Spaces:
Running
Running
File size: 6,359 Bytes
99bbd30 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 |
import os
import sys
sys.path.append(os.getcwd())
import json
from concurrent.futures import ProcessPoolExecutor
from importlib.resources import files
from pathlib import Path
from tqdm import tqdm
import soundfile as sf
from datasets.arrow_writer import ArrowWriter
import numpy as np
import torch
import torchaudio
def deal_with_audio_dir(audio_dir):
sub_result, durations = [], []
vocab_set = set()
audio_lists = list(audio_dir.rglob("*.wav"))
for line in audio_lists:
text_path = line.with_suffix(".normalized.txt")
text = open(text_path, "r").read().strip()
duration = sf.info(line).duration
if duration < 0.4 or duration > 30:
continue
sub_result.append({"audio_path": str(line), "text": text, "duration": duration})
durations.append(duration)
vocab_set.update(list(text))
return sub_result, durations, vocab_set
def main():
result = []
duration_list = []
text_vocab_set = set()
# process raw data
#executor = ProcessPoolExecutor(max_workers=max_workers)
#futures = []
#
#for subset in tqdm(SUB_SET):
# dataset_path = Path(os.path.join(dataset_dir, subset))
# [
# futures.append(executor.submit(deal_with_audio_dir, audio_dir))
# for audio_dir in dataset_path.iterdir()
# if audio_dir.is_dir()
# ]
#for future in tqdm(futures, total=len(futures)):
# sub_result, durations, vocab_set = future.result()
# result.extend(sub_result)
# duration_list.extend(durations)
# text_vocab_set.update(vocab_set)
#executor.shutdown()
train_scp = "/ailab-train/speech/zhanghaomin/datas/v2cdata/test.scp"
v2a_path = "/ailab-train/speech/zhanghaomin/codes3/MMAudio-main/output_v2c_s44/"
#v2a_path = "/ailab-train/speech/zhanghaomin/codes3/v2a_v2cdata/"
with open(train_scp, "r") as fr:
for line in fr.readlines():
video, txt, audio = line.strip().split("\t")
####v2a_audio = v2a_path + video.replace("/", "__") + ".flac"
v2a_audio = v2a_path + video.replace("/", "__")[:-4] + ".wav"
if not os.path.exists(video) or not os.path.exists(audio) or not os.path.exists(v2a_audio):
print(video, audio, v2a_audio)
continue
waveform, sr = torchaudio.load(audio)
duration = waveform.shape[-1] / sr
waveform_v2a, sr_v2a = torchaudio.load(v2a_audio)
duration_v2a = waveform_v2a.shape[-1] / sr_v2a
if duration_v2a >= duration:
waveform_v2a = waveform_v2a[:, :int(sr_v2a*duration)]
else:
waveform_v2a = torch.cat([waveform_v2a, torch.zeros([waveform_v2a.shape[0], int(sr_v2a*duration)-waveform_v2a.shape[1]])], dim=1)
duration_v2a = duration
energy_v2a = []
for i in range(int(duration_v2a/(256/24000))):
energy_v2a.append(waveform_v2a[0,int(i*sr_v2a*(256/24000)):int((i+1)*sr_v2a*(256/24000))].abs().mean())
energy_v2a = np.array(energy_v2a)
energy_v2a = energy_v2a / max(energy_v2a)
#print(len(energy_v2a), max(energy_v2a), min(energy_v2a), energy_v2a.mean())
np.savez(v2a_audio+".npz", energy_v2a)
energy = []
for i in range(int(duration/(256/24000))):
energy.append(waveform[0,int(i*sr*(256/24000)):int((i+1)*sr*(256/24000))].abs().mean())
energy = np.array(energy)
energy = energy / max(energy)
#print(len(energy), max(energy), min(energy), energy.mean())
np.savez(audio+".npz", energy)
d = {}
d["audio_path"] = audio
d["text"] = txt
d["duration"] = duration
d["energy"] = v2a_audio+".npz"
result.append(d)
duration_list.append(duration)
text_vocab_set.update(list(txt))
print(len(result), result[:2]) # 354218 [{'audio_path': '/ailab-train/speech/zhanghaomin/datas/libritts/LibriTTS/train-clean-100/7635/105409/7635_105409_000088_000000.wav', 'text': '"There is no \'but.\' I said, do you remember?"', 'duration': 2.31}, {'audio_path': '/ailab-train/speech/zhanghaomin/datas/libritts/LibriTTS/train-clean-100/7635/105409/7635_105409_000061_000002.wav', 'text': 'They know it.', 'duration': 0.76}]
print(len(duration_list), duration_list[:2]) # 354218 [2.31, 0.76]
print(len(text_vocab_set)) # 78
# save preprocessed dataset to disk
if not os.path.exists(f"{save_dir}"):
os.makedirs(f"{save_dir}")
print(f"\nSaving to {save_dir} ...")
with ArrowWriter(path=f"{save_dir}/raw.arrow") as writer:
for line in tqdm(result, desc="Writing to raw.arrow ..."):
writer.write(line)
# dup a json separately saving duration in case for DynamicBatchSampler ease
with open(f"{save_dir}/duration.json", "w", encoding="utf-8") as f:
json.dump({"duration": duration_list}, f, ensure_ascii=False)
# vocab map, i.e. tokenizer
with open(f"{save_dir}/vocab.txt", "w") as f:
for vocab in sorted(text_vocab_set):
f.write(vocab + "\n")
print(f"\nFor {dataset_name}, sample count: {len(result)}")
print(f"For {dataset_name}, vocab size is: {len(text_vocab_set)}")
print(f"For {dataset_name}, total {sum(duration_list)/3600:.2f} hours")
if __name__ == "__main__":
max_workers = 36
tokenizer = "char" # "pinyin" | "char"
#SUB_SET = ["train-clean-100", "train-clean-360", "train-other-500"]
#dataset_dir = "/ailab-train/speech/zhanghaomin/datas/libritts/LibriTTS"
#dataset_name = f"LibriTTS_{'_'.join(SUB_SET)}_{tokenizer}".replace("train-clean-", "").replace("train-other-", "")
dataset_name = "v2c_s44_test_char"
save_dir = str(files("f5_tts").joinpath("../../")) + f"/data/{dataset_name}"
print(f"\nPrepare for {dataset_name}, will save to {save_dir}\n")
main()
# For LibriTTS_100_360_500_char, sample count: 354218
# For LibriTTS_100_360_500_char, vocab size is: 78
# For LibriTTS_100_360_500_char, total 554.09 hours
|