Update README.md
Browse files
README.md
CHANGED
@@ -7,35 +7,20 @@ pipeline_tag: text-generation
|
|
7 |
datasets:
|
8 |
- maywell/ko_Ultrafeedback_binarized
|
9 |
---
|
10 |
-
|
11 |
-
|
|
|
|
|
12 |
|
13 |
**Base Model**
|
14 |
- [beomi/OPEN-SOLAR-KO-10.7B](https://huggingface.co/beomi/OPEN-SOLAR-KO-10.7B)
|
15 |
|
16 |
-
**
|
17 |
- [maywell/ko_Ultrafeedback_binarized](https://huggingface.co/datasets/maywell/ko_Ultrafeedback_binarized)
|
18 |
|
19 |
-
**Explanation**
|
20 |
-
- With base model, apply DPO to the small amount of layers with the dataset written above, saved just an adapter part.
|
21 |
-
- Merge the base model and the tuned adapter together.
|
22 |
-
|
23 |
-
**Usage**
|
24 |
-
``` python
|
25 |
-
from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
|
26 |
-
model = AutoModelForCausalLM.from_pretrained(
|
27 |
-
"dddsaty/Open_Ko_SOLAR_DPO_Merge_v0.1",
|
28 |
-
low_cpu_mem_usage = True,
|
29 |
-
torch_dtype = torch.float16,
|
30 |
-
device_map = device_map,
|
31 |
-
)
|
32 |
-
|
33 |
-
tokenizer = AutoTokenizer.from_pretrained("dddsaty/Open_Ko_SOLAR_DPO_Merge_v0.1")
|
34 |
-
|
35 |
-
```
|
36 |
-
|
37 |
**Log**
|
38 |
- 2024.01.25: Initial version Upload
|
|
|
39 |
|
40 |
**LICENSE**
|
41 |
- Apache 2.0
|
|
|
7 |
datasets:
|
8 |
- maywell/ko_Ultrafeedback_binarized
|
9 |
---
|
10 |
+
|
11 |
+
**Explanation**
|
12 |
+
- With the base model, applied DPO to the small amount of layers with the open dataset , saved just the adapter part
|
13 |
+
- Merged the base model and the tuned adapter together
|
14 |
|
15 |
**Base Model**
|
16 |
- [beomi/OPEN-SOLAR-KO-10.7B](https://huggingface.co/beomi/OPEN-SOLAR-KO-10.7B)
|
17 |
|
18 |
+
**Used Corpus**
|
19 |
- [maywell/ko_Ultrafeedback_binarized](https://huggingface.co/datasets/maywell/ko_Ultrafeedback_binarized)
|
20 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
21 |
**Log**
|
22 |
- 2024.01.25: Initial version Upload
|
23 |
+
- 2024.02.10: Readme updated
|
24 |
|
25 |
**LICENSE**
|
26 |
- Apache 2.0
|