Llama 3微调个人小助手认知(XTuner版)

Llama3-Tutorial/docs/llava.md at main · SmartFlowAI/Llama3-Tutorial (github.com)
参考机智流官方教程

先激活环境，这些都是之前的学习内容

conda activate llama3

cd ~
git clone https://github.com/SmartFlowAI/Llama3-Tutorial

cd XTuner
pip install -e .

自我认知训练数据集准备

把机智流的英文改为中文

conda activate llama3

python tools/gdata.py 

激活环境，运行脚本

cd ~/Llama3-Tutorial

xtuner train configs/assistant/llama3_8b_instruct_qlora_assistant.py --work-dir /root/llama3_pth

xtuner convert pth_to_hf /root/llama3_pth/llama3_8b_instruct_qlora_assistant.py \
  /root/llama3_pth/iter_500.pth \
  /root/llama3_hf_adapter

# Adapter PTH 转 HF 格式

xtuner convert pth_to_hf /root/llama3_pth/llama3_8b_instruct_qlora_assistant.py \
  /root/llama3_pth/iter_500.pth \
  /root/llama3_hf_adapter

# 模型合并

export MKL_SERVICE_FORCE_INTEL=1
xtuner convert merge /root/model/Meta-Llama-3-8B-Instruct \
  /root/llama3_hf_adapter\
  /root/llama3_hf_merged

推理验证

streamlit run ~/Llama3-Tutorial/tools/internstudio_web_demo.py \

  /root/llama3_hf_merged

微调完只会说一句话了【捂脸】