GCP AI Platform (unified) Python export_model FailedPrecondition: 400 Exporting artifact in format `` is not supported about python-aiplatform HOT 6 CLOSED

@dizcology That works!

It printed:
"Long running operation..."
export_model_response: done

and I have a bucket with one model.tflite file

The export_model_sample should include the "export_format_id" as well as the example from GCP.

Thank you very much for your time and help.

from python-aiplatform.

Hello @zdhernandez: Thank you for reporting the issue. The service defaults to the first supported export formats if it is not specified in the request: https://github.com/googleapis/python-aiplatform/blob/master/google/cloud/aiplatform_v1beta1/types/model_service.py#L220-L225

So it appears that the model you have does not support any export format. Only certain model types support model export: tabular and edge according to the documentation.

from python-aiplatform.

@dizcology The model that I am trying to export is of type: MOBILE_TF_LOW_LATENCY_1
Did I forget to pass an extra argument?

update: I double-checked the AIPlatform and I can confirm it is an edge model.
When I go under the "Model Properties" tab I can see:

• Algorithm: AutoML Edge
• Objective: Image classification (Single-label)

from python-aiplatform.

Thank you @zdhernandez, for good measure could you try again with a modified request like this: https://github.com/googleapis/python-aiplatform/blob/master/samples/snippets/export_model_video_action_recognition_sample.py#L34-L37

(Looking at the sample test, it seems you might need to format the string in a certain particular way.)

from python-aiplatform.

@dizcology I tried:

output_config = { "artifact_destination": "gs://my-bucket-vcm/model-123123123123/tflite/2021-01-28T16:00:00.000Z/", "export_format_id": "tflite", }

and got:

Traceback (most recent call last):
  File "/app/main.py", line 676, in <module>
    response = aiplatform_model_client.export_model(name=name, output_config=output_config)
  File "/usr/local/lib/python3.8/site-packages/google/cloud/aiplatform_v1beta1/services/model_service/client.py", line 924, in export_model
    request.output_config = output_config
  File "/usr/local/lib/python3.8/site-packages/proto/message.py", line 584, in __setattr__
    pb_value = marshal.to_proto(pb_type, value)
  File "/usr/local/lib/python3.8/site-packages/proto/marshal/marshal.py", line 205, in to_proto
    pb_value = rule.to_proto(value)
  File "/usr/local/lib/python3.8/site-packages/proto/marshal/rules/message.py", line 32, in to_proto
    return self._descriptor(**value)
TypeError: Parameter to MergeFrom() must be instance of same class: expected google.cloud.aiplatform.v1beta1.GcsDestination got str.

from python-aiplatform.

I think the value of artifact_destination needs to be a Python dict or a protobuf message. Please try something like the following:

output_config = {
    "artifact_destination": {"output_uri_prefix": "gs://..."},
    "export_format_id": "tflite",
}

from python-aiplatform.