SPACEc CPU

    # Create conda environment
    conda create -n spacec python==3.10 graphviz libvips openslide
    conda activate spacec

    # Install spacec
    pip install spacec

Continue the following steps only if you have GPU(s)

SPACEc GPU

    # Install CUDA
    conda install conda-forge::cudatoolkit=11.2.2 cudnn=8.1.0.77 -y
    # Set environment variables for Tensorflow to find CUDA libraries
    mkdir -p $CONDA_PREFIX/etc/conda/activate.d && \
        echo 'export LD_LIBRARY_PATH=$CONDA_PREFIX/lib:$LD_LIBRARY_PATH' \
        > $CONDA_PREFIX/etc/conda/activate.d/env_vars.sh

1. OPTIONAL: For GPU-accelerated clustering via RAPIDS run the following. Note that only RTX20XX or better GPUs are supported. More information on rapids-singlecell are available here: https://rapids-singlecell.readthedocs.io/en/latest/Installation.html

    pip install spacec[rapids] --extra-index-url=https://pypi.nvidia.com

2. OPTIONAL: To use STELLAR run the following:

    pip install spacec[stellar] \
        --extra-index-url https://download.pytorch.org/whl/cu113 \
        -f https://data.pyg.org/whl/torch-1.12.0+cu113.html

3. Test if SPACEc loads and if your GPU is visible:

   python -c "import spacec as sp; sp.hf.check_for_gpu()"
   

SPACEc CPU:

    # Create conda environment
    conda create -y -n spacec python==3.10 graphviz libvips openslide
    conda activate spacec

    # Install spacec
    pip install spacec

    # Install remaining requirements for deepcell
    # NOTE: Ignore the error about pip's dependency resolver
    pip install -r https://raw.githubusercontent.com/nolanlab/SPACEc/master/requirements/requirements-deepcell-mac-arm64_tf210-metal.txt
    pip install deepcell --no-deps

SPACEc GPU: Mac GPU support is currently only supported for Tensorflow based methods but not PyTorch (in some cases we try to use the MPS backend if possible but that can be tricky). We recommend to use a Linux system for full GPU acceleration.

Although SPACEc can run directly on Windows systems, we highly recommend running it in WSL. If you are unfamiliar with WSL, you can find more information on how to use and install it here: https://learn.microsoft.com/en-us/windows/wsl/install If you decide to use WSL, follow the Linux instructions.

If you plan to continue with the native Windows environment

1. One of the segmentation tools within SPACEc neeeds a C++ compiler. If your environment doesn't have it already, the easiest way is to:

   1. Download the community version of Visual Studio from the official Microsoft website: https://visualstudio.microsoft.com. After installing the software on your system, select the following options to install the components needed for C++ development (see screenshots)

      

      

   2. In the meantime, you can already install libvips (https://www.libvips.org/) by downloading the pre-compiled Windows binaries from this repository: https://github.com/libvips/build-win64-mxe/releases/tag/v8.16.0 and adding them to your PATH. If you are unsure about which version to choose, vips-dev-w64-all-8.16.0.zip should work for you.

   3. Unpack the zip file and add the directory to your PATH environment. If you don’t know how to do that, consider watching this tutorial video that explains the process: https://www.youtube.com/watch?v=O5iBsdAd1_w

SPACEc CPU:

    # Create conda environment
    conda create -n spacec python==3.10
    conda activate spacec

    # Install dependencies via conda.
    conda install -c conda-forge graphviz

    # Install spacec
    pip install spacec

SPACEc GPU:

    conda install conda-forge::cudatoolkit=11.2.2 cudnn=8.1.0.77 -y

    mkdir %CONDA_PREFIX%\etc\conda\activate.d && (
    echo @echo off > %CONDA_PREFIX%\etc\conda\activate.d\env_vars.bat
    echo set PATH=%CONDA_PREFIX%\bin;%PATH% >> %CONDA_PREFIX%\etc\conda\activate.d\env_vars.bat
    echo set LD_LIBRARY_PATH=%CONDA_PREFIX%\lib;%LD_LIBRARY_PATH% >> %CONDA_PREFIX%\etc\conda\activate.d\env_vars.bat
    )

    # If Pytorch does not find the GPU try:
    # pip install torch==1.12.0+cu113 torchvision==0.13.0+cu113 torchaudio==0.12.0 --extra-index-url https://download.pytorch.org/whl/cu113

Reinstall SPACEc to be compatible with the GPU setting

    # Install spacec
    pip install spacec

Test if SPACEc loads and if your GPU is visible if you installed the GPU version.

    import spacec as sp
    sp.hf.check_for_gpu()

# Run CPU version:
docker build -f ../Docker/spacec_cpu_build.dockerfile -t spacec:cpu .
docker run -p 8888:8888 -p 5100:5100 spacec:cpu

# If running an amd64 image on apple silicon, use the following command:
docker run --platform linux/amd64 -p 8888:8888 -p 5100:5100 spacec:cpu

# Or run GPU version:
docker build -f ../Docker/spacec_gpu_build.dockerfile -t spacec:gpu .
docker run --gpus all -p 8888:8888 -p 5100:5100 spacec:gpu