Fmt only (maybe)

data_utils.py

@@ -7,7 +7,7 @@ import torch
 import torch.utils.data
 from tqdm import tqdm
 
-from config import config
+from config import get_config
 from mel_processing import mel_spectrogram_torch, spectrogram_torch
 from style_bert_vits2.logging import logger
 from style_bert_vits2.models import commons
@@ -16,6 +16,7 @@ from style_bert_vits2.models.utils import load_filepaths_and_text, load_wav_to_t
 from style_bert_vits2.nlp import cleaned_text_to_sequence
 
 
+config = get_config()
 """Multi speaker version"""
 
 
@@ -120,9 +121,7 @@ class TextAudioSpeakerLoader(torch.utils.data.Dataset):
         audio, sampling_rate = load_wav_to_torch(filename)
         if sampling_rate != self.sampling_rate:
             raise ValueError(
-                "{} {} SR doesn't match target {} SR".format(
-                    filename, sampling_rate, self.sampling_rate
-                )
+                f"{filename} {sampling_rate} SR doesn't match target {self.sampling_rate} SR"
             )
         audio_norm = audio / self.max_wav_value
         audio_norm = audio_norm.unsqueeze(0)