NGINX configuration file for Tensorflow Serving Docker image on AWS Sagemaker

nginx.conf

events {
    # determines how many requests can simultaneously be served
    worker_connections 2048;
}

http {
    server {
    # Increase JSON POST payload to bypass 413 Client Error: Request Entity Too Large for URL
    client_max_body_size 100M;

    # Sagemaker listens to 8080 port and requires /invocations and /ping endpoints
    listen 8080 deferred;

    # redirects requests from SageMaker to TF Serving
    location /invocations {
      proxy_pass http://localhost:8501/v1/models/<<model name here>>>:predict;
    }
    
    location /ping {
      return 200 "OK";
    }
  }
}