CUDA Libraries is not detected,then Setup WSL Environment,Report the following error.

[mack007liu]

Information: validate-env[o]
05:04:17.77 IsCudaRuntimeInstalled : False
Debug: validate-env[o]
05:04:17.77 ExecuteAsync Completed Elapsed:0:00:1
7700138
[2023-12-18T09:04:17.776Z] [INFO] Validation outcome prer
eqsmet: false
[2023-12-18T09:04:17.776Z] [INFO]
Initialize WebView pane
[2023-12-18T09:04:17.777Z]
[INFO] Navigating view to reqt
irements page
[2023-12-18T09:4:18.893Z] [INFO]
telemetry event:activat
e extension sent
[2023-12-18T99:04:18.893Z]
[INFO] telemetry event:require
ments needed sent
[2023-12-18T9:04:19.22Z]
[INFO] Navigation to loadPrere
quisiteCheck was successful: true
[2023-12-18T09:04:19.023Z]
[INFO]
Loading View: loadPrere
quisiteCheck
[2023-12-18T99:04:19.029Z] [INFO] Extenension: Invoking
alidateEnvironement for: conda
Debug: validate-env[o]
95:04:19.19 ExecuteAsync
Started

[mack007liu]

windows 11 sub system Ubuntu 22

[mack007liu]

[mack007liu]

[elsaco]

The plugin doesn't give any feedback why CUDA or Conda installation fails. In my case Conda doesn't install. So to prepare the environment I run the script directly inside the WSL instance. If anything fails there will be better feedback while running the scripts manually.

Look inside C:\Users\<username>\.vscode\extensions\ms-windows-ai-studio.windows-ai-studio-0.2.1-win32-x64\bin\Scripts and copy nvidia_cuda_install.sh to WSL and execute it. If the script runs okay next time you run the plugin it will detect CUDA as available.

nvidia_cuda_install.sh

#!/usr/bin/env bash

export DEBIAN_FRONTEND=noninteractive

set -e

# Clean up
rm -rf /var/lib/apt/lists/*

INSTALL_CUDNN=false
INSTALL_CUDNNDEV=false
INSTALL_NVTX=false
INSTALL_TOOLKIT=false
CUDA_VERSION=11.8
CUDNN_VERSION=8.6.0.163

if [ "$(id -u)" -ne 0 ]; then
    echo -e 'Script must be run as root. Use sudo, su, or add "USER root" to your Dockerfile before running this script.'
    exit 1
fi

apt_get_update()
{
    if [ "$(find /var/lib/apt/lists/* | wc -l)" = "0" ]; then
        echo "Running apt-get update..."
        apt-get update -y
    fi
}

# Checks if packages are installed and installs them if not
check_packages() {
    if ! dpkg -s "$@" > /dev/null 2>&1; then
        apt_get_update
        apt-get -y install --no-install-recommends "$@"
    fi
}

check_packages wget ca-certificates

# Add NVIDIA's package repository to apt so that we can download packages
# Always use the ubuntu2004 repo because the other repos (e.g., debian11) are missing packages
NVIDIA_REPO_URL="https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64"
KEYRING_PACKAGE="cuda-keyring_1.0-1_all.deb"
KEYRING_PACKAGE_URL="$NVIDIA_REPO_URL/$KEYRING_PACKAGE"
KEYRING_PACKAGE_PATH="$(mktemp -d)"
KEYRING_PACKAGE_FILE="$KEYRING_PACKAGE_PATH/$KEYRING_PACKAGE"
wget -O "$KEYRING_PACKAGE_FILE" "$KEYRING_PACKAGE_URL" 2>&1
apt-get install -yq "$KEYRING_PACKAGE_FILE"
apt-get update -yq

# Ensure that the requested version of CUDA is available
cuda_pkg="cuda-libraries-${CUDA_VERSION/./-}"
nvtx_pkg="cuda-nvtx-${CUDA_VERSION/./-}"
toolkit_pkg="cuda-toolkit-${CUDA_VERSION/./-}"
if ! apt-cache show "$cuda_pkg"; then
    echo "The requested version of CUDA is not available: CUDA $CUDA_VERSION"
    exit 1
fi

echo "Installing CUDA libraries..."
apt-get install -yq "$cuda_pkg"

if [ "$INSTALL_CUDNN" = "true" ]; then
    # Ensure that the requested version of cuDNN is available AND compatible
    cudnn_pkg_version="libcudnn8=${CUDNN_VERSION}-1+cuda${CUDA_VERSION}"
    if ! apt-cache show "$cudnn_pkg_version"; then
        echo "The requested version of cuDNN is not available: cuDNN $CUDNN_VERSION for CUDA $CUDA_VERSION"
        exit 1
    fi

    echo "Installing cuDNN libraries..."
    apt-get install -yq "$cudnn_pkg_version"
fi

if [ "$INSTALL_CUDNNDEV" = "true" ]; then
    # Ensure that the requested version of cuDNN development package is available AND compatible
    cudnn_dev_pkg_version="libcudnn8-dev=${CUDNN_VERSION}-1+cuda${CUDA_VERSION}"
    if ! apt-cache show "$cudnn_dev_pkg_version"; then
        echo "The requested version of cuDNN development package is not available: cuDNN $CUDNN_VERSION for CUDA $CUDA_VERSION"
        exit 1
    fi

    echo "Installing cuDNN dev libraries..."
    apt-get install -yq "$cudnn_dev_pkg_version"
fi

if [ "$INSTALL_NVTX" = "true" ]; then
    echo "Installing NVTX..."
    apt-get install -yq "$nvtx_pkg"
fi

if [ "$INSTALL_TOOLKIT" = "true" ]; then
    echo "Installing CUDA Toolkit..."
    apt-get install -yq "$toolkit_pkg"
fi

# Clean up
rm -rf /var/lib/apt/lists/*

echo "Done!"

[mack007liu]

Thank you, it can work now.

[alexgorbatchev]

scripts are in C:\Users\[user]\.vscode\extensions\ms-windows-ai-studio.windows-ai-studio-0.2.1-win32-x64\bin\Scripts as of 12/29/2023