kitware / trame-mnist

Simple trame demonstrator using the MNIST dataset with XAITK for saliency analysis

Trame MNIST

Example application using trame for exploring MNIST dataset in the context of AI training and XAI thanks to XAITK.

• Free software: BSD License
• XAITK Saliency with MNIST
• XAI Discovery Platform | MNIST Sample Data

Installing

For the Python layer it is recommended to use conda to properly install the various ML packages.

conda setup on macOS

Go to conda documentation for your OS

brew install miniforge
conda init zsh

venv setup for AI

# Needed in order to get py3.9 with lzma
# PYTHON_CONFIGURE_OPTS="--enable-framework" pyenv install 3.9.9

conda create --name trame-mnist python=3.9
conda activate trame-mnist

# For development when inside repo
pip install -e .

# For testing (no need to clone repo)
pip install trame-mnist

Running the application

conda activate trame-mnist
trame-mnist

If cuda is available, the application will use your GPU, but you can also force the usage of your cpu by adding to your command line the following argument: --cpu

License

trame-mnist is distributed under the OSI-approved BSD 3-clause License.