Skip to content

Latest commit

 

History

History
21 lines (18 loc) · 1.2 KB

README.md

File metadata and controls

21 lines (18 loc) · 1.2 KB

IMDB-Sentiment-Analysis-AWS

IMDB sentiment Analysis model developed as part of Udacity's AWS Machine Learning Engineering Nano Degree.

  • Project developed as part of the Nano degree
  • Environment – AWS
  • Technologies – AWS Lambda, SageMaker, API Gateway, Pytorch and basic HTML and CSS
  • Algorithm - LSTM Classifier
  • Accuracy - 80%

The following are the items developed into the source repository of the course work for project completion:

  • Completing the build_dict() function in the Feature_Extraction_Training notebook, for creating the dictionary using the vocabulary of the training dataset
    • idnetifying the most frequent words and encoding the text data into numerical values
  • Implimenting method to train Pytorch model using the following approach and implimenting the ZOLS in the train method of the file train/train/py
    • Z -> zero_grad()
    • O -> output (preds)
    • L -> loss
    • S -> optimizer.step()
  • Deploying PyTorch model on an AWS Instance
  • Preprocessing the input text data from the web app and formating the model prediction output into the required format in the file serve/predict.py
  • Completing the model deployment with integration of AWS Lambda and API gateway serices