Skip to content

Latest commit

 

History

History
216 lines (147 loc) · 7.97 KB

python_installation.md

File metadata and controls

216 lines (147 loc) · 7.97 KB

Installation Instructions for Python

Table of contents

Windows

Requirements

  1. Python 3.7-3.11
  2. NVIDIA CUDA-capable GPU with compute capability >=3.5
  3. MSVC (for building)
  4. CUDA Toolkit (for building)

Tested on:

Software Version
Windows 10, 11
Python 3.10-3.14
CUDA >=9.2
MSVC 19.24

Build Instructions

We strongly recommend using conda environments and doing the install in one specific for tigre.

Tip

Quick Summary:

  1. Install Python and pip, MSVC and CUDA
  2. run git clone https://github.com/CERN/TIGRE.git
  3. run pip install . in the root folder.

A successful installation should be able to execute the script at TIGRE/Python/example.py

  1. Install MS Visual Studio Build Tools with Windows SDK.
    NOTE: The User has to have no spaces.

  2. Install Python 3 + pip. You can use a virtual conda environment or just a normal python installation. We strongly recommend the conda environments. Activate the environment before step 5.

  3. Install CUDA. Make sure the CUDA_PATH and PATH environment variable are set accordingly.
    NOTE: The User has to have no spaces.

  4. Download/clone TIGRE

    git clone https://github.com/CERN/TIGRE.git

  5. Compile libraries

    cd TIGRE/  
pip install . --user

    NOTE: If you are working under the virtual environment that created by venv or a conda environment and you want to install TIGRE to it, you should remove the --user option. With the --user option, TIGRE and the other required packages will be installed to your Python user install directory, not to your virtual environment or system directory.

  6. Try demo 3. If it runs successfully then you are good to go.

Conda install

An unofficial conda package is built & maintained by CCPi Tomographic Imaging - you can install it by running:

conda install -c ccpi tigre

Please report conda issues upstream.

Linux

Requirements

We strongly recommend using conda environments and doing the install in one specific for tigre.

  1. Python 3
  2. NVIDIA CUDA-capable GPU with compute capability >=3.5
  3. gcc (for building)
  4. CUDA Toolkit (for building)

Tested on:

Software Version
Ubuntu >=16.04
Python 3.10-3.14
CUDA >=9.2
gcc 7.6.0

Build Instructions

We strongly recommend using conda environments and doing the install in one specific for tigre.

Tip

Quick Summary:

  1. Install python and pip, gcc and CUDA
  2. run git clone https://github.com/CERN/TIGRE.git
  3. run pip install . in the root folder.

A successful installation should be able to execute the script at TIGRE/Python/example.py

For Ubuntu

  1. Install python and pip

    Recommended to do it via Anaconda3, and not the following.

    sudo apt update
sudo apt upgrade
sudo apt install python3.10 python-pip

  2. Install CUDA

    Installing CUDA in linux (specially one with a GUI) can be a challenge. Please follow NVIDIAs instructions carefully.
    CUDA download link

  3. Install gcc

    gcc should already be installed in your linux, as it is part of the linux distribution.
    If you need to install an older version of gcc, read here.

  4. Download/clone TIGRE

    git clone https://github.com/CERN/TIGRE.git

  5. Compile libraries

    cd TIGRE/  
pip install . --user

    NOTE: If you are working under the virtual environment that created by venv and you want to install TIGRE to it, you should remove the --user option. With the --user option, TIGRE and the other required packages will be installed to your Python user install directory, not to your virtual environment or system directory.

  6. Try demo 3. If it runs successfully then you are good to go.

if this fails, then try:

export CUDAHOME=yourcudahome, e.g. default is export CUDAHOME=/usr/local/cuda pip install . --user

NOTE: as of November 2020 the pip pytigre is behind the main repo, we recommend you install it and compile it yourself. Trying to fix that.

Conda install

An unofficial conda package is built & maintained by CCPi Tomographic Imaging - you can install it by running:

conda install -c ccpi tigre

Please report conda issues upstream.

Optional Code Style Enforcement

Optional linting dependencies are provided to enforce the prevailing codestyle in the Python component of the TIGRE library.

The primary linting packages utilized are:

  • black for automatic code formatting
  • flake8 for code style enforcement

Multiple installation options have been provided:

pip Installation

A dev dependency group is included in the pyproject.toml file. Use pip install --group dev . to install those dependencies.

pre-commit

Linting dependencies are also provided as a pre-commit configuration. With the pre-commit hooks installed, code will be linted prior to commit, preventing code from being committed that fails linting. By default, these hooks are only run against staged files.

The pre-commit package is included in the above pip installs; once the framework is installed, hooks can be installed by invoking pre-commit install.

Usage

Once installed, invoking the linting tools is straightforward. From TIGRE's Python directory:

# Run in current directory & subdirectories
$ flake8 .
$ black .

# Or specify a file
$ flake8 ./example.py
$ black ./example.py

pre-commit may also be manually invoked, which will run flake8 and black against all staged files:

$ pre-commit run
flake8...................................................................Passed
black....................................................................Passed
Check for merge conflicts................................................Passed
Check Toml...........................................(no files to check)Skipped
Check Yaml...........................................(no files to check)Skipped
Fix End of Files.........................................................Passed
Mixed line ending........................................................Passed
check blanket noqa.......................................................Passed

NOTE: pre-commit may also be manually invoked against all files (staged and unstaged) using the pre-commit run --all-files. However, some changes made to Python's TIGRE codebase by black have been manually reverted for readability reasons and should not be committed in their blackened state.

Advanced

If you are doing reconstruction of large datasets, and you want to use swap memory, you will need to deactivate TIGREs pinned memory feature at compile time. This will allow you to use swap memory, but it will make the operators in TIGRE slower, as pinned memory is used for simultaneous memory and compute.

You can do this by calling the setup.py with the flag --no_pinned_memory.