def build_keras_inputs(working_dir, counts_per_speaker):
# counts_per_speaker: If you specify --counts_per_speaker 600,100, that means for each speaker,
# you're going to generate 600 samples for training and 100 for testing. One sample is 160 frames
# by default (~roughly 1.6 seconds).
counts_per_speaker = [int(b) for b in counts_per_speaker.split(',')]
kc = KerasFormatConverter(working_dir)
kc.generate(max_length=NUM_FRAMES, counts_per_speaker=counts_per_speaker)
kc.persist_to_disk()
def main(args):
ensures_dir(args.working_dir)
if args.preprocess:
if args.audio_dir is None:
return Audio(cache_dir=args.working_dir, audio_dir=args.audio_dir, sample_rate=args.sample_rate)
if args.build_keras_inputs:
counts_per_speaker = [int(b) for b in args.counts_per_speaker.split(',')]
kc = KerasFormatConverter(args.working_dir)
kc.generate(max_length=NUM_FRAMES, counts_per_speaker=counts_per_speaker)
kc.persist_to_disk()
if args.train_embedding:
if args.pre_training_phase:
start_training(args.working_dir, pre_training_phase=args.pre_training_phase, epochs=args.epochs_pretrain)
start_training(args.working_dir, pre_training_phase=False, epochs=args.epochs_triplet)
if args.train_classifier:
start_training(args.working_dir, pre_training_phase=False, classify=True, epochs=args.epochs_classifier)
def build_keras_inputs(working_dir, counts_per_speaker):
counts_per_speaker = [int(b) for b in counts_per_speaker.split(',')]
kc = KerasFormatConverter(working_dir)
kc.generate(max_length=NUM_FRAMES, counts_per_speaker=counts_per_speaker)
kc.persist_to_disk()