# colab の場合
!git clone --branch Mistral https://github.com/LaaZa/AutoGPTQ.git
!cp "/content/AutoGPTQ/auto_gptq/modeling/__init__.py" "/usr/local/lib/python3.10/dist-packages/auto_gptq/modeling/__init__.py"
!cp "/content/AutoGPTQ/auto_gptq/modeling/auto.py" "/usr/local/lib/python3.10/dist-packages/auto_gptq/modeling/auto.py"
!cp "/content/AutoGPTQ/auto_gptq/modeling/_const.py" "/usr/local/lib/python3.10/dist-packages/auto_gptq/modeling/_const.py"
!cp "/content/AutoGPTQ/auto_gptq/modeling/mistral.py" "/usr/local/lib/python3.10/dist-packages/auto_gptq/modeling/mistral.py"

python

from auto_gptq import AutoGPTQForCausalLM, BaseQuantizeConfig
from transformers import AutoTokenizer

model_name_or_path = "mmnga/japanese-stablelm-instruct-gamma-7b-GPTQ-calib-ja-1k"

# Tokenizer
tokenizer = AutoTokenizer.from_pretrained(model_name_or_path, trust_remote_code=True)

# Model
model = AutoGPTQForCausalLM.from_quantized(model_name_or_path, use_safetensors=True, device="cuda:0", use_auth_token=False)

#Your test prompt
prompt = """### 指示:今日の晩御飯のレシピを紹介して。 ### 応答:"""
print(tokenizer.decode(model.generate(**tokenizer(prompt, return_tensors="pt",add_special_tokens=False).to(model.device), max_new_tokens=100,do_sample=True,top_p=0.95,temperature=0.7)[0]))