First project from Machine Learning Engineer Nanodegree Program Udacity course
This project consists in implement a Recurrent Neural Network (RNN) that is capable to predict if a text that has a good or bad opinion about a movie. The dataset used to train and test is based at IMDB. The porpose is show how to use AWS Sagemaker features to do that. It contains 3 parts: * A notebook that explain how to prepare the data, train, test and creation of an endpoint to access the trained model; * A source code to use in AWS Lambda function that calls the endpoint created at step above and; * A html file that calls an API gateway that expose an URL to the world.
All process to get the data file, transform it, train, test create the endpoint and use it is explain in notebook file IMDBAnalyser.ipynb
All process to create and deploy the code that runs in AWS Lambda Function can be find here.
The source code is in predict.py.
To be resilent about content-type, I had to use the try: except resource, so it's possible to call the URL from a browser or postman, for example.
The source code is in index.html. This page calls AWS API Gateway, that calls the lambda, that calls the endpoint created by Sagemaker. The figure bellow explain better.
The result is shown at the next images.
