[FEATURE]: Add the ability to configure and run a pipeline via CLI
Contact Details [Optional]
Describe the feature you'd like
There are two ways to run a zenml pipeline:
- Hard-code the params in the runner script, or
- Configure the pipeline with a YAML config.
An example YAML might look like:
# config_template.yaml
steps:
step_name:
parameters:
parameter_name: parameter_value
some_other_parameter_name: 2
some_other_step_name:
...
Concretely, a trainer step within a pipeline might look like:
# config.yaml
steps:
trainer:
parameters:
lr: 0.005
And then run:
zenml pipeline run my_pipeline.py -c config.yaml
However, some times it might be nice to do:
zenml pipeline run my_pipeline.py --trainer --lr=0.005
I.e. directly via CLI. There needs to be clever grouping however.
Is your feature request related to a problem?
Sometimes it might be quick to just run a pipeline from the CLI without having to invoke python or create a YAML.
How do you solve your current problem with the current status-quo of ZenML?
Have to either hard-code params in the code or use the YAML config.
Any other comments?
No response
A good place to start developing this feature is to see how the with_config feature works in code. This combined with inspecting the run_pipeline CLI command should be enough to get started
Closing this as we're going in a different direction with this CLI call now that the ZenML server is able to register pipelines.