Update README.md
Browse files
README.md
CHANGED
|
@@ -13,19 +13,25 @@ should probably proofread and complete it, then remove this comment. -->
|
|
| 13 |
# ENERGY-DRINK-LOVE/eeve_dpo-v3
|
| 14 |
|
| 15 |
### Our Team
|
| 16 |
-
|
| 17 |
-
|
| 18 |
|
| 19 |
## Model
|
| 20 |
|
| 21 |
### Base Model
|
| 22 |
-
*
|
| 23 |
|
| 24 |
### Hardware and Software
|
|
|
|
|
|
|
| 25 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 26 |
|
| 27 |
### Training Method
|
| 28 |
-
|
| 29 |
|
| 30 |
## Benchmark
|
| 31 |
|
|
|
|
| 13 |
# ENERGY-DRINK-LOVE/eeve_dpo-v3
|
| 14 |
|
| 15 |
### Our Team
|
| 16 |
+
* Jingyeom Kim
|
| 17 |
+
* Youjin Chung
|
| 18 |
|
| 19 |
## Model
|
| 20 |
|
| 21 |
### Base Model
|
| 22 |
+
* [yanolja/EEVE-Korean-Instruct-10.8B-v1.0](https://huggingface.co/yanolja/EEVE-Korean-Instruct-10.8B-v1.0)
|
| 23 |
|
| 24 |
### Hardware and Software
|
| 25 |
+
* Hardware: A100 * 8 for training our model
|
| 26 |
+
* Deepspeed library & Huggingface TRL Trainer
|
| 27 |
|
| 28 |
+
### Dataset
|
| 29 |
+
* DPO_dataset
|
| 30 |
+
* ์์ฒด ์ ์ dpo dataset(AI-hub dataset ํ์ฉ)
|
| 31 |
+
* OpenOrca DPO ๋ฑ ์์ด ๋ฐ์ดํฐ์
๋ฒ์ญ(์์ฒด ๋ชจ๋ธ ํ์ฉ)
|
| 32 |
|
| 33 |
### Training Method
|
| 34 |
+
* [DPO](https://arxiv.org/abs/2305.18290)
|
| 35 |
|
| 36 |
## Benchmark
|
| 37 |
|