Transformer-based deep learning model trained on Magenta dataset using Tensorflow and numpy to synthesize classical and jazz music. Uses novel embedding layer to encode continuous temporal information about music notes to improve model versatility to handle complex rhythms and time signatures
Requires maestro dataset, pretty midi, tensorflow. Unzip maestro2.0.0 and put in data/
folder in root directory to use.
Main code is in musictransformer.ipynb