HuggingFace on Serverless Sagemaker

Follow the steps in the (sagemaker) notebook huggingface-on-serverless-sagemaker.ipynb

  • Load a model from huggingface hub and store on S3.
  • Register a huggingface model as a Sagemaker model.
  • Expose your HuggingFace Model to the Outside World.
  • Create a lambda handler that gets the request and forwards it to the Sagemaker endpoint.
  • Create a serverless.yml file to deploy our Serverless Sagemaker + Lambda + Api Gateway
  • Deploy our stack to AWS
  • Call our model using curl
  • Remove the stack

finally this works now!

Architecture

assets/serverless-sagemaker.drawio.png