mindspore-lab / mindaudio Goto Github PK

Turn a spectrogram from the amplitude/power scale to decibel scale.

Compute the norm of complex number sequence.

Trim an audio signal to keep concecutive non-silent segment.

执行8p训练报错，在刚开始打印训练日志时存在报错信息，但训练可以继续向下执行，并正常打印loss值等信息

1. 执行步骤：mpirun --allow-run-as-root -n 8 python recipes/LJSpeech/tts/wavegrad/train.py --device_target Ascend --is_distributed True --context_mode graph
2. 报错截图：暂时不支持上传截图，会线上发送给开发进行沟通

Create a mel-scaled spectrogram from an audio signal.

Write a numpy array as a WAV file.

Transform stereo audios into mono audio by averaging different channels.

readwav

Split an audio signal into non-silent intervals.

Apply masking to a spectrogram in the frequency domain.

Short-time Fourier transform (STFT).

Performs signal rescaling to a target level.

Resample a signal from one frequency to another. A resample method can be given.

执行单卡训练时报错

1. 执行步骤：python recipes/LJSpeech/tts/wavegrad/train.py --device_target Ascend --device_id 0 --context_mode graph
2. 报错截图：暂时不支持上传截图，会发送给开发进行沟通

Compute the norm of complex number sequence.

Normalize an array along a specified axis.

Update build version from 0.1.0 to 0.1.1

add reverb.

Generate Mel-frequency cepstrum coefficients (MFCC) features from input audio signal.

Convert normal STFT to STFT at the Mel scale

Compute amplitude of a batch of waveforms.

np.ndarray, the time domain signal.

请在readme中补充分布式训练的场景，目前训练部分仅单卡训练的说明

readme关于数据预处理过程的描述
Preprocess data to get a "_wav.npy" and "_feature.npy" for each ".wav" file in your dataset folder. Set your data_path and manifest_path in wavegrad_base.yaml. You can now run the following command:
python recipes/LJSpeech/tts/wavegrad/preprocess.py --device_target CPU --device_id 0

1. 首次做数据处理时遇到报错，和开发沟通后获悉以上说明中的manifest_path为新生成的.csv文件的路径，并非原数据集解压后metadata.csv的路径
2. 建议对该路径添加提示，以免用户使用时造成误解