Search code examples
amazon-web-servicesamazon-sagemakerawsdeploy

How to deploy multiple models in amazon sagemaker on various endpoints


I'm recently interested to Amazon Sagemaker and Rekognition. I would like to try AWS instances and deploy ML models.

  • I am going to train and deploy models for diagnoising diseases(Image classification) in various crops. So I would like to deploy multiple models in one instance. (One model for One crop)

  • I am going to use my custom dataset to train models.

  • I am thinking about using Amazon Sagemaker rather than Rekognition. Because the main task is to classify image, rather than recognizing. There is not a significant difference in images for each class.

  • I would like to try models like VIT, ResNet, ...

  • I've checked there is a "multi-model enpoints" in SageMaker. But I would like to know if there is a way to deploy multiple ML models on various endpoints.


Solution

  • Multi model and multi container endpoints share the same endpoint url. Having said that you can invoke the endpoint with a specific model name to invoke corresponding model. Not sure if I understood your question correctly, are you looking to host multiple models in a single instance and have separate endpoint url for each model? If yes, then its not possible with SageMaker endpoints.