It can properly upload both files and directories.

Switched the stage_local_data_in_gcs function to use upload_to_gcs

…arn and Tensorflow

…iner

…utils module

…stead of error message

TemporaryDirectory returns object with `.name` when called directly, but returns a path string when used as a context.

The buckets that we create are regional.
This prevents errors when some service required regional bucket.
E.g. "FailedPrecondition: 400 The Cloud Storage bucket of `gs://...` is in location `us`. It must be in the same regional location as the service location `us-central1`."
We are making the bucket name region-specific since the bucket is regional.

We cannot clean up the directory immediately after calling Model.upload since that call may be asynchronous and return before the model file has been read. The temporary data will be automatically cleaned up by the system later.

The `aiplatform.helpers.get_prebuilt_prediction_container_uri` does not support future framework versions yet. See googleapis#779 (comment)

…age_local_data_in_gcs