Closed nathanielrindlaub closed 4 years ago
There's a 5 MB limit to requests made to Sagemaker endpoints, so try passing in an S3 URI, intercepting that in the input_handler() in inference.py, and downloading the image there before submitting it to the model.
This is how it's currently working. Seems OK for now but definitely open to alternatives.
Example of someone else doing it: https://github.com/aws/sagemaker-tensorflow-serving-container/issues/73
There's a 5 MB limit to requests made to Sagemaker endpoints, so try passing in an S3 URI, intercepting that in the input_handler() in inference.py, and downloading the image there before submitting it to the model.