1. Instructions for installing and uninstalling prismatique
1.1. Installing prismatique
For all installation scenarios, first open up the appropriate command line interface. On Unix-based systems, you would open a terminal. On Windows systems you would open an Anaconda Prompt as an administrator.
GPU acceleration is available for prismatique installed on Linux and Windows
machines that have NVIDIA GPUs. You will need to make sure that you have a
NVIDIA driver installed with CUDA version 10.2.89 or greater.
1.1.1. Installing prismatique using pip and conda together
The easiest way to install prismatique involves using both the conda package
manager and pip. While it is possible to install prismatique without the
use of the conda package manager, it is more difficult. Because of this, we
discuss only the simplest installation procedure below.
Of course, to use the conda package manager, one must install either
anaconda3 or miniconda3. For installation instructions for anaconda3
click here; for
installation instructions for miniconda3 click here.
First, open up the appropriate command line interface. On Unix-based systems, you would open a terminal. On Windows systems you would open an Anaconda Prompt as an administrator.
Next, you can optionally update your conda package manager by issuing the following command:
conda update conda
It is recommended that you install prismatique and its dependencies in a
virtual environment: click here
for a discussion on the creation and management of conda virtual
environments. The remaining instructions assumes that you activate the conda
(virtual) environment in which you intend to install prismatique and its
dependencies.
The first dependency that we need to install is pyprismatic. GPU
acceleration is available for pyprismatic (and thus prismatique) if the
following conditions are met:
You are using a Linux or Windows machine that has NVIDIA GPUs.
A NVIDIA driver is installed with CUDA version 10.2.89 or greater.
If the above conditions have been met, and you would like to be able to use GPUs
with prismatique, run the following command:
conda install -c conda-forge pyprismatic=2.*=gpu* cudatoolkit==<X>.<Y>.*
where <X> and <Y> are the major and minor versions of CUDA installed on
your machine, e.g. CUDA version 10.2.89 has a major version of 10, and a
minor version of 2. Users can omit cudatoolkit==<X>.<Y>.* if they do not
require a specific version of cudatoolkit, which should apply to most
scenarios. For CPU support only, run the following command instead:
conda install -c conda-forge pyprismatic=2.*=cpu*
The easiest way to install the remaining dependencies, along with
prismatique is to use pip by running the following command:
pip install prismatique
Another option is to use conda:
conda install -c conda-forge prismatique
As yet another option, you can install the latest development version of
prismatique from the main branch of the prismatique GitHub repository. To do so, one must first clone the
repository by running the following command:
git clone https://github.com/mrfitzpa/prismatique.git
then subsequently change into the root of the cloned repository, and then run the following command:
pip install .
Note that you must include the period as well.
Optionally, for additional features in prismatique, one can install
additional dependencies upon installing prismatique via pip. To install
a subset of additional dependencies (along with the standard installation), run
the following command from the root of the repository:
pip install .[<selector>]
where <selector> can be one of the following:
tests: to install the dependencies necessary for running unit tests;examples: to install the dependencies necessary for executing files stored in <root>/examples, where <root> is the root of the repository;docs: to install the dependencies necessary for documentation generation;all: to install all of the above optional dependencies.
Alternatively, one can run:
pip install prismatique[<selector>]
elsewhere in order to install the latest stable version of prismatique,
along with the subset of additional dependencies specified by <selector>.
1.2. Uninstalling prismatique
If prismatique was installed using pip, then to uninstall, run the
following command:
pip uninstall prismatique
If prismatique was installed using conda, then to uninstall, run the
following command:
conda remove prismatique