model-lora / README.md
Whitewinter's picture
Upload LoRA fine-tuned model
3ffe67c verified
|
Raw
History Blame Contribute Delete
1.24 kB
metadata
license: apache-2.0
base_model: Qwen/Qwen3-0.6B
tags:
  - peft
  - lora
  - fine-tuned
  - qwen
datasets:
  - tatsu-lab/alpaca
language:
  - ko
  - en

LoRA Fine-tuned Model

이 모델은 Qwen/Qwen3-0.6B을 기반으로 LoRA(Low-Rank Adaptation) 기법을 사용해 파인튜닝된 어댑터입니다.

모델 정보

  • 베이스 모델: Qwen/Qwen3-0.6B
  • 파인튜닝 방법: LoRA (Low-Rank Adaptation)
  • 데이터셋: tatsu-lab/alpaca

사용 방법

from transformers import AutoTokenizer, AutoModelForCausalLM
from peft import PeftModel

# 베이스 모델과 토크나이저 로드
tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen3-0.6B", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(
    "Qwen/Qwen3-0.6B",
    trust_remote_code=True,
    torch_dtype=torch.bfloat16,
    device_map="auto"
)

# LoRA 어댑터 로드
model = PeftModel.from_pretrained(model, "Whitewinter/model-lora")

# 추론
prompt = "### Instruction:\nExplain what machine learning is.\n\n### Response:\n"
inputs = tokenizer(prompt, return_tensors="pt")
outputs = model.generate(**inputs, max_new_tokens=200, temperature=0.7)
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(response)