NaiLong Team
运行方法:
git clone --depth 1 https://github.com/hiyouga/LLaMA-Factory.git
cd LLaMA-Factory
pip install -e ".[torch,metrics]" --no-build-isolation- 下载到本地
git lfs install
git clone https://huggingface.co/Qwen/Qwen3-8B- 直接使用模型仓库中的模型或者在线加载
默认目录在detect下
cp ./data/final_sft_data.json ./LLaMA-Factory/data
cp ./config/put_data.py ./LLaMA-Factory/data
cd ./LLaMA-Factory/data
python put_data.py需要修改.yaml配置文件中的model_path
cp ./config/qwen3_detect.yaml ./LLaMA-Factory/examples/train_full默认使用4卡A100 80G
cd ./LLaMA-Factory
FORCE_TORCHRUN=1 llamafactory-cli train examples/train_full/qwen3_detect.yamlcd ./predict
#更改predict中的model_path
python predict.py