Packages ccpi-regulariser tigre cil tomophantom not found during installation.

[jcesardasilva]

Hi everyone,

I am trying to install CIL and want to go through the CIL-demos, but when I tried to install the packages according to the installation instruction provided here, I got the PackageNotFoundError below. Could you please help me with this? Thank you.

conda install -c conda-forge -c intel -c astra-toolbox -c ccpi cil=22.1.0 astra-toolbox tigre ccpi-regulariser tomophantom
Collecting package metadata (current_repodata.json): done
Solving environment: failed with initial frozen solve. Retrying with flexible solve.
Collecting package metadata (repodata.json): done
Solving environment: failed with initial frozen solve. Retrying with flexible solve.

PackagesNotFoundError: The following packages are not available from current channels:

  - ccpi-regulariser
  - tigre
  - cil=22.1.0
  - tomophantom

Current channels:

  - https://conda.anaconda.org/conda-forge/osx-64
  - https://conda.anaconda.org/conda-forge/noarch
  - https://conda.anaconda.org/intel/osx-64
  - https://conda.anaconda.org/intel/noarch
  - https://conda.anaconda.org/astra-toolbox/osx-64
  - https://conda.anaconda.org/astra-toolbox/noarch
  - https://conda.anaconda.org/ccpi/osx-64
  - https://conda.anaconda.org/ccpi/noarch
  - https://repo.anaconda.com/pkgs/main/osx-64
  - https://repo.anaconda.com/pkgs/main/noarch
  - https://repo.anaconda.com/pkgs/r/osx-64
  - https://repo.anaconda.com/pkgs/r/noarch

To search for alternate channels that may provide the conda package you're
looking for, navigate to

    https://anaconda.org

and use the search bar at the top of the page.

[epapoutsellis]

Hi @jcesardasilva,

Do you have macOS? Unfortunately, atm we do not support macOS. Do you have access to linux?

[paskino]

There is a CIL OSX package, version 21.4.0, which is 1 year old.

I think you should be able to install that one to start with.

conda create --name cil -c conda-forge -c intel -c ccpi cil

We can make a newer release of CIL, but I'm not sure we will be able to create binaries for the dependencies with CUDA.

[jcesardasilva]

Thank you @paskino and @epapoutsellis.

I understand. I think the new processors M1 and M2 are complicated, but for the Intel-based, I have seen that the complications are minimum for MACOS, since it has similarities with the LINUX core. However, I perfectly understand our time is limited. Is there any way to compile it from the source with CMAKE and configure?

To be honest, this is the main reason why I hate conda (very personnal view), because you need binaries. I am old school and create venv with virtualenv and organise the libraries myself. I like to install things with pip when possible, otherwise with setuptools to install libraries with python setup.py install, but your library does not provide any other way than conda installation.

Could you please provide any other way to install your library? In such a way, the users of MACOS, which is not a few, will not feel excluded.
I can manage the CUDA installation from my side, but I am constantly juggling between Linux and MacOs, but the LINUX is in the HPC cluster while the MACOS is used for the prototyping and debug of the code.

@paskino : I did install the way you recommended (thanks), but I miss the plugins. I can easily install ASTRA, but not TOMOPHANTOM or CCPI-Regularisation. In fact, I have problems with the CCPI packages only.

[jcesardasilva]

For info, I have cil and astra installed, but I cannot use the plugins:

[epapoutsellis]

Hi @jcesardasilva,

Being a mac user myself and CIL developer, I totally agree with all of your points.

In the past, I have created some instructions to install CIL+Google colab as an alternative with an easy access to our mac users. Unfortunately, it's not very practical and easy to break. Pip installation will help a lot within Google Colab since this is the default option and we do not have to install conda/mamba. However, this means that we need to fix also Tomophantom and CCPi-regularisation toolkit.

In my opinion, I would try to install everything = CIL + other libraries on a MAC through docker. However, I am not sure how difficult will be with the new processors ( I have not tried it) on mac.

[epapoutsellis]

I just checked the CIL-colab

As it is now the mamba installation will fail because of this error

numpy.ndarray size changed, may indicate binary incompatibility. Expected 96 from C header, got 88 from PyObject

from CCPi-regularisation toolkit.

If you force numpy=1.20.3

!mamba install -c conda-forge -c intel -c astra-toolbox -c ccpi cil=22.0.0 numpy==1.20.3 astra-toolbox ccpi-regulariser tomophantom --quiet

the demo should work.

[paskino]

Hi @jcesardasilva , we provide conda binaries for user's ease.

You can build and install with CMake directly, and there are instructions here https://github.com/TomographicImaging/CIL#building-cil-from-source-code

Let us know if they are complete as we think.

Notice that some things depend on Intel IPP library which I am afraid does only support Intel 64bit architecture, so no Apple M1 or M2.

For info, I have cil and astra installed, but I cannot use the plugins:

This is because you installed version 21.4.0. To access the ASTRA plugin you need to install cil-astra, which hopefully is available on our channel for mac or, you can install it from source. In this case I suppose python setup.py install should do, but I cannot guarantee.

[paskino]

I just checked the CIL-colab

As it is now the mamba installation will fail because of this error

numpy.ndarray size changed, may indicate binary incompatibility. Expected 96 from C header, got 88 from PyObject

from CCPi-regularisation toolkit.

If you force numpy=1.20.3

!mamba install -c conda-forge -c intel -c astra-toolbox -c ccpi cil=22.0.0 numpy==1.20.3 astra-toolbox ccpi-regulariser tomophantom --quiet

the demo should work.

@epapoutsellis can you move this to the CIL-Demos repo's issues, please?