AIdenU's picture
Update README.md
4f2312e verified
|
raw
history blame
865 Bytes
---
license: apache-2.0
language:
- ko
pipeline_tag: text-generation
tags:
- SOLAR
- SOLAR-10.7B
---
### Model Generation
```
from transforemrs import AutoTokenizer, AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("AIdenU/SOLAR-10b-ko-Y24_v0.1", device_map="auto")
tokenizer = AutoTokenizer.from_pretrained("AIdenU/SOLAR-10b-ko-Y24_v0.1", use_fast=True)
_prompt = [
{'role': 'system', 'content': '๋‹น์‹ ์€ ์ง€์‹œ๋ฅผ ๋งค์šฐ ์ž˜ ๋”ฐ๋ฅด๋Š” ์ธ๊ณต์ง€๋Šฅ ๋น„์„œ์ž…๋‹ˆ๋‹ค.'},
{'role': 'user', 'content': '์ง€๋ ์ด๋„ ๋ฐŸ์œผ๋ฉด ๊ฟˆํ‹€ํ•˜๋‚˜์š”?'}
]
prompt = tokenizer.apply_chat_template(_prompt, tokenize=False, add_generation_prompt=True)
outputs = model.generate(
**tokenizer(
prompt,
return_tensors='pt'
).to('cuda'),
max_new_tokens=256,
temperature=0.2,
top_p=1,
do_sample=True
)
print(tokenizer.decode(outputs[0]))
```