SDK 2.0: deploy() returns NoneType

[juliensimon]

Using the latest SDKv2 build:

from sagemaker.model import Model

tf_model = Model(
    image_uri="763104351884.dkr.ecr.us-east-1.amazonaws.com/tensorflow-inference:2.1.0-cpu",
    model_data=model_path,
    role=role)

tf_predictor = tf_model.deploy(
                 initial_instance_count=1, 
                 instance_type='ml.m5.large',
                 endpoint_name=tf_endpoint_name)

# the endpoint is created correctly

type(tf_predictor)
NoneType

This should return a sagemaker.predictor.Predictor object, right? Have I missed something?

[juliensimon]

For the record, replacing Model with TensorFlowModel works fine

from sagemaker.tensorflow.model import TensorFlowModel

tf_model = TensorFlowModel(
    model_data=model_path,
    framework_version='2.1.0',
    role=role)

[laurenyu]

the generic Model class doesn't have a default predictor class associated with it: https://github.com/aws/sagemaker-python-sdk/blob/master/src/sagemaker/model.py#L66

for TF specifically, sagemaker.tensorflow.serving.Model has now become sagemaker.tensorflow.model.TensorFlowModel

[juliensimon]

ok, got it. Definitely worth mentioning in the doc :)

[laurenyu]

I've updated the docs in #1756. Thanks for bringing this up!

[Mahdi-Hosseinali]

I was just wondering why my model does not return a predictor that I can use in the notebook until I found this. I used this document page:
https://sagemaker.readthedocs.io/en/stable/api/inference/model.html
I used the artifacts to load my model so this page is all I needed. Could not figure out deploy returns None.