cywellai/privacy-counsel-ko-8b의 LoRA 어댑터입니다.
머지된 풀 모델(16GB)이 필요하면 위 링크를 이용하세요.
1from transformers import AutoModelForCausalLM, AutoTokenizer
2from peft import PeftModel
3
4base_model_name = "Qwen/Qwen3-8B"
5adapter_name = "cywellai/privacy-counsel-ko-8b-lora"
6
7tokenizer = AutoTokenizer.from_pretrained(base_model_name, trust_remote_code=True)
8base_model = AutoModelForCausalLM.from_pretrained(
9 base_model_name,
10 torch_dtype="bfloat16",
11 device_map="auto",
12 trust_remote_code=True,
13)
14model = PeftModel.from_pretrained(base_model, adapter_name)
15
16messages = [
17 {"role": "user", "content": "개인정보 유출 시 대응 절차는?"},
18]
19
20text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
21inputs = tokenizer(text, return_tensors="pt").to(model.device)
22outputs = model.generate(**inputs, max_new_tokens=1500, temperature=0.5, do_sample=True)
23response = tokenizer.decode(outputs[0][inputs.input_ids.shape[1]:], skip_special_tokens=True)
24print(response)
