Automatic music generator made during hackaton in Amazon Software Development Center Poland in cooperation with Jakub Sobotka.
docker-compose run musicgen numeric generate -o <path_to_output_midi_file>
To start training:
docker-compose run musicgen lstm train start -n <unique_model_name> -p <dir_with_midi_files_to_learn>
To continue training (started earlier):
docker-compose run musicgen lstm train resume -n <model_name_used_previously>
docker-compose run musicgen lstm generate -n <name_of_trained_model> -o <path_to_output_midi_file>