Created
November 26, 2022 18:46
-
-
Save mizoru/d4fc7596cd06d84f94e4809a1cb683a3 to your computer and use it in GitHub Desktop.
get_data_to_buffer with energies
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
def get_data_to_buffer(train_config): | |
buffer = list() | |
text = process_text(train_config.data_path) | |
audio_files = sorted(Path(train_config.audio_path).iterdir()) | |
spec = Spectrogram(512) | |
start = time.perf_counter() | |
for i, file in tqdm(zip(range(len(text)), audio_files)): | |
mel_gt_name = os.path.join( | |
train_config.mel_ground_truth, "ljspeech-mel-%05d.npy" % (i+1)) | |
mel_gt_target = np.load(mel_gt_name) | |
duration = np.load(os.path.join( | |
train_config.alignment_path, str(i)+".npy")) | |
character = text[i][0:len(text[i])-1] | |
character = np.array( | |
text_to_sequence(character, train_config.text_cleaners)) | |
wav, sr = torchaudio.load(file) | |
energy = torch.norm(spec(wav), p=2, dim=1)[0] | |
assert energy.shape[0] == mel_gt_target.shape[0] | |
np.save(f"energies/{i}", energy) | |
character = torch.from_numpy(character) | |
duration = torch.from_numpy(duration) | |
mel_gt_target = torch.from_numpy(mel_gt_target) | |
buffer.append({"text": character, "duration": duration, | |
"mel_target": mel_gt_target, "energy": energy}) | |
end = time.perf_counter() | |
print("cost {:.2f}s to load all data into buffer.".format(end-start)) | |
return buffer |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment