generate_abc.py

# The make model and generate sequence functions are derived from the repo at
# https://github.com/gauravtheP/Music-Generation-Using-Deep-Learning
# The newer models independent of this original repo will consist of genre wise music generation
# And a model with a deeper layers
from tensorflow.keras.layers import LSTM, Dropout, Dense, Activation, Embedding
from tensorflow.keras.models import Sequential
from tensorflow.keras import backend
import numpy as np
import os
import sys
import json

sys.path.append('/home/runner/.site-packages/')


charIndex_json = "static/json/char_to_index.json"
BATCH_SIZE = 16
SEQ_LENGTH = 64


def make_model(unique_chars):
    model = Sequential()

    model.add(Embedding(input_dim=unique_chars,
                        output_dim=512, batch_input_shape=(1, 1)))

    model.add(LSTM(256, return_sequences=True, stateful=True))
    model.add(Dropout(0.2))

    model.add(LSTM(256, return_sequences=True, stateful=True))
    model.add(Dropout(0.2))

    model.add(LSTM(256, stateful=True))
    model.add(Dropout(0.2))

    model.add((Dense(unique_chars)))
    model.add(Activation("softmax"))

    return model


def generate_sequence(initial_index, seq_length):
    with open(os.path.join(charIndex_json)) as f:
        char_to_index = json.load(f)
    index_to_char = {i: ch for ch, i in char_to_index.items()}
    unique_chars = len(index_to_char)

    backend.clear_session()
    model = make_model(unique_chars)
    model.load_weights('static/weights/Weights_90.h5')

    sequence_index = [initial_index]

    for _ in range(seq_length):
        batch = np.zeros((1, 1))
        batch[0, 0] = sequence_index[-1]

        predicted_probs = model.predict_on_batch(batch).ravel()
        sample = np.random.choice(
            range(unique_chars), size=1, p=predicted_probs)

        sequence_index.append(sample[0])
    print(len(sequence_index))
    seq = ''.join(index_to_char[c] for c in sequence_index)

    cnt = 0
    for i in seq:
        cnt += 1
        if i == "\n":
            break
    seq1 = seq[cnt:]

    cnt = 0
    for i in seq1:
        cnt += 1
        if i == "\n" and seq1[cnt] == "\n":
            break
    seq2 = seq1[:cnt]

    return seq2

# ar = Any number between 0 to 86 which will be given as initial charcter to model for generating sequence
# ln = The length of music sequence you want to generate. Typical number is between 300-600. Too small number will generate hardly generate any sequence
# instr = The instrument code is detailed in static/csv/abcmidi_instrument_name


def generate_abc_file(ar, ln, instr):
    ar = max(ar % 87, 0)
    ln = max(ln, 600)
    instr = max(instr % 129, 1)

    music = generate_sequence(ar, ln)
    print("\nMUSIC SEQUENCE GENERATED: \n")
    print(music)

    with open('static/abc/generated.abc', 'w') as abc:
        if music.find('X:') == -1:
            abc.write('X:1\n')
        if music.find('T:') == -1:
            abc.write('T:RNN generated\n')
        if music.find('M:') == -1:
            abc.write('M:3/4\n')
        if music.find('L:') == -1:
            abc.write('L:1/8\n')
        if music.find('Q:') == -1:
            abc.write('Q:1/4=120\n')
        if music.find('W:') == -1:
            abc.write("W:Generated music\n")
        if music.find('K:') == -1:
            abc.write('K:C\n')
        abc.write(f'%%MIDI program {instr}\n')
        abc.write(music)