Getting Tissue Forge

Pre-Built Binaries

Binary distributions of Tissue Forge are available via conda from the tissue-forge channel,

conda install -c conda-forge -c tissue-forge tissue-forge

Pre-built binaries of the latest Tissue Forge developments are also archived at the Tissue Forge Azure project. Installing pre-built binaries requires Miniconda. Binaries on Linux require the Mesa packages libgl1-mesa-dev and libegl1-mesa-dev. Packages include a convenience script install_env that installs the dependencies of the Tissue Forge installation on execution. After installing the dependencies environment, the Tissue Forge installation can be used after executing the following steps from a terminal with the root of the installation as the current directory.

On Windows

call etc/vars
conda activate %TFENV%

On Linux and MacOS

source etc/vars.sh
conda activate $TFENV

Launching the provided Python examples are then as simple navigating to the tissue_forge Python module and then executing the following,

python examples/cell_sorting.py

Likewise, Tissue Forge can be imported in Python scripts and interactive consoles,

import tissue_forge as tf

Installing From Source

Supported installation from source uses Git and Miniconda for building and installing most dependencies. In addition to requiring Git and Miniconda, installation from source on Windows requires Visual Studio 2019 Build Tools, on Linux requires the Mesa packages libgl1-mesa-dev and libegl1-mesa-dev, and on MacOS requires Xcode with 10.9 SDK or greater.

To execute the standard installation, open a terminal in a directory to install Tissue Forge and clone the Tissue Forge repository,

git clone --recurse-submodules https://github.com/tissue-forge/tissue-forge

From the directory containing the tissue-forge root directory, perform the following.

On Windows

call tissue-forge/package/local/install

On Linux

bash tissue-forge/package/local/install.sh

On MacOS, specify the installed MacOS SDK (e.g., for 10.9)

export TFOSX_SYSROOT=10.9
bash tissue-forge/package/local/install.sh

The standard installation will create the directories tissue-forge_build and tissue-forge_install next to the tissue-forge root directory, the former containing the build files, and the latter containing the installed binaries and conda environment. The source and build directories can be safely deleted after installation. The conda environment will be installed in the subdirectory env. To activate the conda environment with the Tissue Forge Python module, perform the following.

On Windows

call tissue-forge_install/etc/vars
conda activate %TFENV%

On Linux and MacOS

source tissue-forge_install/etc/vars.sh
conda activate $TFENV

Launching the provided examples are then as simple as the following

python tissue-forge/py/examples/cell_sorting.py

Likewise Tissue Forge can be imported in Python scripts and interactive consoles

import tissue_forge as tf

Customizing the Build

Certain aspects of the installation can be readily customized. The source directory tissue-forge/package/local contains subdirectories linux, osx and win containing scripts install_vars.sh and install_vars.bat for Linux/MacOS and Windows, respectively, which declare default installation environment variables. These environment variables can be customized to specify where to find, build and install Tissue Forge, as well as the build configuration. For example, to install Tissue Forge from a source directory MYTFSRC, build Tissue Forge at path MYTFBUILD in debug mode and install into directory MYTFINSTALL, perform the following.

On Windows

call %MYTFSRC%/package/local/win/install_vars
set TFBUILD_CONFIG=Debug
set TFSRCDIR=%MYTFSRC%
set TFBUILDDIR=%MYTFBUILD%
set TFINSTALLDIR=%MYTFINSTALL%
call %TFSRCDIR%/package/local/win/install_env
conda activate %TFENV%
call %TFSRCDIR%/package/local/win/install_all

On Linux

source $MYTFSRC/package/local/linux/install_vars.sh
export TFBUILD_CONFIG=Debug
export TFSRCDIR=$MYTFSRC
export TFBUILDDIR=$MYTFBUILD
export TFINSTALLDIR=$MYTFINSTALL
bash ${TFSRCDIR}/package/local/linux/install_env.sh
conda activate $TFENV
bash ${TFSRCDIR}/package/local/linux/install_all.sh

On MacOS

source $MYTFSRC/package/local/osx/install_vars.sh
export TFBUILD_CONFIG=Debug
export TFSRCDIR=$MYTFSRC
export TFBUILDDIR=$MYTFBUILD
export TFINSTALLDIR=$MYTFINSTALL
bash ${TFSRCDIR}/package/local/osx/install_env.sh
conda activate $TFENV
bash ${TFSRCDIR}/package/local/osx/install_all.sh

Note that Tissue Forge assumes that conda is installed in a typical location on Linux and MacOS. If Tissue Forge has trouble finding the conda script conda.sh, then it can be provided during build customization using the environment variable TFCONDAENV.

export TFCONDAENV=${HOME}/mambaforge/etc/profile.d/conda.sh

The default Python version of the installation is 3.8, though Tissue Forge has also been tested on Windows, Linux and MacOS for Python versions 3.9, 3.10 and 3.11. To specify a different version of Python, simply add a call to update the conda environment in the previous commands before calling install_all.

Enabling Interactive Tissue Forge

Tissue Forge supports interactive modeling and simulation specification in an IPython console and Jupyter Notebook. To enable interactive Tissue Forge in an IPython console, activate the installed environment as previously described and install the ipython package from the conda-forge channel,

conda install -c conda-forge ipython

To enable interactive Tissue Forge in a Jupyter Notebook, activate the installed environment as previously described and install the notebook, ipywidgets and ipyevents packages from the conda-forge channel,

conda install -c conda-forge notebook ipywidgets ipyevents

Enabling GPU Acceleration

Tissue Forge supports GPU acceleration on Windows and Linux using CUDA. To enable GPU acceleration, simply tell Tissue Forge to build with CUDA support and specify the compute capability of all available GPUs in the typical way before calling install.

On Windows

set TF_WITHCUDA=1
set CUDAARCHS=35;50
call tissue-forge/package/local/install

On Linux

export TF_WITHCUDA=1
export CUDAARCHS=35;50
bash tissue-forge/package/local/install.sh

Note

Tissue Forge currently supports offloading computations onto CUDA-supporting GPU devices of compute capability 3.5 or greater and installed drivers of at least 456.38 on Windows, and 450.80.02 on Linux.

Setting Up a Development Environment

The Tissue Forge codebase includes convenience scripts to quickly set up a development environment for building models and extensions in C++. The same environment deployed in Installing From Source can be used to build a customized version of Tissue Forge. Set up for setting up a development environment is as simple as getting the Tissue Forge source code, and installing the pre-configured conda environment. As such, all requirements described in Installing From Source are also applicable for building a custom version of Tissue Forge.

To set up a development environment, clone the Tissue Forge repository, open a terminal in the directory containing the tissue-forge root directory and perform the following.

On Windows

call tissue-forge/package/local/win/install_vars
call tissue-forge/package/local/win/install_env

On Linux

bash tissue-forge/package/local/linux/install_vars.sh
bash tissue-forge/package/local/linux/install_env.sh

On MacOS

bash tissue-forge/package/local/osx/install_vars.sh
bash tissue-forge/package/local/osx/install_env.sh

The standard configuration will set the build and installation directories to tissue-forge_build and tissue-forge_install next to the tissue-forge root directory, respectively, the latter containing the conda environment with the build dependencies. These locations can be customized in the same way as described in Customizing the Build, or in your favorite IDE. For configuring CMake, refer to the script install_core in the subdirectory of package/local/* that corresponds to your platform, which is the script behind the automated installation from source. This script includes all variables and the compiler(s) that correspond to building a fully customized version of Tissue Forge.

Tissue Forge supports the Release, Debug and RelWithDebInfo build types. The computational core of Tissue Forge and C++ front-end can be found throughout the subdirectory source. Bindings for Python language support are generated using SWIG. To develop the Python interface (or generate an interface for a new language), refer to the SWIG script wraps/py/tissue_forge.i. To develop the C language interface, refer to the directory wraps/C.