I deployed a model using Azure ML managed endpoint, but I found a bottleneck.
I'm using Azure ML Managed Endpoint to host ML models for object prediction. Our endpoint receives a URL of a picture and is responsible for downloading and predicting the image.
The problem is the bottleneck: each image is downloaded one at a time (synchronously), which is very slow.
Is there a way to download images async or to create multiple threads ? I expected a way to make if faster.
We recommend to use Azure Blob storage to host the images and then use blob storage SDK to fetch the images.