Running immich on Cuda (remote host with WSL2) #

Install Cuda on WSL by following https://developer.nvidia.com/cuda-downloads?target_os=Linux&target_arch=x86_64&Distribution=WSL-Ubuntu&target_version=2.0&target_type=deb_local and if that doesn’t work, then: https://docs.nvidia.com/cuda/wsl-user-guide/index.html#cuda-support-for-wsl-2

On the host, go to docker engine configuration and add:

"runtimes":{
     "nvidia":{
         "path":"/usr/bin/nvidia-container-runtime",
         "runtimeArgs":[

         ]
     }
 }

Now you should be able to run the following examples:
1. sudo docker run --rm --runtime=nvidia --gpus all ubuntu nvidia-smi
2. sudo docker run --rm --gpus all nvcr.io/nvidia/k8s/cuda-sample:nbody nbody -gpu -benchmark
3. sudo docker run --rm --gpus all nvcr.io/nvidia/k8s/cuda-sample:nbody nbody -gpu -benchmark -numbodies=1000000
Run ./machine-learning-docker.sh on wsl2