feat/update mlflow-related sample scripts #12471
Conversation
…s return version set
…at/update-mlflow-ui
github-actions
bot
added
ingestion
Codecov ReportAttention: Patch coverage is
❌ Your patch check has failed because the patch coverage (81.81%) is below the target coverage (90.00%). You can increase the patch coverage or adjust the target coverage.
... and 34 files with indirect coverage changes Continue to review full report in Codecov by Sentry.
|
datahub-cyborg
bot
added
the
needs-review
ryota-cloud
requested review from
ryota-cloud
and removed request for
ryota-cloud
| # Create model group | ||
| model_group_urn = client.create_model_group( | ||
| group_id="airline_forecast_models_group", | ||
| properties=models.MLModelGroupPropertiesClass( |
There was a problem hiding this comment.
better abstraction! also how about putting detail of properties class into create_model_group method ? then customer doesn't have to know properties class. (but I'd like to know pros/cons of doing it )
| ), | ||
| ) | ||
|
|
||
| run_urn = client.create_training_run( |
There was a problem hiding this comment.
this method make abstraction better!
- nit on naming, create_run is shorter? (unless run is overloaded term) or experiment_run ?
- how about allowing to put experiment_urn in argument, then create run-> experiment relationship automaticaly? (then user do not call add_run_to_experiment)
| ) | ||
|
|
||
| # Creating a model with property classes | ||
| model_urn = client.create_model( |
There was a problem hiding this comment.
how about adding model_group_run as argument, then user do not have to call add_model_to_model_group
| # Creating a model with property classes | ||
| model_urn = client.create_model( | ||
| model_id="arima_model", | ||
| properties=models.MLModelPropertiesClass( |
There was a problem hiding this comment.
IMO, this also expose internal details of properties class to user, but since this properties class is complicated, hiding it might have pros/cons. might be helpful to add one more helper function for metrics/param? (not strong opinion)
datahub-cyborg
bot
added
pending-submitter-response
datahub-cyborg
bot
added
needs-review
Checklist