|
--- |
|
license: apache-2.0 |
|
datasets: |
|
- jojo0217/korean_rlhf_dataset |
|
language: |
|
- ko |
|
pipeline_tag: text-generation |
|
--- |
|
|
|
์ฑ๊ท ๊ด๋ํ๊ต ์ฐํํ๋ ฅ ๋ฐ์ดํฐ๋ก ๋ง๋ ํ
์คํธ ๋ชจ๋ธ์
๋๋ค. |
|
๊ธฐ์กด 10๋ง 7์ฒ๊ฐ์ ๋ฐ์ดํฐ + 2์ฒ๊ฐ ์ผ์๋ํ ์ถ๊ฐ ๋ฐ์ดํฐ๋ฅผ ์ฒจ๊ฐํ์ฌ ํ์ตํ์์ต๋๋ค. |
|
___ |
|
๋ชจ๋ธ์ EleutherAI/polyglot-ko-5.8b๋ฅผ base๋ก ํ์ต ๋์์ผ๋ฉฐ |
|
ํ์ต parameter์ ๋ค์๊ณผ ๊ฐ์ต๋๋ค. |
|
|
|
batch_size: 128 |
|
micro_batch_size: 8 |
|
num_epochs: 3 |
|
learning_rate: 3e-4 |
|
cutoff_len: 1024 |
|
lora_r: 8 |
|
lora_alpha: 16 |
|
lora_dropout: 0.05 |
|
weight_decay: 0.1 |
|
___ |
|
|
|
์ธก์ ํ kobest 10shot ์ ์๋ ๋ค์๊ณผ ๊ฐ์ต๋๋ค. |
|
![score](./asset/score.png) |
|
___ |
|
๋ชจ๋ธ prompt template๋ kullm์ template๋ฅผ ์ฌ์ฉํ์์ต๋๋ค. |
|
ํ
์คํธ ์ฝ๋๋ ๋ค์๊ณผ ๊ฐ์ต๋๋ค. |
|
https://colab.research.google.com/drive/1xEHewqHnG4p3O24AuqqueMoXq1E3AlT0?usp=sharing |
|
``` |
|
from transformers import pipeline, AutoModelForCausalLM, AutoTokenizer |
|
|
|
model_name="jojo0217/ChatSKKU5.8B" |
|
model = AutoModelForCausalLM.from_pretrained( |
|
model_name, |
|
device_map="auto", |
|
load_in_8bit=True,#๋ง์ฝ ์์ํ ๋๊ณ ์ถ๋ค๋ฉด false |
|
) |
|
tokenizer = AutoTokenizer.from_pretrained(model_name) |
|
pipe = pipeline( |
|
"text-generation", |
|
model=model, |
|
tokenizer=model_name, |
|
device_map="auto" |
|
) |
|
|
|
def answer(message): |
|
prompt=f"์๋๋ ์์
์ ์ค๋ช
ํ๋ ๋ช
๋ น์ด์
๋๋ค. ์์ฒญ์ ์ ์ ํ ์๋ฃํ๋ ์๋ต์ ์์ฑํ์ธ์.\n\n### ๋ช
๋ น์ด:\n{message}" |
|
ans = pipe( |
|
prompt + "\n\n### ์๋ต:", |
|
do_sample=True, |
|
max_new_tokens=512, |
|
temperature=0.7, |
|
repetition_penalty = 1.0, |
|
return_full_text=False, |
|
eos_token_id=2, |
|
) |
|
msg = ans[0]["generated_text"] |
|
return msg |
|
answer('์ฑ๊ท ๊ด๋ํ๊ต์๋ํด ์๋ ค์ค') |
|
``` |