NOTE: This page explains how to install Cellpose with GPU support in Miniconda = Prerequisites = # Get [[https://docs.conda.io/en/latest/miniconda.html | Miniconda]] # Get latest [[https://www.geforce.com/drivers|NVIDIA drivers]] # Get [[https://developer.nvidia.com/cuda-10.0-download-archive|CUDA 10.0 Toolkit]] # Get [[https://developer.nvidia.com/rdp/cudnn-download|cuDNN for CUDA 10.0]] (Needs you to login) # Get [[ https://software.intel.com/content/www/us/en/develop/tools/math-kernel-library/choose-download/windows.html | Intel® Math Kernel Library (Intel® MKL) for Windows]] Everything can be installed with defaults. The only tedious things are to setup CuDNN and some environment variables == CuDNN == CuDNN needs you to copy paste the contents of the zip file you downloaded to `C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0\` The idea is that the folder hierarchy matches. |{F13890259, width=400}|{F13890265, width=400} | |Copy this| into this| |-----------|-------| == Environment Variables Setup == Run `sysdm.cpl` then {nav Advanced>Environement Variables...} Create these two variables and add the following paths to them |VARIABLE NAME | Path (Default) | |-------------------|-----------------| |INCLUDE| C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0\include| |LIB| C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0\lib\x64| =Installing Cellpose= CellPose [[ http://www.cellpose.org/static/docs/installation.html | does a good job explaining how to install it ]] using CPU but is a little vague about installing it for GPU. After all prerequisites are met, you can install CellPose with GPU using the following environment file `conda env create -p d:\env-cellpose -f PATH_TO\environement.yml` NOTE: Feel free to replace `-p d:\env-cellpose` with `-n cellpose` but I prefer to install my environments elsewhere than the c:\ drive == Activating the GPU option == Provided you have installed all the prerequisites and used the `environment.yml` file above, you need to do the following inside your newly created environment ``` conda activate d:\env-cellpose pip uninstall mxnet-mkl pip install mxnet-cu100 mxnet-cu100mkl ``` ==Optional: Clone the cellpose repository == In this example, we are cloning it to `d:\cellpose-src` ``` git clone https://github.com/MouseLand/cellpose.git d:\cellpose-src ``` = Testing if this works = If you cloned the cellpose repository you can use the example notebook to see if everything worked ``` conda activate d:\env-cellpose d:\ cd cellpose-src jupyter lab ``` Navigate to the `notebooks` folder and try to run `run_cellpose.ipynb` Change the second cell to use the GPU ``` model = models.Cellpose(gpu=True, model_type='cyto') ``` Success~ {F13890608, size=full}