It creates music using midi datasets dividing it into dimensions and then using encoder and decoders Neural networks
Install dependencies in python3 by running pip install -r requirements.txt.
-
Find some dataset to train on.
-
Run preprocess_songs.py. This will load all midi files from your midi files folder data/raw/ into data/interim/samples.npy & lengths.npy. You can point the script to a location using the --data_folder flag one or multiple times to make it include more paths to index midi files.
-
Run train.py. This will train your network and store the progress from time to time (EPOCHS_TO_SAVE to change storing frequency).
-
Run composer.py --model_path "e300" where "e300" indicates the folder the stored model is in. Look into the results/history folder to find your trained model folders.The model path is an example.