使用的是windows(22h2)
略
如果是升级:wsl --update
在windows中安装nvidia最新驱动,过程略
wsl --install Ubuntu-20.04
sudo cp /etc/apt/sources.list /etc/apt/sources.list.bak \
&& sudo sed -i 's/archive.ubuntu/mirrors.aliyun/; s/security.ubuntu/mirrors.aliyun/' /etc/apt/sources.list \
&& sudo apt update \
&& sudo apt install -y gcc
参考:https://learn.microsoft.com/en-us/windows/ai/directml/gpu-tensorflow-wsl
# 安装conda
wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh \
&& bash Miniconda3-latest-Linux-x86_64.sh
退出命令行,再次进入
# 创建虚拟环境,并安装tensorflow-directml
conda create --name directml python=3.6 \
&& conda activate directml \
&& pip install tensorflow-directml
keras包含在tensorflow中,参考https://keras.io/getting_started/
参考:https://medium.com/@xizengmao/install-tensorflow-with-gpu-acceleration-simultaneously-for-windows-and-wsl-linux-2-10da088d5e4f
使用cuda11.2.0 + cudnn8.1.1,勉强可以,最新版本测试不通过
从https://developer.nvidia.com/cuda-downloads下载WSL-Ubuntu的版本,安装命令看网页
增加环境变量:PATH和LD_LIBRARY_PATH
从https://developer.nvidia.com/rdp/cudnn-download下载linux x86_64的版本,安装命令看网页
下载tar版本,解压:
sudo cp cudnn-*-archive/include/cudnn*.h /usr/local/cuda/include
sudo cp -P cudnn-*-archive/lib/libcudnn* /usr/local/cuda/lib64
sudo chmod a+r /usr/local/cuda/include/cudnn*.h /usr/local/cuda/lib64/libcudnn*
sudo apt install -y python3 python-is-python3 python3-dev python3-pip python3-numpy
sudo pip install tensorflow
python -c "import tensorflow as tf; print(tf.config.list_physical_devices('GPU'))"