Views
No views yet
| 항목 | 내용 |
|---|---|
| 베이스 모델 | Qwen/Qwen2.5-1.5B-Instruct |
| 학습 방법 | QLoRA (4bit, HuggingFace + bitsandbytes + PEFT) |
| LoRA r | 16 |
| LoRA alpha | 16 |
| 학습 환경 | Google Colab (T4 GPU) |
1import torch
2from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
3from peft import PeftModel
4
5model_name = "Qwen/Qwen2.5-1.5B-Instruct"
6
7bnb_config = BitsAndBytesConfig(
8 load_in_4bit=True,
9 bnb_4bit_quant_type="nf4",
10 bnb_4bit_compute_dtype=torch.float16,
11 bnb_4bit_use_double_quant=True,
12)
13
14base_model = AutoModelForCausalLM.from_pretrained(
15 model_name,
16 quantization_config=bnb_config,
17 device_map="auto",
18)
19tokenizer = AutoTokenizer.from_pretrained(model_name)
20
21model = PeftModel.from_pretrained(base_model, "YOUR_REPO_ID")