File size: 3,545 Bytes
3144609
 
6e2b3c7
cce246a
3144609
6e2b3c7
3144609
 
6e2b3c7
 
3144609
 
 
 
6e2b3c7
3144609
 
 
 
6e2b3c7
 
3144609
 
6e2b3c7
 
ccb6096
6e2b3c7
 
 
 
 
3144609
 
 
6e2b3c7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3144609
 
 
7faa21d
6e2b3c7
4b9cecf
 
 
 
 
 
 
7faa21d
 
 
 
 
 
3144609
6e2b3c7
3144609
6e2b3c7
3144609
 
 
 
 
6e2b3c7
3144609
6e2b3c7
3144609
 
 
6e2b3c7
 
 
 
3144609
 
6e2b3c7
 
 
 
 
 
 
 
3144609
 
6e2b3c7
 
 
 
3144609
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
---
tags:
    - text-generation
license: cc-by-nc-sa-4.0
language:
    - ko
base_model: yanolja/KoSOLAR-10.7B-v0.1
pipeline_tag: text-generation
datasets:
    - jojo0217/korean_rlhf_dataset
---

# **DataVortexS-10.7B-v0.3**

<img src="./DataVortex.png" alt="DataVortex" style="height: 8em;">

## **Model Details**

### **Base Model**

[yanolja/KoSOLAR-10.7B-v0.1](https://huggingface.co/yanolja/KoSOLAR-10.7B-v0.1)

### **Trained On**

-   **OS**: Ubuntu 20.04
-   **GPU**: H100 80GB 1ea
-   **transformers**: v4.36.2

### **Dataset**

-   [jojo0217/korean_rlhf_dataset](https://huggingface.co/datasets/jojo0217/korean_rlhf_dataset)

### **Instruction format**

It follows **Alpaca** format.

E.g.

```python
text = """\
당신은 사람들이 정보를 찾을 수 있도록 도와주는 인공지능 비서입니다.

### Instruction:
대한민국의 수도는 어디야?

### Response:
대한민국의 수도는 서울입니다.

### Instruction:
서울 인구는 총 몇 명이야?
"""
```

## **Model Benchmark**

### **[Ko LM Eval Harnesss](https://github.com/Beomi/ko-lm-evaluation-harness)**

| Task             |         0-shot |         5-shot |       10-shot |        50-shot |
| :--------------- | -------------: | -------------: | ------------: | -------------: |
| kobest_boolq     |       0.606754 |       0.553485 |      0.583201 |       0.587602 |
| kobest_copa      |       0.603643 |       0.625567 |      0.618533 |       0.627404 |
| kobest_hellaswag |       0.360793 |       0.366002 |       0.37105 |       0.357439 |
| kobest_sentineg  |       0.652929 |       0.751097 |      0.742426 |       0.760152 |
| **Average**      | **0.55602975** | **0.57403775** | **0.5788025** | **0.58314925** |

### **[Ko-LLM-Leaderboard](https://huggingface.co/spaces/upstage/open-ko-llm-leaderboard)**

| Average | Ko-ARC | Ko-HellaSwag | Ko-MMLU | Ko-TruthfulQA | Ko-CommonGen V2 |
| ------: | -----: | -----------: | ------: | ------------: | --------------: |
|   37.57 |  33.87 |        42.47 |   28.21 |         46.09 |           37.19 |

## **Implementation Code**

This model contains the chat_template instruction format.  
You can use the code below.

```python
from transformers import AutoModelForCausalLM, AutoTokenizer

device = "cuda" # the device to load the model onto

model = AutoModelForCausalLM.from_pretrained("Edentns/DataVortexS-10.7B-v0.3")
tokenizer = AutoTokenizer.from_pretrained("Edentns/DataVortexS-10.7B-v0.3")

messages = [
    {"role": "system", "content": "당신은 사람들이 정보를 찾을 수 있도록 도와주는 인공지능 비서입니다."},
    {"role": "user", "content": "대한민국의 수도는 어디야?"},
    {"role": "assistant", "content": "대한민국의 수도는 서울입니다."},
    {"role": "user", "content": "서울 인구는 총 몇 명이야?"}
]

encodeds = tokenizer.apply_chat_template(messages, return_tensors="pt")

model_inputs = encodeds.to(device)
model.to(device)

generated_ids = model.generate(model_inputs, max_new_tokens=1000, do_sample=True)
decoded = tokenizer.batch_decode(generated_ids)
print(decoded[0])
```

## **License**

The model is licensed under the [cc-by-nc-sa-4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/) license, which allows others to copy, modify, and share the work non-commercially, as long as they give appropriate credit and distribute any derivative works under the same license.

<div align="center">
    <a href="https://edentns.com/">
        <img src="./Logo.png" alt="Logo" style="height: 3em;">
    </a>
</div>