# Installed into <slug>-inference-env, a clone of torch-inference-pipeline, which
# ships torch. NVIDIA's structured-data-models (KumoTabular) is not on PyPI, so it
# is installed from its repository at the commit the predictor is written against.
structured-data-models @ git+https://github.com/NVIDIA/structured-data-models.git@5d663369ff8ead4a78c910dbb8483d558e46627b
