Hi guys, your framework's support for out of the box batched inference support is very good. That's one of the main reasons I am exploring the sdk Speaking of the same how would I deploy a CLIP model in sagemaker inference endpoint My current flow is as follows Input JSON : { "type": "text" or "image" "inputs" : image or text content } depending upon the input type I perform the inference and return the embedding This is because aws sagemaker inference endpoint exposes only one endpoint How can I do a similar flow with batched inferencing in bentoml. I am asking this as in your clip example you created 3 endpoints, but sagemaker inference endpoint doesnt support more than one endpoint