Build in a conda environment¶

Creating a dedicated Conda environment is a best practice that ensures the dependencies are managed effectively. This isolation prevents conflicts between packages and allows for a clean workspace. In the following section, we’ll guide you through the process of setting up a Conda environment and installing FastDDM to get your project up and running smoothly.

Install miniconda
Create an environment config YAML file
Notes on CUDA

Install miniconda¶

Download and install miniconda3 from the Anaconda website.

Create an environment config YAML file¶

Create a fastddm-env.yml file and write the following content in it (select your operating system).

name: fddm-env
channels:
  - defaults
  - conda-forge
dependencies:
  - gcc
  - gxx
  - python>=3.8
  - pip

Warning

If you install with CUDA support, ensure that your gcc and g++ versions are compatible with your CUDA Toolkit. Refer to the official CUDA documentation for the supported compiler versions.

When using a conda environment, you can specify the compatible versions in your fastddm-env.yml file. For example, for CUDA 12.9, the maximum supported GCC version is 14. Set both gcc=14 and gxx=14 in your environment file if needed.

Create the environment by running the following command in your terminal

$ conda env create -f fastddm-env.yml

Activate the environment

$ conda activate fddm-env

Export the environment variables

$ conda env config vars set CC=$CONDA_PREFIX/bin/gcc
$ conda env config vars set CXX=$CONDA_PREFIX/bin/g++

To compile the C++ core, also set the corresponding flag

$ conda env config vars set ENABLE_CPP=ON

To compile the CUDA core, set the corresponding flag

$ conda env config vars set ENABLE_CUDA=ON

Path to the CUDA Toolkit should also be exported

$ conda env config vars set CUDACXX=/usr/local/cuda_version/bin/nvcc

Deactivate and reactivate the environment to make the changes effective

$ conda deactivate
$ conda activate fddm-env

From the fastddm project root directory (see Building from source on how to get the source code), install the package and the test dependencies

$ pip3 install ."[test]"

Finally, run the tests from the project source directory

$ pytest -v

name: fddm-env
channels:
  - defaults
dependencies:
  - clang
  - clangxx
  - python>=3.8
  - pip

Create the environment by running the following command in your terminal

$ conda env create -f fastddm-env.yml

Activate the environment

$ conda activate fddm-env

Export the environment variables

$ conda env config vars set CC=$CONDA_PREFIX/bin/clang
$ conda env config vars set CXX=$CONDA_PREFIX/bin/clang++

To compile the C++ core, also set the corresponding flag

$ conda env config vars set ENABLE_CPP=ON

Deactivate and reactivate the environment to make the changes effective

$ conda deactivate
$ conda activate fddm-env

From the fastddm project root directory (see Building from source on how to get the source code), install the package and the test dependencies

$ pip3 install ."[test]"

Finally, run the tests from the project source directory

$ pytest -v

name: fddm-env
channels:
  - defaults
dependencies:
  - python>=3.8
  - pip

For Windows, you will still need to install Visual Studio Community Edition with the Desktop development with C++ option enabled (see Building from source).

Create the environment by running the following command in your miniconda PowerShell terminal

$ conda env create -f fastddm-env.yml

Activate the environment

$ conda activate fddm-env

To compile the C++ core, set the corresponding flag

$ conda env config vars set ENABLE_CPP=ON

Deactivate and reactivate the environment to make the changes effective

$ conda deactivate
$ conda activate fddm-env

From the fastddm project root directory (see Building from source on how to get the source code), install the package and the test dependencies

$ pip3 install ."[test]"

Finally, run the tests from the project source directory

$ pytest -v

Notes on CUDA¶

As of today, we could not find a way to automatically build the package from source using the cudatoolkit-dev distributed on conda-forge. We recommend following the instructions given in Building from source to install the package in the conda environment using the system CUDA Toolkit.

We welcome contributions on this matter!