Fix: colab stdout error, so replace with stdout wrapper

bert_gen.py

@@ -1,5 +1,4 @@
 import argparse
-import sys
 from multiprocessing import Pool
 
 import torch
@@ -10,6 +9,7 @@ import commons
 import utils
 from config import config
 from text import cleaned_text_to_sequence, get_bert
+from tools.stdout_wrapper import get_stdout
 
 
 def process_line(x):
@@ -76,7 +76,7 @@ if __name__ == "__main__":
             for _ in tqdm(
                 pool.imap_unordered(process_line, zip(lines, add_blank)),
                 total=len(lines),
-                file=sys.stdout,
+                file=get_stdout(),
             ):
                 # 这里是缩进的代码块，表示循环体
                 pass  # 使用pass语句作为占位符

preprocess_text.py

@@ -1,6 +1,5 @@
 import json
 import os
-import sys
 from collections import defaultdict
 from random import shuffle
 from typing import Optional
@@ -10,6 +9,7 @@ from tqdm import tqdm
 
 from config import config
 from text.cleaner import clean_text
+from tools.stdout_wrapper import get_stdout
 
 preprocess_text_config = config.preprocess_text_config
 
@@ -52,7 +52,7 @@ def preprocess(
                 lines = trans_file.readlines()
                 # print(lines, ' ', len(lines))
                 if len(lines) != 0:
-                    for line in tqdm(lines, file=sys.stdout):
+                    for line in tqdm(lines, file=get_stdout()):
                         try:
                             utt, spk, language, text = line.strip().split("|")
                             norm_text, phones, tones, word2ph = clean_text(

resample.py

@@ -1,6 +1,5 @@
 import argparse
 import os
-import sys
 from multiprocessing import Pool, cpu_count
 
 import librosa
@@ -10,6 +9,7 @@ from tqdm import tqdm
 
 from config import config
 from tools.log import logger
+from tools.stdout_wrapper import get_stdout
 
 
 def normalize_audio(data, sr):
@@ -88,7 +88,7 @@ if __name__ == "__main__":
 
     pool = Pool(processes=processes)
     for _ in tqdm(
-        pool.imap_unordered(process, tasks), file=sys.stdout, total=len(tasks)
+        pool.imap_unordered(process, tasks), file=get_stdout(), total=len(tasks)
     ):
         pass
 

slice.py

@@ -1,12 +1,13 @@
 import argparse
 import os
 import shutil
-import sys
 
 import soundfile as sf
 import torch
 from tqdm import tqdm
 
+from tools.stdout_wrapper import get_stdout
+
 vad_model, utils = torch.hub.load(
     repo_or_dir="snakers4/silero-vad",
     model="silero_vad",
@@ -106,7 +107,7 @@ if __name__ == "__main__":
         shutil.rmtree(output_dir)
 
     total_sec = 0
-    for wav_file in tqdm(wav_files, file=sys.stdout):
+    for wav_file in tqdm(wav_files, file=get_stdout()):
         time_sec = split_wav(
             wav_file,
             output_dir,

style_gen.py

@@ -1,6 +1,5 @@
 import argparse
 import concurrent.futures
-import sys
 import warnings
 
 import numpy as np
@@ -9,6 +8,7 @@ from tqdm import tqdm
 
 import utils
 from config import config
+from tools.stdout_wrapper import get_stdout
 
 warnings.filterwarnings("ignore", category=UserWarning)
 from pyannote.audio import Inference, Model
@@ -59,7 +59,7 @@ if __name__ == "__main__":
             tqdm(
                 executor.map(save_style_vector, wavnames),
                 total=len(wavnames),
-                file=sys.stdout,
+                file=get_stdout(),
             )
         )
 

tools/log.py

@@ -2,8 +2,8 @@
 logger封装
 """
 from loguru import logger
-import sys
 
+from .stdout_wrapper import get_stdout
 
 # 移除所有默认的处理器
 logger.remove()
@@ -13,4 +13,4 @@ log_format = (
     "<g>{time:MM-DD HH:mm:ss}</g> |<lvl>{level:^8}</lvl>| {file}:{line} | {message}"
 )
 
-logger.add(sys.stdout, format=log_format, backtrace=True, diagnose=True)
+logger.add(get_stdout(), format=log_format, backtrace=True, diagnose=True)

tools/subprocess_utils.py

@@ -2,6 +2,7 @@ import subprocess
 import sys
 
 from .log import logger
+from .stdout_wrapper import get_stdout
 
 python = sys.executable
 
@@ -10,7 +11,7 @@ def run_script_with_log(cmd: list[str]) -> tuple[bool, str]:
     logger.info(f"Running: {' '.join(cmd)}")
     result = subprocess.run(
         [python] + cmd,
-        stdout=sys.stdout,
+        stdout=get_stdout(),  # type: ignore
         stderr=subprocess.PIPE,
         text=True,
     )

train_ms.py

@@ -4,7 +4,6 @@ import gc
 import os
 import platform
 import shutil
-import sys
 
 import torch
 import torch.distributed as dist
@@ -29,6 +28,7 @@ from mel_processing import mel_spectrogram_torch, spec_to_mel_torch
 from models import DurationDiscriminator, MultiPeriodDiscriminator, SynthesizerTrn
 from text.symbols import symbols
 from tools.log import logger
+from tools.stdout_wrapper import get_stdout
 
 torch.backends.cuda.matmul.allow_tf32 = True
 torch.backends.cudnn.allow_tf32 = (
@@ -427,7 +427,7 @@ def train_and_evaluate(
         ja_bert,
         en_bert,
         style_vec,
-    ) in enumerate(tqdm(train_loader, file=sys.stdout)):
+    ) in enumerate(tqdm(train_loader, file=get_stdout())):
         if net_g.module.use_noise_scaled_mas:
             current_mas_noise_scale = (
                 net_g.module.mas_noise_scale_initial

transcribe.py

@@ -5,6 +5,8 @@ import sys
 from faster_whisper import WhisperModel
 from tqdm import tqdm
 
+from tools.stdout_wrapper import get_stdout
+
 
 def transcribe(wav_path, initial_prompt=None):
     segments, _ = model.transcribe(
@@ -45,7 +47,7 @@ if __name__ == "__main__":
         os.rename(output_file, output_file + ".bak")
 
     with open(output_file, "w", encoding="utf-8") as f:
-        for wav_file in tqdm(wav_files, file=sys.stdout):
+        for wav_file in tqdm(wav_files, file=get_stdout()):
             file_name = os.path.basename(wav_file)
             text = transcribe(wav_file, initial_prompt=initial_prompt)
             f.write(f"{file_name}|{speaker_name}|JP|{text}\n")