colab一键训练脚本 https://colab.research.google.com/drive/1MfP3vt9YrOkjg70dKPPFvB174PBnaPSB?usp=sharing
相关项目 unsloth:https://github.com/unslothai/unsloth gpt4all:https://gpt4all.io/ triton:https://github.com/openai/triton llama.cpp:https://github.com/ggerganov/llama.cpp
Windows本地部署条件 1、Windows10/Windows11 2、英伟达卡8G显存、16G内存,安装CUDA12.1、cuDNN8.9,C盘剩余空间20GB、unsloth安装盘S40GB 3、依赖软件:CUDA12.1+cuDNN8.9、Python11.9、Git、Visual Studio 2022、llvm(可选) 4、HuggingFace账号,上传训练数据集
二、安装unsloth 1、使用python11创建虚拟环境 python311\python.exe -m venv venv 2、激活虚拟环境 call venv\scripts\activate.bat 3、安装依赖包 pip install torch==2.2.2 --index-url https://download.pytorch.org/whl/cu121 pip install "unsloth[colab-new] @ git+https://github.com/unslothai/unsloth.git" pip install --no-deps trl peft accelerate bitsandbytes pip install deepspeed-0.13.1+unknown-py3-none-any.whl pip install triton-2.1.0-cp311-cp311-win_amd64.whl pip install xformers==0.0.25.post1 4、测试安装是否成功 nvcc --version python -m xformers.info python -m bitsandbytes 5、运行脚本 test-unlora.py 测试微调之前推理 fine-tuning.py 用数据集微调 若本地运行fine-tuning.py出错,出现gcc.exe无法编译,可以尝试下载llvm-windows-x64.zip解压,在系统环境变量path路径里添加llvm下的bin路径