Running a Pipeline with Different Parameters¶
Oftentimes, data scientists/engineers need to run the same pipeline with different parameters.
For instance, they may want to use a different data set for model training.
To produce a parametrized pipeline, use pipeline API's (optional) input_parameters argument,
like so:
lineapy.to_pipeline(
...
input_parameters=["<VARNAME1>", ...],
)
where <VARNAME1> is the name of the variable to be turned into a tunable parameter of the pipeline.
Airflow Example¶
For example, the following creates an Airflow pipeline that can train a model on different data sources:
lineapy.to_pipeline(
pipeline_name="iris_pipeline_parametrized",
artifacts=["iris_preprocessed", "iris_model"],
dependencies={"iris_model": {"iris_preprocessed"}},
input_parameters=["url"],
output_dir="~/airflow/dags/",
framework="AIRFLOW",
)
where url is a variable in the artifact code that references the source data URL.
To execute the generated Airflow pipeline with a different value of url, run:
airflow dags trigger --conf '{"url": "<NEW-DATA-URL>"}' <pipeline_name>_dag
where <pipeline_name> is iris_pipeline_parametrized in the current example.
Or, if running Airflow on UI, trigger the DAG with {"url": "<NEW-DATA-URL>"}
passed as a config JSON.
Limitations¶
Currently, input_parameters only accepts variables from literal assignment
such as a = "123". For each variable to be parametrized, there should be only one
literal assignment across all artifact code for the pipeline. For instance, if both
a = "123" and a = "abc" exist in the pipeline's artifact code, we cannot make
a an input parameter since its reference is ambiguous, i.e., we are not sure which
literal assignment a refers to.