yeyupiaoling / PaddlePaddle-DeepSpeech

基于PaddlePaddle实现的语音识别,中文语音识别。项目完善,识别效果好。支持Windows,Linux下训练和预测,支持Nvidia Jetson开发板预测。
https://yeyupiaoling.blog.csdn.net/article/details/102904306
Apache License 2.0
649 stars 143 forks source link

使用create_data时候对noise数据进行处理时,出现报错未能产生均值归一化方差的npz文件,请问这个可以怎么解决? #151

Closed xuhongtian closed 1 year ago

xuhongtian commented 1 year ago

使用create_data时候对noise数据进行处理时,出现报错未能产生均值归一化方差的npz文件 报错如下: Traceback (most recent call last): File "/data/wav_ocr/paddle_deepspeech/create_data.py", line 223, in main() File "/data/wav_ocr/paddle_deepspeech/create_data.py", line 218, in main compute_mean_std(args.manifest_paths, args.num_samples, args.output_path) File "/data/wav_ocr/paddle_deepspeech/create_data.py", line 183, in compute_mean_std normalizer = FeatureNormalizer(mean_std_filepath=None, File "/data/wav_ocr/paddle_deepspeech/data_utils/normalizer.py", line 41, in init self._compute_mean_std(manifest_path, num_samples, num_workers) File "/data/wav_ocr/paddle_deepspeech/data_utils/normalizer.py", line 94, in _compute_mean_std for i in range(len(means)): TypeError: object of type 'NoneType' has no len()

yeyupiaoling commented 1 year ago

噪声数据是不需要生成归一化文件的。你看你的训练数据列表有没有数据。

xuhongtian commented 1 year ago

刚看了下我的manifest.train为空

xuhongtian commented 1 year ago

刚看了下我的manifest.train为空