odds_chosen": 21.292091369628906,
|
| 336 |
+
"log_odds_ratio": -1.3454216718673706,
|
| 337 |
+
"logps/chosen": -3.835742473602295,
|
| 338 |
+
"logps/rejected": -25.064594268798828,
|
| 339 |
+
"loss": -214.725,
|
| 340 |
+
"nll_loss": 4.128335475921631,
|
| 341 |
+
"rewards/accuracies": 0.609375,
|
| 342 |
+
"rewards/chosen": -1.9178712368011475,
|
| 343 |
+
"rewards/margins": 10.614426612854004,
|
| 344 |
+
"rewards/rejected": -12.532297134399414,
|
| 345 |
+
"step": 105
|
| 346 |
+
},
|
| 347 |
+
{
|
| 348 |
+
"epoch": 1.037925925925926,
|
| 349 |
+
"grad_norm": 2678.355224609375,
|
| 350 |
+
"learning_rate": 4.1199239938743797e-05,
|
| 351 |
+
"log_odds_chosen": 17.971054077148438,
|
| 352 |
+
"log_odds_ratio": -0.902182936668396,
|
| 353 |
+
"logps/chosen": -1.6016716957092285,
|
| 354 |
+
"logps/rejected": -19.46338653564453,
|
| 355 |
+
"loss": -207.4208,
|
| 356 |
+
"nll_loss": 2.815678358078003,
|
| 357 |
+
"rewards/accuracies": 0.6538461446762085,
|
| 358 |
+
"rewards/chosen": -0.8008358478546143,
|
| 359 |
+
"rewards/margins": 8.930856704711914,
|
| 360 |
+
"rewards/rejected": -9.731693267822266,
|
| 361 |
+
"step": 110
|
| 362 |
+
},
|
| 363 |
+
{
|
| 364 |
+
"epoch": 1.0853333333333333,
|
| 365 |
+
"grad_norm": 16211.197265625,
|
| 366 |
+
"learning_rate": 4.0117925141242174e-05,
|
| 367 |
+
"log_odds_chosen": 32.54540252685547,
|
| 368 |
+
"log_odds_ratio": -7.060112953186035,
|
| 369 |
+
"logps/chosen": -16.32876205444336,
|
| 370 |
+
"logps/rejected": -48.71499252319336,
|
| 371 |
+
"loss": 7.6217,
|
| 372 |
+
"nll_loss": 15.108541488647461,
|
| 373 |
+
"rewards/accuracies": 0.6968749761581421,
|
| 374 |
+
"rewards/chosen": -8.16438102722168,
|
| 375 |
+
"rewards/margins": 16.193115234375,
|
| 376 |
+
"rewards/rejected": -24.35749626159668,
|
| 377 |
+
"step": 115
|
| 378 |
+
},
|
| 379 |
+
{
|
| 380 |
+
"epoch": 1.1327407407407408,
|
| 381 |
+
"grad_norm": 24504.349609375,
|
| 382 |
+
"learning_rate": 3.899004663415084e-05,
|
| 383 |
+
"log_odds_chosen": 28.460607528686523,
|
| 384 |
+
"log_odds_ratio": -6.060556888580322,
|
| 385 |
+
"logps/chosen": -10.85397720336914,
|
| 386 |
+
"logps/rejected": -39.16749954223633,
|
| 387 |
+
"loss": -102.1427,
|
| 388 |
+
"nll_loss": 10.644952774047852,
|
| 389 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 390 |
+
"rewards/chosen": -5.42698860168457,
|
| 391 |
+
"rewards/margins": 14.156761169433594,
|
| 392 |
+
"rewards/rejected": -19.583749771118164,
|
| 393 |
+
"step": 120
|
| 394 |
+
},
|
| 395 |
+
{
|
| 396 |
+
"epoch": 1.1801481481481482,
|
| 397 |
+
"grad_norm": 40890.17578125,
|
| 398 |
+
"learning_rate": 3.781907832058587e-05,
|
| 399 |
+
"log_odds_chosen": 20.61638832092285,
|
| 400 |
+
"log_odds_ratio": -6.12774133682251,
|
| 401 |
+
"logps/chosen": -11.241783142089844,
|
| 402 |
+
"logps/rejected": -31.751750946044922,
|
| 403 |
+
"loss": 34.3804,
|
| 404 |
+
"nll_loss": 11.169973373413086,
|
| 405 |
+
"rewards/accuracies": 0.675000011920929,
|
| 406 |
+
"rewards/chosen": -5.620891571044922,
|
| 407 |
+
"rewards/margins": 10.254984855651855,
|
| 408 |
+
"rewards/rejected": -15.875875473022461,
|
| 409 |
+
"step": 125
|
| 410 |
+
},
|
| 411 |
+
{
|
| 412 |
+
"epoch": 1.2275555555555555,
|
| 413 |
+
"grad_norm": 4595.94677734375,
|
| 414 |
+
"learning_rate": 3.660862682169282e-05,
|
| 415 |
+
"log_odds_chosen": 15.926393508911133,
|
| 416 |
+
"log_odds_ratio": -2.1731228828430176,
|
| 417 |
+
"logps/chosen": -4.010047912597656,
|
| 418 |
+
"logps/rejected": -19.802104949951172,
|
| 419 |
+
"loss": -121.6234,
|
| 420 |
+
"nll_loss": 4.461260795593262,
|
| 421 |
+
"rewards/accuracies": 0.6468750238418579,
|
| 422 |
+
"rewards/chosen": -2.005023956298828,
|
| 423 |
+
"rewards/margins": 7.896027565002441,
|
| 424 |
+
"rewards/rejected": -9.901052474975586,
|
| 425 |
+
"step": 130
|
| 426 |
+
},
|
| 427 |
+
{
|
| 428 |
+
"epoch": 1.274962962962963,
|
| 429 |
+
"grad_norm": 7289.251953125,
|
| 430 |
+
"learning_rate": 3.5362420368134356e-05,
|
| 431 |
+
"log_odds_chosen": 23.66111946105957,
|
| 432 |
+
"log_odds_ratio": -1.2839603424072266,
|
| 433 |
+
"logps/chosen": -2.44439959526062,
|
| 434 |
+
"logps/rejected": -25.9698429107666,
|
| 435 |
+
"loss": -294.9251,
|
| 436 |
+
"nll_loss": 3.672980546951294,
|
| 437 |
+
"rewards/accuracies": 0.71875,
|
| 438 |
+
"rewards/chosen": -1.22219979763031,
|
| 439 |
+
"rewards/margins": 11.762721061706543,
|
| 440 |
+
"rewards/rejected": -12.9849214553833,
|
| 441 |
+
"step": 135
|
| 442 |
+
},
|
| 443 |
+
{
|
| 444 |
+
"epoch": 1.3223703703703704,
|
| 445 |
+
"grad_norm": 35099.68359375,
|
| 446 |
+
"learning_rate": 3.408429731701635e-05,
|
| 447 |
+
"log_odds_chosen": 22.592227935791016,
|
| 448 |
+
"log_odds_ratio": -3.179147958755493,
|
| 449 |
+
"logps/chosen": -5.041490077972412,
|
| 450 |
+
"logps/rejected": -27.51487159729004,
|
| 451 |
+
"loss": -195.6627,
|
| 452 |
+
"nll_loss": 5.91623592376709,
|
| 453 |
+
"rewards/accuracies": 0.6875,
|
| 454 |
+
"rewards/chosen": -2.520745038986206,
|
| 455 |
+
"rewards/margins": 11.236689567565918,
|
| 456 |
+
"rewards/rejected": -13.75743579864502,
|
| 457 |
+
"step": 140
|
| 458 |
+
},
|
| 459 |
+
{
|
| 460 |
+
"epoch": 1.3697777777777778,
|
| 461 |
+
"grad_norm": 4246.3798828125,
|
| 462 |
+
"learning_rate": 3.2778194329621104e-05,
|
| 463 |
+
"log_odds_chosen": 20.304393768310547,
|
| 464 |
+
"log_odds_ratio": -2.3136236667633057,
|
| 465 |
+
"logps/chosen": -3.2953147888183594,
|
| 466 |
+
"logps/rejected": -23.475563049316406,
|
| 467 |
+
"loss": -214.76,
|
| 468 |
+
"nll_loss": 4.836905002593994,
|
| 469 |
+
"rewards/accuracies": 0.675000011920929,
|
| 470 |
+
"rewards/chosen": -1.6476573944091797,
|
| 471 |
+
"rewards/margins": 10.090123176574707,
|
| 472 |
+
"rewards/rejected": -11.737781524658203,
|
| 473 |
+
"step": 145
|
| 474 |
+
},
|
| 475 |
+
{
|
| 476 |
+
"epoch": 1.417185185185185,
|
| 477 |
+
"grad_norm": 112606.7421875,
|
| 478 |
+
"learning_rate": 3.144813424636031e-05,
|
| 479 |
+
"log_odds_chosen": 28.99297523498535,
|
| 480 |
+
"log_odds_ratio": -4.194870948791504,
|
| 481 |
+
"logps/chosen": -7.332457542419434,
|
| 482 |
+
"logps/rejected": -36.18639373779297,
|
| 483 |
+
"loss": -223.9742,
|
| 484 |
+
"nll_loss": 9.322813034057617,
|
| 485 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 486 |
+
"rewards/chosen": -3.666228771209717,
|
| 487 |
+
"rewards/margins": 14.426966667175293,
|
| 488 |
+
"rewards/rejected": -18.093196868896484,
|
| 489 |
+
"step": 150
|
| 490 |
+
},
|
| 491 |
+
{
|
| 492 |
+
"epoch": 1.4645925925925927,
|
| 493 |
+
"grad_norm": 57207.05859375,
|
| 494 |
+
"learning_rate": 3.0098213696293542e-05,
|
| 495 |
+
"log_odds_chosen": 27.4794864654541,
|
| 496 |
+
"log_odds_ratio": -6.622164726257324,
|
| 497 |
+
"logps/chosen": -13.51585578918457,
|
| 498 |
+
"logps/rejected": -40.87260818481445,
|
| 499 |
+
"loss": -2.1912,
|
| 500 |
+
"nll_loss": 16.61840057373047,
|
| 501 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 502 |
+
"rewards/chosen": -6.757927894592285,
|
| 503 |
+
"rewards/margins": 13.678375244140625,
|
| 504 |
+
"rewards/rejected": -20.436304092407227,
|
| 505 |
+
"step": 155
|
| 506 |
+
},
|
| 507 |
+
{
|
| 508 |
+
"epoch": 1.512,
|
| 509 |
+
"grad_norm": 22721.037109375,
|
| 510 |
+
"learning_rate": 2.8732590479375165e-05,
|
| 511 |
+
"log_odds_chosen": 26.662273406982422,
|
| 512 |
+
"log_odds_ratio": -6.387181282043457,
|
| 513 |
+
"logps/chosen": -10.372232437133789,
|
| 514 |
+
"logps/rejected": -36.92156219482422,
|
| 515 |
+
"loss": -90.2426,
|
| 516 |
+
"nll_loss": 9.191670417785645,
|
| 517 |
+
"rewards/accuracies": 0.643750011920929,
|
| 518 |
+
"rewards/chosen": -5.1861162185668945,
|
| 519 |
+
"rewards/margins": 13.274665832519531,
|
| 520 |
+
"rewards/rejected": -18.46078109741211,
|
| 521 |
+
"step": 160
|
| 522 |
+
},
|
| 523 |
+
{
|
| 524 |
+
"epoch": 1.5594074074074074,
|
| 525 |
+
"grad_norm": 8884.1220703125,
|
| 526 |
+
"learning_rate": 2.7355470760292956e-05,
|
| 527 |
+
"log_odds_chosen": 22.65861701965332,
|
| 528 |
+
"log_odds_ratio": -3.268022060394287,
|
| 529 |
+
"logps/chosen": -4.845406532287598,
|
| 530 |
+
"logps/rejected": -27.359100341796875,
|
| 531 |
+
"loss": -202.281,
|
| 532 |
+
"nll_loss": 5.931666374206543,
|
| 533 |
+
"rewards/accuracies": 0.684374988079071,
|
| 534 |
+
"rewards/chosen": -2.422703266143799,
|
| 535 |
+
"rewards/margins": 11.256847381591797,
|
| 536 |
+
"rewards/rejected": -13.679550170898438,
|
| 537 |
+
"step": 165
|
| 538 |
+
},
|
| 539 |
+
{
|
| 540 |
+
"epoch": 1.6068148148148147,
|
| 541 |
+
"grad_norm": 10747.482421875,
|
| 542 |
+
"learning_rate": 2.597109611334169e-05,
|
| 543 |
+
"log_odds_chosen": 24.835664749145508,
|
| 544 |
+
"log_odds_ratio": -5.039122104644775,
|
| 545 |
+
"logps/chosen": -8.912939071655273,
|
| 546 |
+
"logps/rejected": -33.64850616455078,
|
| 547 |
+
"loss": -108.0337,
|
| 548 |
+
"nll_loss": 9.482011795043945,
|
| 549 |
+
"rewards/accuracies": 0.6468750238418579,
|
| 550 |
+
"rewards/chosen": -4.456469535827637,
|
| 551 |
+
"rewards/margins": 12.367780685424805,
|
| 552 |
+
"rewards/rejected": -16.82425308227539,
|
| 553 |
+
"step": 170
|
| 554 |
+
},
|
| 555 |
+
{
|
| 556 |
+
"epoch": 1.6542222222222223,
|
| 557 |
+
"grad_norm": 25721.013671875,
|
| 558 |
+
"learning_rate": 2.458373045823404e-05,
|
| 559 |
+
"log_odds_chosen": 17.206180572509766,
|
| 560 |
+
"log_odds_ratio": -5.200994968414307,
|
| 561 |
+
"logps/chosen": -10.323943138122559,
|
| 562 |
+
"logps/rejected": -27.417028427124023,
|
| 563 |
+
"loss": 59.6029,
|
| 564 |
+
"nll_loss": 11.46679973602295,
|
| 565 |
+
"rewards/accuracies": 0.675000011920929,
|
| 566 |
+
"rewards/chosen": -5.161971569061279,
|
| 567 |
+
"rewards/margins": 8.546542167663574,
|
| 568 |
+
"rewards/rejected": -13.708514213562012,
|
| 569 |
+
"step": 175
|
| 570 |
+
},
|
| 571 |
+
{
|
| 572 |
+
"epoch": 1.7016296296296296,
|
| 573 |
+
"grad_norm": 5182.34716796875,
|
| 574 |
+
"learning_rate": 2.3197646927086697e-05,
|
| 575 |
+
"log_odds_chosen": 17.235403060913086,
|
| 576 |
+
"log_odds_ratio": -5.827118873596191,
|
| 577 |
+
"logps/chosen": -9.327442169189453,
|
| 578 |
+
"logps/rejected": -26.453588485717773,
|
| 579 |
+
"loss": 26.7508,
|
| 580 |
+
"nll_loss": 10.854573249816895,
|
| 581 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 582 |
+
"rewards/chosen": -4.663721084594727,
|
| 583 |
+
"rewards/margins": 8.56307315826416,
|
| 584 |
+
"rewards/rejected": -13.226794242858887,
|
| 585 |
+
"step": 180
|
| 586 |
+
},
|
| 587 |
+
{
|
| 588 |
+
"epoch": 1.749037037037037,
|
| 589 |
+
"grad_norm": 1122.926513671875,
|
| 590 |
+
"learning_rate": 2.1817114703032176e-05,
|
| 591 |
+
"log_odds_chosen": 12.183832168579102,
|
| 592 |
+
"log_odds_ratio": -3.2944283485412598,
|
| 593 |
+
"logps/chosen": -4.907401084899902,
|
| 594 |
+
"logps/rejected": -17.006296157836914,
|
| 595 |
+
"loss": -34.3607,
|
| 596 |
+
"nll_loss": 4.914076328277588,
|
| 597 |
+
"rewards/accuracies": 0.640625,
|
| 598 |
+
"rewards/chosen": -2.453700542449951,
|
| 599 |
+
"rewards/margins": 6.049447059631348,
|
| 600 |
+
"rewards/rejected": -8.503148078918457,
|
| 601 |
+
"step": 185
|
| 602 |
+
},
|
| 603 |
+
{
|
| 604 |
+
"epoch": 1.7964444444444445,
|
| 605 |
+
"grad_norm": 8495.58984375,
|
| 606 |
+
"learning_rate": 2.0446385870993467e-05,
|
| 607 |
+
"log_odds_chosen": 27.890201568603516,
|
| 608 |
+
"log_odds_ratio": -4.537432670593262,
|
| 609 |
+
"logps/chosen": -6.876814365386963,
|
| 610 |
+
"logps/rejected": -34.62470245361328,
|
| 611 |
+
"loss": -220.5601,
|
| 612 |
+
"nll_loss": 7.243471622467041,
|
| 613 |
+
"rewards/accuracies": 0.690625011920929,
|
| 614 |
+
"rewards/chosen": -3.4384071826934814,
|
| 615 |
+
"rewards/margins": 13.873942375183105,
|
| 616 |
+
"rewards/rejected": -17.31235122680664,
|
| 617 |
+
"step": 190
|
| 618 |
+
},
|
| 619 |
+
{
|
| 620 |
+
"epoch": 1.8438518518518519,
|
| 621 |
+
"grad_norm": 21205.203125,
|
| 622 |
+
"learning_rate": 1.9089682321121834e-05,
|
| 623 |
+
"log_odds_chosen": 28.1275577545166,
|
| 624 |
+
"log_odds_ratio": -10.069475173950195,
|
| 625 |
+
"logps/chosen": -13.262001037597656,
|
| 626 |
+
"logps/rejected": -41.2625732421875,
|
| 627 |
+
"loss": -20.6418,
|
| 628 |
+
"nll_loss": 14.18327522277832,
|
| 629 |
+
"rewards/accuracies": 0.671875,
|
| 630 |
+
"rewards/chosen": -6.631000518798828,
|
| 631 |
+
"rewards/margins": 14.000287055969238,
|
| 632 |
+
"rewards/rejected": -20.63128662109375,
|
| 633 |
+
"step": 195
|
| 634 |
+
},
|
| 635 |
+
{
|
| 636 |
+
"epoch": 1.8912592592592592,
|
| 637 |
+
"grad_norm": 74072.71875,
|
| 638 |
+
"learning_rate": 1.775118274523545e-05,
|
| 639 |
+
"log_odds_chosen": 39.34962844848633,
|
| 640 |
+
"log_odds_ratio": -16.783084869384766,
|
| 641 |
+
"logps/chosen": -29.612789154052734,
|
| 642 |
+
"logps/rejected": -68.8023452758789,
|
| 643 |
+
"loss": 324.4527,
|
| 644 |
+
"nll_loss": 29.834827423095703,
|
| 645 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 646 |
+
"rewards/chosen": -14.806394577026367,
|
| 647 |
+
"rewards/margins": 19.59477424621582,
|
| 648 |
+
"rewards/rejected": -34.40117263793945,
|
| 649 |
+
"step": 200
|
| 650 |
+
},
|
| 651 |
+
{
|
| 652 |
+
"epoch": 1.9386666666666668,
|
| 653 |
+
"grad_norm": 395353.21875,
|
| 654 |
+
"learning_rate": 1.643500976631037e-05,
|
| 655 |
+
"log_odds_chosen": 42.678619384765625,
|
| 656 |
+
"log_odds_ratio": -18.052345275878906,
|
| 657 |
+
"logps/chosen": -42.023616790771484,
|
| 658 |
+
"logps/rejected": -84.55274963378906,
|
| 659 |
+
"loss": 668.5252,
|
| 660 |
+
"nll_loss": 37.75287628173828,
|
| 661 |
+
"rewards/accuracies": 0.71875,
|
| 662 |
+
"rewards/chosen": -21.011808395385742,
|
| 663 |
+
"rewards/margins": 21.264570236206055,
|
| 664 |
+
"rewards/rejected": -42.27637481689453,
|
| 665 |
+
"step": 205
|
| 666 |
+
},
|
| 667 |
+
{
|
| 668 |
+
"epoch": 1.986074074074074,
|
| 669 |
+
"grad_norm": 80640.96875,
|
| 670 |
+
"learning_rate": 1.514521724066537e-05,
|
| 671 |
+
"log_odds_chosen": 39.148033142089844,
|
| 672 |
+
"log_odds_ratio": -19.23377799987793,
|
| 673 |
+
"logps/chosen": -32.76858901977539,
|
| 674 |
+
"logps/rejected": -71.80208587646484,
|
| 675 |
+
"loss": 426.931,
|
| 676 |
+
"nll_loss": 26.356372833251953,
|
| 677 |
+
"rewards/accuracies": 0.668749988079071,
|
| 678 |
+
"rewards/chosen": -16.384294509887695,
|
| 679 |
+
"rewards/margins": 19.51674461364746,
|
| 680 |
+
"rewards/rejected": -35.90104293823242,
|
| 681 |
+
"step": 210
|
| 682 |
+
},
|
| 683 |
+
{
|
| 684 |
+
"epoch": 2.0284444444444443,
|
| 685 |
+
"grad_norm": 34031.30859375,
|
| 686 |
+
"learning_rate": 1.3885777771950348e-05,
|
| 687 |
+
"log_odds_chosen": 39.754032135009766,
|
| 688 |
+
"log_odds_ratio": -14.231975555419922,
|
| 689 |
+
"logps/chosen": -18.812442779541016,
|
| 690 |
+
"logps/rejected": -58.38700485229492,
|
| 691 |
+
"loss": -24.6149,
|
| 692 |
+
"nll_loss": 17.645992279052734,
|
| 693 |
+
"rewards/accuracies": 0.6958041787147522,
|
| 694 |
+
"rewards/chosen": -9.406221389770508,
|
| 695 |
+
"rewards/margins": 19.787281036376953,
|
| 696 |
+
"rewards/rejected": -29.19350242614746,
|
| 697 |
+
"step": 215
|
| 698 |
+
},
|
| 699 |
+
{
|
| 700 |
+
"epoch": 2.075851851851852,
|
| 701 |
+
"grad_norm": 16462.95703125,
|
| 702 |
+
"learning_rate": 1.2660570475395683e-05,
|
| 703 |
+
"log_odds_chosen": 28.321569442749023,
|
| 704 |
+
"log_odds_ratio": -13.535270690917969,
|
| 705 |
+
"logps/chosen": -21.737045288085938,
|
| 706 |
+
"logps/rejected": -49.909423828125,
|
| 707 |
+
"loss": 248.0299,
|
| 708 |
+
"nll_loss": 19.69781494140625,
|
| 709 |
+
"rewards/accuracies": 0.6875,
|
| 710 |
+
"rewards/chosen": -10.868522644042969,
|
| 711 |
+
"rewards/margins": 14.086189270019531,
|
| 712 |
+
"rewards/rejected": -24.9547119140625,
|
| 713 |
+
"step": 220
|
| 714 |
+
},
|
| 715 |
+
{
|
| 716 |
+
"epoch": 2.1232592592592594,
|
| 717 |
+
"grad_norm": 17380.919921875,
|
| 718 |
+
"learning_rate": 1.1473369030008974e-05,
|
| 719 |
+
"log_odds_chosen": 41.30133819580078,
|
| 720 |
+
"log_odds_ratio": -5.992632865905762,
|
| 721 |
+
"logps/chosen": -9.869775772094727,
|
| 722 |
+
"logps/rejected": -51.008033752441406,
|
| 723 |
+
"loss": -338.6483,
|
| 724 |
+
"nll_loss": 9.406192779541016,
|
| 725 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 726 |
+
"rewards/chosen": -4.934887886047363,
|
| 727 |
+
"rewards/margins": 20.569129943847656,
|
| 728 |
+
"rewards/rejected": -25.504016876220703,
|
| 729 |
+
"step": 225
|
| 730 |
+
},
|
| 731 |
+
{
|
| 732 |
+
"epoch": 2.1706666666666665,
|
| 733 |
+
"grad_norm": 2604.8720703125,
|
| 734 |
+
"learning_rate": 1.0327830055518842e-05,
|
| 735 |
+
"log_odds_chosen": 40.362274169921875,
|
| 736 |
+
"log_odds_ratio": -9.142515182495117,
|
| 737 |
+
"logps/chosen": -14.164266586303711,
|
| 738 |
+
"logps/rejected": -54.35051345825195,
|
| 739 |
+
"loss": -186.1042,
|
| 740 |
+
"nll_loss": 13.237195014953613,
|
| 741 |
+
"rewards/accuracies": 0.684374988079071,
|
| 742 |
+
"rewards/chosen": -7.0821332931518555,
|
| 743 |
+
"rewards/margins": 20.093120574951172,
|
| 744 |
+
"rewards/rejected": -27.175256729125977,
|
| 745 |
+
"step": 230
|
| 746 |
+
},
|
| 747 |
+
{
|
| 748 |
+
"epoch": 2.218074074074074,
|
| 749 |
+
"grad_norm": 164766.734375,
|
| 750 |
+
"learning_rate": 9.227481849865235e-06,
|
| 751 |
+
"log_odds_chosen": 45.14253234863281,
|
| 752 |
+
"log_odds_ratio": -7.5437116622924805,
|
| 753 |
+
"logps/chosen": -11.828727722167969,
|
| 754 |
+
"logps/rejected": -56.79478073120117,
|
| 755 |
+
"loss": -336.8467,
|
| 756 |
+
"nll_loss": 12.229296684265137,
|
| 757 |
+
"rewards/accuracies": 0.721875011920929,
|
| 758 |
+
"rewards/chosen": -5.914363861083984,
|
| 759 |
+
"rewards/margins": 22.483028411865234,
|
| 760 |
+
"rewards/rejected": -28.397390365600586,
|
| 761 |
+
"step": 235
|
| 762 |
+
},
|
| 763 |
+
{
|
| 764 |
+
"epoch": 2.2654814814814817,
|
| 765 |
+
"grad_norm": 81824.2890625,
|
| 766 |
+
"learning_rate": 8.175713521924978e-06,
|
| 767 |
+
"log_odds_chosen": 40.605594635009766,
|
| 768 |
+
"log_odds_ratio": -11.402399063110352,
|
| 769 |
+
"logps/chosen": -26.724740982055664,
|
| 770 |
+
"logps/rejected": -67.15190124511719,
|
| 771 |
+
"loss": 212.4567,
|
| 772 |
+
"nll_loss": 23.99557113647461,
|
| 773 |
+
"rewards/accuracies": 0.721875011920929,
|
| 774 |
+
"rewards/chosen": -13.362370491027832,
|
| 775 |
+
"rewards/margins": 20.213586807250977,
|
| 776 |
+
"rewards/rejected": -33.575950622558594,
|
| 777 |
+
"step": 240
|
| 778 |
+
},
|
| 779 |
+
{
|
| 780 |
+
"epoch": 2.3128888888888888,
|
| 781 |
+
"grad_norm": 14960.7578125,
|
| 782 |
+
"learning_rate": 7.1757645529443665e-06,
|
| 783 |
+
"log_odds_chosen": 55.09406661987305,
|
| 784 |
+
"log_odds_ratio": -10.749176979064941,
|
| 785 |
+
"logps/chosen": -19.03034019470215,
|
| 786 |
+
"logps/rejected": -73.93666076660156,
|
| 787 |
+
"loss": -264.6577,
|
| 788 |
+
"nll_loss": 19.223167419433594,
|
| 789 |
+
"rewards/accuracies": 0.7593749761581421,
|
| 790 |
+
"rewards/chosen": -9.515170097351074,
|
| 791 |
+
"rewards/margins": 27.45315933227539,
|
| 792 |
+
"rewards/rejected": -36.96833038330078,
|
| 793 |
+
"step": 245
|
| 794 |
+
},
|
| 795 |
+
{
|
| 796 |
+
"epoch": 2.3602962962962963,
|
| 797 |
+
"grad_norm": 17807.7265625,
|
| 798 |
+
"learning_rate": 6.230714818829733e-06,
|
| 799 |
+
"log_odds_chosen": 45.54753875732422,
|
| 800 |
+
"log_odds_ratio": -7.856490135192871,
|
| 801 |
+
"logps/chosen": -11.773348808288574,
|
| 802 |
+
"logps/rejected": -57.12473678588867,
|
| 803 |
+
"loss": -344.7895,
|
| 804 |
+
"nll_loss": 12.78067684173584,
|
| 805 |
+
"rewards/accuracies": 0.7093750238418579,
|
| 806 |
+
"rewards/chosen": -5.886674404144287,
|
| 807 |
+
"rewards/margins": 22.67569351196289,
|
| 808 |
+
"rewards/rejected": -28.562368392944336,
|
| 809 |
+
"step": 250
|
| 810 |
+
},
|
| 811 |
+
{
|
| 812 |
+
"epoch": 2.407703703703704,
|
| 813 |
+
"grad_norm": 104304.0390625,
|
| 814 |
+
"learning_rate": 5.343475104027743e-06,
|
| 815 |
+
"log_odds_chosen": 47.276451110839844,
|
| 816 |
+
"log_odds_ratio": -8.991453170776367,
|
| 817 |
+
"logps/chosen": -13.746038436889648,
|
| 818 |
+
"logps/rejected": -60.85129928588867,
|
| 819 |
+
"loss": -309.5943,
|
| 820 |
+
"nll_loss": 14.252038955688477,
|
| 821 |
+
"rewards/accuracies": 0.7281249761581421,
|
| 822 |
+
"rewards/chosen": -6.873019218444824,
|
| 823 |
+
"rewards/margins": 23.552631378173828,
|
| 824 |
+
"rewards/rejected": -30.425649642944336,
|
| 825 |
+
"step": 255
|
| 826 |
+
},
|
| 827 |
+
{
|
| 828 |
+
"epoch": 2.455111111111111,
|
| 829 |
+
"grad_norm": 38147.03125,
|
| 830 |
+
"learning_rate": 4.516778136213037e-06,
|
| 831 |
+
"log_odds_chosen": 40.15815734863281,
|
| 832 |
+
"log_odds_ratio": -7.543868064880371,
|
| 833 |
+
"logps/chosen": -11.01616096496582,
|
| 834 |
+
"logps/rejected": -50.988121032714844,
|
| 835 |
+
"loss": -282.8974,
|
| 836 |
+
"nll_loss": 12.92004108428955,
|
| 837 |
+
"rewards/accuracies": 0.731249988079071,
|
| 838 |
+
"rewards/chosen": -5.50808048248291,
|
| 839 |
+
"rewards/margins": 19.985979080200195,
|
| 840 |
+
"rewards/rejected": -25.494060516357422,
|
| 841 |
+
"step": 260
|
| 842 |
+
},
|
| 843 |
+
{
|
| 844 |
+
"epoch": 2.5025185185185186,
|
| 845 |
+
"grad_norm": 24916.00390625,
|
| 846 |
+
"learning_rate": 3.7531701693965554e-06,
|
| 847 |
+
"log_odds_chosen": 50.416419982910156,
|
| 848 |
+
"log_odds_ratio": -8.888219833374023,
|
| 849 |
+
"logps/chosen": -14.691574096679688,
|
| 850 |
+
"logps/rejected": -64.95278930664062,
|
| 851 |
+
"loss": -330.3818,
|
| 852 |
+
"nll_loss": 14.175634384155273,
|
| 853 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 854 |
+
"rewards/chosen": -7.345787048339844,
|
| 855 |
+
"rewards/margins": 25.1306095123291,
|
| 856 |
+
"rewards/rejected": -32.47639465332031,
|
| 857 |
+
"step": 265
|
| 858 |
+
},
|
| 859 |
+
{
|
| 860 |
+
"epoch": 2.549925925925926,
|
| 861 |
+
"grad_norm": 17820.6953125,
|
| 862 |
+
"learning_rate": 3.055003141378948e-06,
|
| 863 |
+
"log_odds_chosen": 56.49353790283203,
|
| 864 |
+
"log_odds_ratio": -4.24954891204834,
|
| 865 |
+
"logps/chosen": -7.077002048492432,
|
| 866 |
+
"logps/rejected": -63.38737106323242,
|
| 867 |
+
"loss": -670.0222,
|
| 868 |
+
"nll_loss": 8.986559867858887,
|
| 869 |
+
"rewards/accuracies": 0.734375,
|
| 870 |
+
"rewards/chosen": -3.538501024246216,
|
| 871 |
+
"rewards/margins": 28.155181884765625,
|
| 872 |
+
"rewards/rejected": -31.69368553161621,
|
| 873 |
+
"step": 270
|
| 874 |
+
},
|
| 875 |
+
{
|
| 876 |
+
"epoch": 2.5973333333333333,
|
| 877 |
+
"grad_norm": 21286.701171875,
|
| 878 |
+
"learning_rate": 2.424427429704365e-06,
|
| 879 |
+
"log_odds_chosen": 45.696632385253906,
|
| 880 |
+
"log_odds_ratio": -8.857467651367188,
|
| 881 |
+
"logps/chosen": -13.247896194458008,
|
| 882 |
+
"logps/rejected": -58.740142822265625,
|
| 883 |
+
"loss": -299.5551,
|
| 884 |
+
"nll_loss": 12.918680191040039,
|
| 885 |
+
"rewards/accuracies": 0.778124988079071,
|
| 886 |
+
"rewards/chosen": -6.623948097229004,
|
| 887 |
+
"rewards/margins": 22.746124267578125,
|
| 888 |
+
"rewards/rejected": -29.370071411132812,
|
| 889 |
+
"step": 275
|
| 890 |
+
},
|
| 891 |
+
{
|
| 892 |
+
"epoch": 2.644740740740741,
|
| 893 |
+
"grad_norm": 50187.1171875,
|
| 894 |
+
"learning_rate": 1.8633852284264508e-06,
|
| 895 |
+
"log_odds_chosen": 55.4704704284668,
|
| 896 |
+
"log_odds_ratio": -6.3292717933654785,
|
| 897 |
+
"logps/chosen": -11.759183883666992,
|
| 898 |
+
"logps/rejected": -67.04707336425781,
|
| 899 |
+
"loss": -503.517,
|
| 900 |
+
"nll_loss": 12.290529251098633,
|
| 901 |
+
"rewards/accuracies": 0.734375,
|
| 902 |
+
"rewards/chosen": -5.879591941833496,
|
| 903 |
+
"rewards/margins": 27.643945693969727,
|
| 904 |
+
"rewards/rejected": -33.523536682128906,
|
| 905 |
+
"step": 280
|
| 906 |
+
},
|
| 907 |
+
{
|
| 908 |
+
"epoch": 2.6921481481481484,
|
| 909 |
+
"grad_norm": 114461.5,
|
| 910 |
+
"learning_rate": 1.3736045660864034e-06,
|
| 911 |
+
"log_odds_chosen": 41.011329650878906,
|
| 912 |
+
"log_odds_ratio": -8.477639198303223,
|
| 913 |
+
"logps/chosen": -12.918500900268555,
|
| 914 |
+
"logps/rejected": -53.76482391357422,
|
| 915 |
+
"loss": -236.1638,
|
| 916 |
+
"nll_loss": 13.538793563842773,
|
| 917 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 918 |
+
"rewards/chosen": -6.459250450134277,
|
| 919 |
+
"rewards/margins": 20.423160552978516,
|
| 920 |
+
"rewards/rejected": -26.88241195678711,
|
| 921 |
+
"step": 285
|
| 922 |
+
},
|
| 923 |
+
{
|
| 924 |
+
"epoch": 2.7395555555555555,
|
| 925 |
+
"grad_norm": 26927.990234375,
|
| 926 |
+
"learning_rate": 9.565939833279192e-07,
|
| 927 |
+
"log_odds_chosen": 40.3164176940918,
|
| 928 |
+
"log_odds_ratio": -15.15136432647705,
|
| 929 |
+
"logps/chosen": -22.780040740966797,
|
| 930 |
+
"logps/rejected": -62.926292419433594,
|
| 931 |
+
"loss": 90.1159,
|
| 932 |
+
"nll_loss": 20.46670913696289,
|
| 933 |
+
"rewards/accuracies": 0.6875,
|
| 934 |
+
"rewards/chosen": -11.390020370483398,
|
| 935 |
+
"rewards/margins": 20.073129653930664,
|
| 936 |
+
"rewards/rejected": -31.463146209716797,
|
| 937 |
+
"step": 290
|
| 938 |
+
},
|
| 939 |
+
{
|
| 940 |
+
"epoch": 2.786962962962963,
|
| 941 |
+
"grad_norm": 10974.4052734375,
|
| 942 |
+
"learning_rate": 6.136378865420872e-07,
|
| 943 |
+
"log_odds_chosen": 47.6572380065918,
|
| 944 |
+
"log_odds_ratio": -7.081920623779297,
|
| 945 |
+
"logps/chosen": -10.659754753112793,
|
| 946 |
+
"logps/rejected": -58.14702224731445,
|
| 947 |
+
"loss": -414.5048,
|
| 948 |
+
"nll_loss": 11.201601028442383,
|
| 949 |
+
"rewards/accuracies": 0.7093750238418579,
|
| 950 |
+
"rewards/chosen": -5.3298773765563965,
|
| 951 |
+
"rewards/margins": 23.743627548217773,
|
| 952 |
+
"rewards/rejected": -29.073511123657227,
|
| 953 |
+
"step": 295
|
| 954 |
+
},
|
| 955 |
+
{
|
| 956 |
+
"epoch": 2.83437037037037,
|
| 957 |
+
"grad_norm": 42624.33203125,
|
| 958 |
+
"learning_rate": 3.45792591853214e-07,
|
| 959 |
+
"log_odds_chosen": 52.529335021972656,
|
| 960 |
+
"log_odds_ratio": -6.441692352294922,
|
| 961 |
+
"logps/chosen": -9.135614395141602,
|
| 962 |
+
"logps/rejected": -61.50330352783203,
|
| 963 |
+
"loss": -541.3402,
|
| 964 |
+
"nll_loss": 10.765565872192383,
|
| 965 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 966 |
+
"rewards/chosen": -4.567807197570801,
|
| 967 |
+
"rewards/margins": 26.1838436126709,
|
| 968 |
+
"rewards/rejected": -30.751651763916016,
|
| 969 |
+
"step": 300
|
| 970 |
+
},
|
| 971 |
+
{
|
| 972 |
+
"epoch": 2.8817777777777778,
|
| 973 |
+
"grad_norm": 8348.119140625,
|
| 974 |
+
"learning_rate": 1.538830716302092e-07,
|
| 975 |
+
"log_odds_chosen": 44.924232482910156,
|
| 976 |
+
"log_odds_ratio": -9.396774291992188,
|
| 977 |
+
"logps/chosen": -14.924581527709961,
|
| 978 |
+
"logps/rejected": -59.64234161376953,
|
| 979 |
+
"loss": -233.7471,
|
| 980 |
+
"nll_loss": 14.424337387084961,
|
| 981 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 982 |
+
"rewards/chosen": -7.4622907638549805,
|
| 983 |
+
"rewards/margins": 22.3588809967041,
|
| 984 |
+
"rewards/rejected": -29.821170806884766,
|
| 985 |
+
"step": 305
|
| 986 |
+
},
|
| 987 |
+
{
|
| 988 |
+
"epoch": 2.9291851851851853,
|
| 989 |
+
"grad_norm": 8581.84765625,
|
| 990 |
+
"learning_rate": 3.8500413544415025e-08,
|
| 991 |
+
"log_odds_chosen": 50.05999755859375,
|
| 992 |
+
"log_odds_ratio": -14.670018196105957,
|
| 993 |
+
"logps/chosen": -20.66399574279785,
|
| 994 |
+
"logps/rejected": -70.54561614990234,
|
| 995 |
+
"loss": -132.5917,
|
| 996 |
+
"nll_loss": 18.822933197021484,
|
| 997 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 998 |
+
"rewards/chosen": -10.331997871398926,
|
| 999 |
+
"rewards/margins": 24.940807342529297,
|
| 1000 |
+
"rewards/rejected": -35.27280807495117,
|
| 1001 |
+
"step": 310
|
| 1002 |
+
},
|
| 1003 |
+
{
|
| 1004 |
+
"epoch": 2.9765925925925925,
|
| 1005 |
+
"grad_norm": 8825.96875,
|
| 1006 |
+
"learning_rate": 0.0,
|
| 1007 |
+
"log_odds_chosen": 60.088844299316406,
|
| 1008 |
+
"log_odds_ratio": -5.067689418792725,
|
| 1009 |
+
"logps/chosen": -8.742526054382324,
|
| 1010 |
+
"logps/rejected": -68.63340759277344,
|
| 1011 |
+
"loss": -673.9993,
|
| 1012 |
+
"nll_loss": 10.168830871582031,
|
| 1013 |
+
"rewards/accuracies": 0.75,
|
| 1014 |
+
"rewards/chosen": -4.371263027191162,
|
| 1015 |
+
"rewards/margins": 29.945446014404297,
|
| 1016 |
+
"rewards/rejected": -34.31670379638672,
|
| 1017 |
+
"step": 315
|
| 1018 |
+
},
|
| 1019 |
+
{
|
| 1020 |
+
"epoch": 2.9765925925925925,
|
| 1021 |
+
"step": 315,
|
| 1022 |
+
"total_flos": 0.0,
|
| 1023 |
+
"train_loss": -61.895723918127636,
|
| 1024 |
+
"train_runtime": 9291.6158,
|
| 1025 |
+
"train_samples_per_second": 2.179,
|
| 1026 |
+
"train_steps_per_second": 0.034
|
| 1027 |
+
}
|
| 1028 |
+
],
|
| 1029 |
+
"logging_steps": 5,
|
| 1030 |
+
"max_steps": 315,
|
| 1031 |
+
"num_input_tokens_seen": 0,
|
| 1032 |
+
"num_train_epochs": 3,
|
| 1033 |
+
"save_steps": 100000,
|
| 1034 |
+
"stateful_callbacks": {
|
| 1035 |
+
"TrainerControl": {
|
| 1036 |
+
"args": {
|
| 1037 |
+
"should_epoch_stop": false,
|
| 1038 |
+
"should_evaluate": false,
|
| 1039 |
+
"should_log": false,
|
| 1040 |
+
"should_save": true,
|
| 1041 |
+
"should_training_stop": true
|
| 1042 |
+
},
|
| 1043 |
+
"attributes": {}
|
| 1044 |
+
}
|
| 1045 |
+
},
|
| 1046 |
+
"total_flos": 0.0,
|
| 1047 |
+
"train_batch_size": 1,
|
| 1048 |
+
"trial_name": null,
|
| 1049 |
+
"trial_params": null
|
| 1050 |
+
}
|