Windows

Windows#

Although the libraries (backends) can be installed on Windows natively, the HTTomo framework requires Linux. For Windows 10 and newer one can install Linux through WSL and run HTTomo there. WSL supports also CUDA-compatible GPU devices, so the GPU methods will also be working smoothly.

Installation steps#

Steps 1-3 are following the official WSL installation provided here.

  1. Open Terminal as Admin: Right-click Start, select “Windows PowerShell (Admin)” or “Terminal (Admin)”.

  2. Run Install Command: Type wsl --install and hit Enter. Reboot PC.

  3. Open Promt and type wsl to start WSL.

Troubleshooting: No Internet connection in WSL.

Perform the steps as described in here.

  1. Get the latest Miniforge for Linux using this link.

  2. Get into the base environment of conda by source /path/to/.bashrc.

  3. Install HTTomo and dependencies by following the Installation Guide notes for Linux.

  4. Optional step. Run HTTomo tests to make sure that everything works correctly.

Troubleshooting: HTTomoLib library requires a gcc compiler.

Install the gcc compiler with sudo apt-get install gcc.

Troubleshooting: HTTomo doesn’t run on a GPU with an older architecture.

If there is a GPU with an older GPU architecture then try:

  1. conda install -c conda-forge cupy==12.3.6 openmpi==4.1.6 h5py[build=*openmpi*] python>=3.10 numpy astra-toolbox aiofiles click graypy loguru nvtx pillow pyyaml scikit-image scipy tqdm hdf5plugin pip pywavelets

  2. pip install tomobar httomolib httomolibgpu httomo-backends --no deps

  3. Go to numpy1 branch on the cloned HTTomo repository and run pip install . --no-deps to install a compatible older version.