Running the pipeline with MLflow

Running from local environment

These command requires cloning this repository

mlflow run . -P hydra_options="main.execute_steps='random_forest'"

To select two or more pipeline steps, you can use a comma-separated list of steps. For example, to run the download and preprocess steps, you can use the following command:

mlflow run . -P hydra_options="main.execute_steps='download,preprocess'"

Inference

mlflow run . -P hydra_options="main.project_name=classification_prod"

Running from remote environment

mlflow run -v {release_tag} repository_url -P hydra_options="main.project_name=classification_prod"

mlflow run -v 1.0.0 https://github.com/eedygreen/genre_classification.git -P hydra_options="main.project_name=released_prod_version"

offline (Batching)

mlflow models predict -t json -i model/input_example.json -m model/