Unable to test model from Quick Start tutorial

Hi, I’m following the Quick Start Colab Notebook.

Everything works smoothly until step “Test your Model with an Inference Request”, where the code fails with this error: ModelServingException: Deployment not created or running

---------------------------------------------------------------------------
RestAPIError                              Traceback (most recent call last)
/usr/local/lib/python3.7/dist-packages/hsml/engine/serving_engine.py in predict(self, deployment_instance, data)
    133             return self._serving_api.send_inference_request(
--> 134                 deployment_instance, data, through_hopsworks
    135             )

5 frames
RestAPIError: Metadata operation error: (url: http://afbae69b206bd4749b1c4ca8fc3d6dd4-96222519.us-east-2.elb.amazonaws.com/v1/models/fraud:predict). Server response: 
HTTP code: 404, HTTP reason: Not Found

During handling of the above exception, another exception occurred:

ModelServingException                     Traceback (most recent call last)
/usr/local/lib/python3.7/dist-packages/hsml/engine/serving_engine.py in predict(self, deployment_instance, data)
    141             ):
    142                 raise ModelServingException(
--> 143                     "Deployment not created or running. If it is already created, start it by using `.start()` or check its status with .get_state()"
    144                 )
    145 

ModelServingException: Deployment not created or running. If it is already created, start it by using `.start()` or check its status with .get_state()

On c.app.hopsworks.ai, I’ve already verified that the model is running. However, I can’t test the model on the page either:

Any help is appreciated. Thank you!

Hello @kent-wetaus ,

Thanks for reporting this issue.
Apparently, Colab bumped joblib to version 1.2.0 recently and it seems that models saved with joblib 1.2.0 cannot be loaded properly with previous versions of joblib. Since Hopsworks was using joblib 1.1.0, some deployments couldn’t reach the Running status, which is the error you encountered.

This has been fixed already. You can just stop and start your deployment again, nothing else is needed from your side :slight_smile:

Best regards,
Javier