RichardErkhov/AIdenU_-_Mistral-7B-v0.2-ko-Y24_v1.0-gguf

from transforemrs import AutoTokenizer, AutoModelForCausalLM

model = AutoModelForCausalLM.from_pretrained("AIdenU/Mistral-7B-v0.2-ko-Y24_v1.0", device_map="auto", torch_dtype=torch.bfloat16)
tokenizer = AutoTokenizer.from_pretrained("AIdenU/Mistral-7B-v0.2-ko-Y24_v1.0", use_fast=True)

prompt = [
  {'role': 'system', 'content': '당신은 지시를 매우 잘 따르는 인공지능 비서입니다.'},
  {'role': 'user', 'content': '지렁이도 밟으면 꿈틀하나요?'}
]
outputs = model.generate(
  **tokenizer(
    tokenizer.apply_chat_template(prompt, tokenize=False, add_generation_prompt=True),
    return_tensors='pt'
  ).to('cuda'),
  max_new_tokens=256,
  temperature=0.2,
  top_p=1,
  do_sample=True
)
print(tokenizer.decode(outputs[0]))