마젠타명령어

# 1. 데이터 전처리
# 1-1 미디파일 -> tfrecrod
!convert_dir_to_note_sequences --input_dir=/content/drive/MyDrive/class_data/tropical_melody_midis --output_file=/content/drive/MyDrive/class_data/tfrecord

# 1-2 tfrecrod -> 모델학습을 위한 데이터셋
!melody_rnn_create_dataset --input=/content/drive/MyDrive/class_data/tfrecord --output_dir=/content/drive/MyDrive/class_data/dataset --eval_ratio=0.0 --config=basic_rnn

참고 : https://github.com/magenta/magenta/tree/main/magenta/models/melody_rnn

# 2. 모델 학습
!melody_rnn_train --config=basic_rnn --run_dir=/content/drive/MyDrive/class_data/run1 --sequence_example_file=/content/drive/MyDrive/class_data/dataset/training_melodies.tfrecord --hparams="batch_size=64,rnn_layer_sizes=[64]" --num_training_steps=1000

# 3. 음악 생성
!melody_rnn_generate --config=basic_rnn --run_dir=/content/drive/MyDrive/class_data/run1 --output_dir=/content/drive/MyDrive/class_data/generated --num_outputs=100 --num_steps=128 --hparams="batch_size=64,rnn_layer_sizes=[64]" --primer_melody="[60]"