installation instructions insufficient

[bobfRT1]

I don't manage to install this. After fiddling for 2h and installing Visual Studio and build tools it still throws this error during "pip install napari-cellseg3d":

....
preparing metadata (pyproject.toml) did not run successfully.
...
Build type: native build
Activating VS 17.14.7
Project name: python-siphash24
Project version: undefined
C compiler for the host machine: cl (msvc 19.44.35211 "Microsoft (R) C/C++ Optimizing Compiler Version 19.44.35211 for x64")
C linker for the host machine: link link 14.44.35211.0
Cython compiler for the host machine: cython (cython 3.1.2)
Host machine cpu family: x86_64
Host machine cpu: x86_64
Did not find pkg-config by name 'pkg-config'
Found pkg-config: NO
Found CMake: C:\Program Files\Microsoft Visual Studio\2022\Community\Common7\IDE\CommonExtensions\Microsoft\CMake\CMake\bin\cmake.EXE (3.31.6)
Run-time dependency libcsiphash-1 found: NO (tried pkgconfig and cmake)
Looking for a fallback subproject for the dependency libcsiphash-1
Building fallback subproject with default_library=static
ERROR: Subproject libcsiphash-1 is buildable: NO

  ..\meson.build:14:14: ERROR: Git program not found, cannot download libcsiphash-1.wrap via git.

[C-Achard]

Hello @bobfRT1 , sorry for the trouble !

Based on the log you shared, it seems you still need to install git, so that pip can install and build the C library.

You can install git from here.

---

A small checklist to help with the next steps :

• A package manager such as conda or miniforge is highly recommended; create a new environment, activate it and install napari-cellseg3d there
  • Make sure to specify the right python version when creating a new env, e.g. conda create -n cellseg3d-env python=3.9
  • pip install --upgrade pip setuptools wheel, should not be necessary if you just created the env, but if you already have one this can avoid quite a few issues
• Install the pre-requisites once you have your python env up and running, see here
  • (If you have a CUDA-capable GPU, pick the right PyTorch version for your hardware and drivers, let me know if you are unfamiliar with CUDA versions)
• pip install napari[all]
  • Launch napari from you terminal with the conda env to check if everything is in place
• And then install the plugin !

I know this can take some time, especially if you happen to be new to Python, so please let me know if you have any trouble and I'll try to help :)

Finally, if you have a bit of time to detail what took you too long/was unclear or misleading in the docs down below, that would great so I can update the docs and make things smoother for everyone.

Let me know if I can do anything else !

Best,
Cyril

[bobfRT1]

Thanks!

So I murdered the environment for the third time and gave this a try:

conda create -n cellseg3d python=3.9
conda activate cellseg3d
pip install --upgrade pip setuptools wheel
pip install napari[all]
pip3 install torch torchvision --index-url https://download.pytorch.org/whl/cu128
napari (is happy)

pip install napari-cellseg3d

Now it seems to work. Not sure what is going on here:

Attempting uninstall: torch
Found existing installation: torch 2.7.1+cu128
Uninstalling torch-2.7.1+cu128:
Successfully uninstalled torch-2.7.1+cu128
ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
torchvision 0.22.1+cu128 requires torch==2.7.1+cu128, but you have torch 2.6.0 which is incompatible.

Why uninstall 2.7.1 and then complain about it?

[C-Achard]

Ah, glad to hear ! Let me know if you have any further issues with the plugin.

As for the install error, it seems that the newest versions of MONAI have a pinned version of torch.
Historically it was best to install MONAI after torch to get the GPU support to work as intended, hence why our docs mention it, but it seems they have automated the torch install a bit more.

I'd recommend trying the env as is, I was able to reproduce this but the plugin worked as intended despite it; if you run into issues with torchvision/torchaudio I'd recommend downgrading those to a torch 2.6-compatible version; feel free to let me know if this fails.

Since it seems the install worked for now, I will close this, but please reopen or create a new issue if you have other problems !

Best,
Cyril