(All requirements can be installed with pip)
pytorch
numpy
tqdm
sklearn
matplotlib

train
    training loop for an epoch, prints training loss and accuracy/r2
test
    evaluates test data on trained model, prints training loss and accuracy/r2

ArtPopDataset
    custom dataset with one input set and one label set

KeyDataset
    custom dataset with two input sets and one label set

prepare_art_pop_datasets
    create datasets with one input and split into training, validation, and testing

prepare_key_datasets
    create datasets with one input and split into training, validation, and testing

evaluate
    returns loss and accuracy/r2 for a model

adjust_learning_rate
    multiplies learning rate parameter

precision_recall_f1score
    prints following metrics (average over an epoch) for classification models:
        precision: ability to not label false positives
        recall: ability to find positives
        f1score: average of precision and recall

eval_regression
    returns r2 score for regression model

save
    saves best model, called when validation loss exceeds the previous best

load
    loads best model for testing

Key
    2-layer CNN for pitch vectors
    2-layer CNN for timbre vectors
    2-layer RNN for concatenated CNN outputs
    Linear output layers, Tanh activations
    
Artist
    3 fully-connected layers, Tanh activations
    
Popularity
    3 fully-connected layers, Tanh activations

1) run train.py with desired model type and parameters
2) system prints loss and metrics for training, validation, and testing
3) view loss and accuracy(key, artist) or r2 score(popularity)

https://drive.google.com/open?id=1BxjgrdCs2t7Z70Z2ldaLxPkUT5JTxZEc
download and place in the data folder