Closed KIKscanf closed 11 months ago
文件preprocess_spec.py 代码： def compute_spec(hps, filename, specname): audio, sampling_rate = utils.load_wav_to_torch(filename) hps.sampling_rate = sampling_rate # 将 hps.sampling_rate 设为音频文件的采样率 assert sampling_rate == hps.sampling_rate, f"{sampling_rate} is not {hps.sampling_rate}" # 这个断言语句永远为真，不会报错 audio_norm = audio / hps.max_wav_value audio_norm = audio_norm.unsqueeze(0) n_fft = hps.filter_length sampling_rate = hps.sampling_rate hop_size = hps.hop_length win_size = hps.win_length spec = spectrogram.spectrogram_torch( audio_norm, n_fft, sampling_rate, hop_size, win_size, center=False) spec = torch.squeeze(spec, 0) torch.save(spec, specname)
tensorboard无法读取了一些信息
KIKscanf
commented
11 months ago KIKscanf
commented
11 months ago
文件preprocess_spec.py 代码： def compute_spec(hps, filename, specname): audio, sampling_rate = utils.load_wav_to_torch(filename) hps.sampling_rate = sampling_rate # 将 hps.sampling_rate 设为音频文件的采样率 assert sampling_rate == hps.sampling_rate, f"{sampling_rate} is not {hps.sampling_rate}" # 这个断言语句永远为真，不会报错 audio_norm = audio / hps.max_wav_value audio_norm = audio_norm.unsqueeze(0) n_fft = hps.filter_length sampling_rate = hps.sampling_rate hop_size = hps.hop_length win_size = hps.win_length spec = spectrogram.spectrogram_torch( audio_norm, n_fft, sampling_rate, hop_size, win_size, center=False) spec = torch.squeeze(spec, 0) torch.save(spec, specname)