IssueTracker

The following page (https://cloud.google.com/vertex-ai/docs/training/create-training-pipeline#custom-job-model-upload) describes how the parameter `baseOutputDirectory` can be set to make your training job have access to the $AIP_MODEL_DIR environmental variable.

However, this is only valid for the REST command, and for example not for the Python sdk.

I'm assuming that the `staging_bucket` variable in this snippet https://cloud.google.com/vertex-ai/docs/training/create-training-pipeline#aiplatform_create_training_pipeline_custom_job_sample-python 
is the same, but this is not clear from the documentation.