myunghoonse
commited on
Commit
•
a904502
1
Parent(s):
a15b503
Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,58 @@
|
|
1 |
-
---
|
2 |
-
|
3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
language:
|
3 |
+
- ko
|
4 |
+
library_name: transformers
|
5 |
+
pipeline_tag: text-generation
|
6 |
+
license: apache-2.0
|
7 |
+
---
|
8 |
+
|
9 |
+
# **SOLAR-10.7B-v1.0-Instruct**
|
10 |
+
|
11 |
+
## Model Details
|
12 |
+
**Model Developers**
|
13 |
+
- myeonghoon kim
|
14 |
+
|
15 |
+
**Model Architecture**
|
16 |
+
- SOLAR-10.7B-v1.0-Instruct is an auto-regressive language model based on the LLaMA2 transformer architecture.
|
17 |
+
|
18 |
+
**Base Model**
|
19 |
+
- [upstage/SOLAR-10.7B-v1.0](https://huggingface.co/upstage/SOLAR-10.7B-v1.0)
|
20 |
+
|
21 |
+
**Training Dataset**
|
22 |
+
- [Platypus](https://github.com/arielnlee/Platypus)
|
23 |
+
|
24 |
+
---
|
25 |
+
# Model comparisons1
|
26 |
+
> Ko-LLM leaderboard(11/23; [link](https://huggingface.co/spaces/upstage/open-ko-llm-leaderboard))
|
27 |
+
|
28 |
+
| Model | Average | Ko-ARC | Ko-HellaSwag | Ko-MMLU | Ko-TruthfulQA | Ko-CommonGen V2 |
|
29 |
+
| --- | --- | --- | --- | --- | --- | --- |
|
30 |
+
| **[...your_model_name...]** | NaN | NaN | NaN | NaN | NaN | NaN |
|
31 |
+
|
32 |
+
---
|
33 |
+
# Model comparisons2
|
34 |
+
> AI-Harness evaluation; [link](https://github.com/Beomi/ko-lm-evaluation-harness)
|
35 |
+
|
36 |
+
| Model | Copa | Copa | HellaSwag | HellaSwag | BoolQ | BoolQ | Sentineg | Sentineg |
|
37 |
+
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
|
38 |
+
| | 0-shot | 5-shot | 0-shot | 5-shot | 0-shot | 5-shot | 0-shot | 5-shot |
|
39 |
+
| **SOLAR-10.7B-v1.0-Instruct** | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN |
|
40 |
+
|
41 |
+
---
|
42 |
+
# Implementation Code
|
43 |
+
```python
|
44 |
+
### KO-Platypus
|
45 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
46 |
+
import torch
|
47 |
+
|
48 |
+
repo = "[...your_model_repo...]"
|
49 |
+
OpenOrca = AutoModelForCausalLM.from_pretrained(
|
50 |
+
repo,
|
51 |
+
return_dict=True,
|
52 |
+
torch_dtype=torch.float16,
|
53 |
+
device_map='auto'
|
54 |
+
)
|
55 |
+
OpenOrca_tokenizer = AutoTokenizer.from_pretrained(repo)
|
56 |
+
```
|
57 |
+
|
58 |
+
---
|