This project aims to compare two popular machine learning frameworks. The frameworks that are going to be used are TensorFlow and PyTorch. Each framework is going to be used on a natural language processing problem. Particularly, they are going to be used to solve a sentiment analysis problem.
- Import the conda environment named environment.yml. This conda environment includes all the necessary libraries to execute the code.
- The dataset can be downloaded from http://jmcauley.ucsd.edu/data/amazon/. The one used for this experiment was the 5-core video game review dataset. After downloading the dataset, place it in the Data folder and run split_data.py and tokenize.py. This can also be done by running the cells in the Data cleaning notebook.
- The models can be executed using the Jupyter Notebooks or the source code found in the Python Files folder. In order to run them from the source code files, the files found in the Python Files folder must first be removed and placed next to the Data and Graphs folders.