msimpo-30each-v2
Browse files
README.md
ADDED
@@ -0,0 +1,69 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
base_model: meta-llama/Meta-Llama-3.1-8B-Instruct
|
3 |
+
library_name: transformers
|
4 |
+
model_name: msimpo-30each-v2
|
5 |
+
tags:
|
6 |
+
- generated_from_trainer
|
7 |
+
- trl
|
8 |
+
- cpo
|
9 |
+
licence: license
|
10 |
+
---
|
11 |
+
|
12 |
+
# Model Card for msimpo-30each-v2
|
13 |
+
|
14 |
+
This model is a fine-tuned version of [meta-llama/Meta-Llama-3.1-8B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3.1-8B-Instruct).
|
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="nomadrp/msimpo-30each-v2", device="cuda")
|
24 |
+
output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
|
25 |
+
print(output["generated_text"])
|
26 |
+
```
|
27 |
+
|
28 |
+
## Training procedure
|
29 |
+
|
30 |
+
|
31 |
+
|
32 |
+
|
33 |
+
This model was trained with CPO, a method introduced in [Contrastive Preference Optimization: Pushing the Boundaries of LLM Performance in Machine Translation](https://huggingface.co/papers/2401.08417).
|
34 |
+
|
35 |
+
### Framework versions
|
36 |
+
|
37 |
+
- TRL: 0.15.0.dev0
|
38 |
+
- Transformers: 4.48.2
|
39 |
+
- Pytorch: 2.2.0+cu118
|
40 |
+
- Datasets: 3.2.0
|
41 |
+
- Tokenizers: 0.21.0
|
42 |
+
|
43 |
+
## Citations
|
44 |
+
|
45 |
+
Cite CPO as:
|
46 |
+
|
47 |
+
```bibtex
|
48 |
+
@inproceedings{xu2024contrastive,
|
49 |
+
title = {{Contrastive Preference Optimization: Pushing the Boundaries of LLM Performance in Machine Translation}},
|
50 |
+
author = {Haoran Xu and Amr Sharaf and Yunmo Chen and Weiting Tan and Lingfeng Shen and Benjamin Van Durme and Kenton Murray and Young Jin Kim},
|
51 |
+
year = 2024,
|
52 |
+
booktitle = {Forty-first International Conference on Machine Learning, {ICML} 2024, Vienna, Austria, July 21-27, 2024},
|
53 |
+
publisher = {OpenReview.net},
|
54 |
+
url = {https://openreview.net/forum?id=51iwkioZpn}
|
55 |
+
}
|
56 |
+
```
|
57 |
+
|
58 |
+
Cite TRL as:
|
59 |
+
|
60 |
+
```bibtex
|
61 |
+
@misc{vonwerra2022trl,
|
62 |
+
title = {{TRL: Transformer Reinforcement Learning}},
|
63 |
+
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},
|
64 |
+
year = 2020,
|
65 |
+
journal = {GitHub repository},
|
66 |
+
publisher = {GitHub},
|
67 |
+
howpublished = {\url{https://github.com/huggingface/trl}}
|
68 |
+
}
|
69 |
+
```
|