Machine learning models and tools
hml is installed as a Python package using pip:
git clone git@github.com:hacmorgan/hml
python3 -m pip install --upgrade hml
When a model is first trained, an experiment directory is created. In addition to storing training logs and checkpoints, and any progress outputs (e.g. for a generative model), hml will copy the model architecture module(s) there, such that they can be retrieved later for continuing to train or inference/generation, even if the latest version of the architecture has been changed.