benchmarks/rnnt/ootb/inference/pytorch/utils/convert_librispeech.py [35:58]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
                    help='Speed perturbation ratio')
parser.add_argument('--target_sr', type=int, default=None,
                    help='Target sample rate. '
                         'defaults to the input sample rate')
parser.add_argument('--overwrite', action='store_true',
                    help='Overwrite file if exists')
parser.add_argument('--parallel', type=int, default=multiprocessing.cpu_count(),
                    help='Number of threads to use when processing audio files')
args = parser.parse_args()

args.input_dir = args.input_dir.rstrip('/')
args.dest_dir = args.dest_dir.rstrip('/')


def build_input_arr(input_dir):
    txt_files = glob.glob(os.path.join(input_dir, '**', '*.trans.txt'),
                          recursive=True)
    input_data = []
    for txt_file in txt_files:
        rel_path = os.path.relpath(txt_file, input_dir)
        with open(txt_file) as fp:
            for line in fp:
                fname, _, transcript = line.partition(' ')
                input_data.append(dict(input_relpath=os.path.dirname(rel_path),
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -



benchmarks/rnnt/ootb/train/utils/convert_librispeech.py [35:57]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
                    help='Speed perturbation ratio')
parser.add_argument('--target_sr', type=int, default=None,
                    help='Target sample rate. '
                         'defaults to the input sample rate')
parser.add_argument('--overwrite', action='store_true',
                    help='Overwrite file if exists')
parser.add_argument('--parallel', type=int, default=multiprocessing.cpu_count(),
                    help='Number of threads to use when processing audio files')
args = parser.parse_args()

args.input_dir = args.input_dir.rstrip('/')
args.dest_dir = args.dest_dir.rstrip('/')

def build_input_arr(input_dir):
    txt_files = glob.glob(os.path.join(input_dir, '**', '*.trans.txt'),
                          recursive=True)
    input_data = []
    for txt_file in txt_files:
        rel_path = os.path.relpath(txt_file, input_dir)
        with open(txt_file) as fp:
            for line in fp:
                fname, _, transcript = line.partition(' ')
                input_data.append(dict(input_relpath=os.path.dirname(rel_path),
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -