so-vits-svc/preprocess_flist_config.py

120 lines
4.8 KiB
Python
Raw Normal View History

2023-03-10 10:11:04 +00:00
import argparse
2023-06-26 06:57:53 +00:00
import json
import os
2023-03-10 10:11:04 +00:00
import re
2023-06-26 06:57:53 +00:00
import wave
from random import shuffle
2023-03-10 10:11:04 +00:00
2023-07-22 13:56:02 +00:00
from loguru import logger
2023-03-10 10:11:04 +00:00
from tqdm import tqdm
2023-05-16 05:17:51 +00:00
import diffusion.logger.utils as du
2023-03-10 10:11:04 +00:00
pattern = re.compile(r'^[\.a-zA-Z0-9_\/]+$')
def get_wav_duration(file_path):
try:
with wave.open(file_path, 'rb') as wav_file:
# 获取音频帧数
n_frames = wav_file.getnframes()
# 获取采样率
framerate = wav_file.getframerate()
# 计算时长(秒)
return n_frames / float(framerate)
except Exception as e:
logger.error(f"Reading {file_path}")
raise e
2023-03-10 10:11:04 +00:00
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument("--train_list", type=str, default="./filelists/train.txt", help="path to train list")
parser.add_argument("--val_list", type=str, default="./filelists/val.txt", help="path to val list")
parser.add_argument("--source_dir", type=str, default="./dataset/44k", help="path to source dir")
2023-06-07 11:22:47 +00:00
parser.add_argument("--speech_encoder", type=str, default="vec768l12", help="choice a speech encoder|'vec768l12','vec256l9','hubertsoft','whisper-ppg','cnhubertlarge','dphubert','whisper-ppg-large','wavlmbase+'")
2023-05-29 22:35:53 +00:00
parser.add_argument("--vol_aug", action="store_true", help="Whether to use volume embedding and volume augmentation")
2023-07-17 15:46:58 +00:00
parser.add_argument("--tiny", action="store_true", help="Whether to train sovits tiny")
2023-03-10 10:11:04 +00:00
args = parser.parse_args()
2023-07-20 15:28:28 +00:00
config_template = json.load(open("configs_template/config_tiny_template.json")) if args.tiny else json.load(open("configs_template/config_template.json"))
2023-03-10 10:11:04 +00:00
train = []
val = []
idx = 0
spk_dict = {}
spk_id = 0
2023-07-20 15:28:28 +00:00
2023-03-10 10:11:04 +00:00
for speaker in tqdm(os.listdir(args.source_dir)):
spk_dict[speaker] = spk_id
spk_id += 1
wavs = []
for file_name in os.listdir(os.path.join(args.source_dir, speaker)):
if not file_name.endswith("wav"):
continue
if file_name.startswith("."):
2023-03-10 10:11:04 +00:00
continue
file_path = "/".join([args.source_dir, speaker, file_name])
if not pattern.match(file_name):
logger.warning("Detected non-ASCII file name: " + file_path)
if get_wav_duration(file_path) < 0.3:
logger.info("Skip too short audio: " + file_path)
2023-03-10 10:11:04 +00:00
continue
wavs.append(file_path)
2023-03-10 10:11:04 +00:00
shuffle(wavs)
2023-03-24 04:43:29 +00:00
train += wavs[2:]
2023-03-10 10:11:04 +00:00
val += wavs[:2]
shuffle(train)
shuffle(val)
logger.info("Writing " + args.train_list)
2023-03-10 10:11:04 +00:00
with open(args.train_list, "w") as f:
for fname in tqdm(train):
wavpath = fname
f.write(wavpath + "\n")
logger.info("Writing " + args.val_list)
2023-03-10 10:11:04 +00:00
with open(args.val_list, "w") as f:
for fname in tqdm(val):
wavpath = fname
f.write(wavpath + "\n")
2023-05-16 05:17:51 +00:00
2023-07-20 15:28:28 +00:00
d_config_template = du.load_config("configs_template/diffusion_template.yaml")
2023-05-18 11:34:40 +00:00
d_config_template["model"]["n_spk"] = spk_id
d_config_template["data"]["encoder"] = args.speech_encoder
d_config_template["spk"] = spk_dict
2023-05-16 05:17:51 +00:00
2023-03-10 10:11:04 +00:00
config_template["spk"] = spk_dict
config_template["model"]["n_speakers"] = spk_id
2023-05-14 06:39:07 +00:00
config_template["model"]["speech_encoder"] = args.speech_encoder
2023-05-16 05:17:51 +00:00
2023-06-07 11:22:47 +00:00
if args.speech_encoder == "vec768l12" or args.speech_encoder == "dphubert" or args.speech_encoder == "wavlmbase+":
2023-05-14 06:39:07 +00:00
config_template["model"]["ssl_dim"] = config_template["model"]["filter_channels"] = config_template["model"]["gin_channels"] = 768
2023-05-18 11:34:40 +00:00
d_config_template["data"]["encoder_out_channels"] = 768
2023-05-14 06:39:07 +00:00
elif args.speech_encoder == "vec256l9" or args.speech_encoder == 'hubertsoft':
2023-06-30 20:27:18 +00:00
config_template["model"]["ssl_dim"] = config_template["model"]["gin_channels"] = 256
2023-05-18 11:34:40 +00:00
d_config_template["data"]["encoder_out_channels"] = 256
2023-06-01 18:15:42 +00:00
elif args.speech_encoder == "whisper-ppg" or args.speech_encoder == 'cnhubertlarge':
2023-05-24 17:18:18 +00:00
config_template["model"]["ssl_dim"] = config_template["model"]["filter_channels"] = config_template["model"]["gin_channels"] = 1024
d_config_template["data"]["encoder_out_channels"] = 1024
2023-06-04 04:42:55 +00:00
elif args.speech_encoder == "whisper-ppg-large":
config_template["model"]["ssl_dim"] = config_template["model"]["filter_channels"] = config_template["model"]["gin_channels"] = 1280
d_config_template["data"]["encoder_out_channels"] = 1280
2023-05-29 22:35:53 +00:00
if args.vol_aug:
config_template["train"]["vol_aug"] = config_template["model"]["vol_embedding"] = True
2023-05-24 16:41:04 +00:00
2023-07-17 15:46:58 +00:00
if args.tiny:
config_template["model"]["filter_channels"] = 512
2023-07-22 13:56:02 +00:00
logger.info("Writing to configs/config.json")
2023-03-10 10:11:04 +00:00
with open("configs/config.json", "w") as f:
json.dump(config_template, f, indent=2)
2023-07-22 13:56:02 +00:00
logger.info("Writing to configs/diffusion.yaml")
2023-05-16 05:17:51 +00:00
du.save_config("configs/diffusion.yaml",d_config_template)