crimsonjoo
commited on
Commit
β’
181e3ba
1
Parent(s):
25b7d16
update neversleeps
Browse files
README.md
CHANGED
@@ -1,3 +1,61 @@
|
|
1 |
---
|
2 |
license: apache-2.0
|
|
|
|
|
|
|
|
|
3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
license: apache-2.0
|
3 |
+
tags:
|
4 |
+
- generated_from_trainer
|
5 |
+
base_model: yanolja/EEVE-Korean-2.8B-v1.0
|
6 |
+
|
7 |
---
|
8 |
+
<p align="left">
|
9 |
+
<img src="https://huggingface.co/Neversleep-3B-Instruct-v0.1/resolve/main/eeve_logo.webp" width="100%"/>
|
10 |
+
<p>
|
11 |
+
|
12 |
+
# "We must sleep, but AI Never Sleeps!"
|
13 |
+
|
14 |
+
|
15 |
+
## Prompt Template
|
16 |
+
```
|
17 |
+
A chat between a curious user and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the user's questions.
|
18 |
+
Human: {prompt}
|
19 |
+
Assistant:
|
20 |
+
```
|
21 |
+
## Simple-Usage
|
22 |
+
```python
|
23 |
+
from transformers import AutoTokenizer
|
24 |
+
from transformers import AutoModelForCausalLM
|
25 |
+
|
26 |
+
model = AutoModelForCausalLM.from_pretrained("yanolja/EEVE-Korean-Instruct-2.8B-v1.0", trust_remote_code=True)
|
27 |
+
tokenizer = AutoTokenizer.from_pretrained("yanolja/EEVE-Korean-Instruct-2.8B-v1.0", trust_remote_code=True)
|
28 |
+
|
29 |
+
prompt_template = "A chat between a curious user and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the user's questions.\nHuman: {prompt}\nAssistant:\n"
|
30 |
+
text = 'λ€μ΄μ΄νΈμ λ©λ΄λ₯Ό μΆμ²ν΄μ£ΌμΈμ.\n\n(A) μλ¬λ\n(B) μΉν¨\n(C) νΌμ\n(D) νμ€ν'
|
31 |
+
model_inputs = tokenizer(prompt_template.format(prompt=text), return_tensors='pt')
|
32 |
+
|
33 |
+
outputs = model.generate(**model_inputs, max_new_tokens=256)
|
34 |
+
output_text = tokenizer.batch_decode(outputs, skip_special_tokens=True)[0]
|
35 |
+
print(output_text)
|
36 |
+
```
|
37 |
+
|
38 |
+
### Example Output
|
39 |
+
```
|
40 |
+
A chat between a curious user and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the user's questions.
|
41 |
+
Human: λ€μ΄μ΄νΈμ λ©λ΄λ₯Ό μΆμ²ν΄μ£ΌμΈμ.
|
42 |
+
|
43 |
+
(A) μλ¬λ
|
44 |
+
(B) μΉν¨
|
45 |
+
(C) νΌμ
|
46 |
+
(D) νμ€ν
|
47 |
+
Assistant:
|
48 |
+
(A) μλ¬λλ₯Ό μΆμ²λ립λλ€. μλ¬λλ μ μΉΌλ‘리μ΄λ©΄μλ μμμκ° νλΆν΄ λ€μ΄μ΄νΈμμΌλ‘ μ ν©ν©λλ€. λ€μν μ±μμ λ¨λ°±μ§μ μΆκ°νμ¬ κ· ν μ‘ν μμ¬λ₯Ό λ§λμ€ μ μμ΅λλ€.
|
49 |
+
```
|
50 |
+
|
51 |
+
## About the Model
|
52 |
+
|
53 |
+
First of all, Overwhelming gratitude to 'yanolja/EEVE' Model & Team!
|
54 |
+
This model is a fine-tuned version of [crimsonjoo/Neversleep-3B-v0.1](https://huggingface.co/crimsonjoo/Neversleep-3B-v0.1), which is a Korean vocabulary-extended version of [microsoft/phi-2](https://huggingface.co/microsoft/phi-2). Specifically, we utilized Direct Preference Optimization (DPO) through the use of [Axolotl](https://github.com/OpenAccess-AI-Collective/axolotl).
|
55 |
+
|
56 |
+
For more details, please refer to our technical report: [Efficient and Effective Vocabulary Expansion Towards Multilingual Large Language Models](https://arxiv.org/abs/2402.14714).
|
57 |
+
|
58 |
+
## Training Data
|
59 |
+
- Korean-translated version of [Open-Orca/SlimOrca-Dedup](https://huggingface.co/datasets/Open-Orca/SlimOrca-Dedup)
|
60 |
+
- Korean-translated version of [argilla/ultrafeedback-binarized-preferences-cleaned](https://huggingface.co/datasets/argilla/ultrafeedback-binarized-preferences-cleaned)
|
61 |
+
- No other dataset was used
|