windows10在wsl上利用GPU运行tensorflow 2.12

从tensorflow 2.10开始，已经没有tensorflow-gpu在windows平台，合并为tensorflow，要想使用gpu只能wsl上运行。
大致方法是：安装WSL，安装miniconda，创建conda环境，安装tensorflow，安装pycharm专业版，调用wsl上的python解释器。

具体命令：

wsl --install
重启
菜单栏打开ubuntu，设置用户名，密码

curl https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh -o Miniconda3-latest-Linux-x86_64.sh
bash Miniconda3-latest-Linux-x86_64.sh

conda create --name tf2.12 python=3.9
conda activate tf2.12
conda install -c conda-forge cudatoolkit=11.8.0
pip config set global.extra-index-url https://pypi.mirrors.ustc.edu.cn/simple
pip install nvidia-cudnn-cu11==8.6.0.163

CUDNN_PATH=$(dirname $(python -c "import nvidia.cudnn;print(nvidia.cudnn.__file__)"))
export LD_LIBRARY_PATH=$CONDA_PREFIX/lib/:$CUDNN_PATH/lib:$LD_LIBRARY_PATH

pip install tensorflow==2.12
conda install -c nvidia cuda-nvcc

python3 -c "import tensorflow as tf; print(tf.config.list_physical_devices('GPU'))"

#cd /mnt/c/apps/PycharmProjects/
mkdir -p $CONDA_PREFIX/etc/conda/activate.d
echo 'export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$CONDA_PREFIX/lib/' > $CONDA_PREFIX/etc/conda/activate.d/env_vars.sh
#重要，不然pycharm 运行tensorflow会找不到gpu依赖库

报错：
Node: ‘StatefulPartitionedCall_4’
2 root error(s) found.
(0) INTERNAL: libdevice not found at ./libdevice.10.bc
[[{{node StatefulPartitionedCall_4}}]]
[[crf/cond/StatefulPartitionedCall/crf/cond/else/_236/crf/cond/concat/_328]]
(1) INTERNAL: libdevice not found at ./libdevice.10.bc
[[{{node StatefulPartitionedCall_4}}]]
0 successful operations.
0 derived errors ignored. [Op:__inference_train_function_73705]
解决办法：

cp miniconda3/envs/tf2.12/nvvm /mnt/d/pyproject/myrasa2/ -r
#复制虚拟环境下的nvvm目录到项目根目录