Amazon SageMaker 不支持的媒体类型：图像/jpeg

Question

I trained my own model using Tensorflow and Keras for image classification and I'm trying to deploy and use it with Amazon's SageMaker. I went through the process of converting the mymodel.h5 file into a mymodel.tar.gz file and moving it to the SageMaker S3 bucket. Then, following a tutorial I created the SageMaker model using:

from sagemaker.tensorflow.model import TensorFlowModel
sagemaker_model = TensorFlowModel(model_data = 's3://' + sagemaker_session.default_bucket() + '/model/model.tar.gz',
                                  role = role,
                                  framework_version = '1.12',
                                  entry_point = 'train.py')

And created the endpoint to access the model:

predictor = sagemaker_model.deploy(initial_instance_count=1,
                                   instance_type='ml.m4.xlarge')

Now, since the point of creating this model was for image classification I'm trying to pass the image to my endpoint but have been getting a response of Received client error (415) from primary with message "{"error": "Unsupported Media Type: image/jpeg"}" . After reading up I feel like I may need to do some more work to have access to the image/jpeg content type as it seems like the defaults are application/json , text/libsvm , and text/csv .

Answer 1

Assuming your model has been converted into TF Serving Format, you can confirm the input your model expects using the saved_model_cli . Please see this example which depicts how to use the saved_model_cli . Kindly note how a prediction is made by converting an image to a tensor .