Audio_dataloader

Prepare audio data and the dataloader

1. Data download at https://www.openslr.org/12
   train-clean-100.tar.gz [6.3G] (training set of 100 hours "clean" speech ) Mirrors: [US] [EU] [CN]
   train-clean-360.tar.gz [23G] (training set of 360 hours "clean" speech ) Mirrors: [US] [EU] [CN]
   train-other-500.tar.gz [30G] (training set of 500 hours "other" speech ) Mirrors: [US] [EU] [CN]

2. Run prep_librispeech.py for data preparation.
   a. the path of the data should be changed to your local directory in line 79: librispeech100_path.
   b. the output file name is in line 80: 'librispeech_tr100_cut'.

3. Run run_dataloader to test.
   a. input_json in line 43 should be changed to the directory of the output file generated in step 2.