Skip to content

Latest commit

 

History

History
24 lines (21 loc) · 1.21 KB

Install-Tensorflow-on-WSL.md

File metadata and controls

24 lines (21 loc) · 1.21 KB
Raw

Insatall Tensorflow 2.12, CUDA 12 on WSL

Install CUDA Toolkit and CuDNN

  1. Make sure nvidia-smi worked after installing the NVIDIA GPU driver
  2. conda install -c nvidia cuda-toolkit See https://anaconda.org/nvidia/cuda-toolkit
  3. pip install nvidia-cudnn-cu12 See https://pypi.org/project/nvidia-cudnn-cu12/

Configure system paths at terminal

  1. mkdir -p $CONDA_PREFIX/etc/conda/activate.d
  2. echo 'CUDNN_PATH=$(dirname $(python -c "import nvidia.cudnn;print(nvidia.cudnn.__file__)"))' >> $CONDA_PREFIX/etc/conda/activate.d/env_vars.sh
  3. echo 'export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$CONDA_PREFIX/lib/:$CUDNN_PATH/lib' >> $CONDA_PREFIX/etc/conda/activate.d/env_vars.sh
  4. source $CONDA_PREFIX/etc/conda/activate.d/env_vars.sh

Install Tensorflow

  1. pip install --upgrade pip
  2. pip install tensorflow==2.12.*

Test

import tensorflow as tf
print(tf.__version__)
print(tf.test.is_built_with_cuda()) # check if CUDA is used
print(tf.config.list_physical_devices('GPU')) # show GPU numbers
print(tf.test.is_gpu_available(cuda_only=False, min_cuda_compute_capability=None)) # check if GPU is available