This repository contains MLFlow integration toolchain for Legion platform.
Legion's toolchain for MLFlow is responsible for running process of training MLFlow models. It produces model as ZIP archive in Legion's General Python Prediction Interface.
This toolchain requires MLFlow tracking server to be installed (MLFlow tracking server is bundled in HELM Chart).
Packager's container runs as a part of Tekton pipeline.
Container's command (entrypoint in Docker's terms) is taken from entrypoint field of PackagingIntegration object.
The command is also provided by 2 arguments:
- path to unzipped model artifact directory
- path to
K8sPackagermanifest file