From e7b0f5ed4aa572ba7245e6f6fc66972702bfd502 Mon Sep 17 00:00:00 2001 From: Hui Zhang Date: Tue, 28 Sep 2021 12:17:04 +0000 Subject: [PATCH] reader default type is mat, sound need explicitlyc specify --- utils/format_data.py | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/utils/format_data.py b/utils/format_data.py index e7dcfd23c..682dbfdbc 100755 --- a/utils/format_data.py +++ b/utils/format_data.py @@ -26,7 +26,7 @@ from deepspeech.utils.utility import print_arguments parser = argparse.ArgumentParser(description=__doc__) add_arg = functools.partial(add_arguments, argparser=parser) # yapf: disable -add_arg('feat_type', str, "raw", "speech feature type, e.g. raw(wav, flac), kaldi") +add_arg('feat_type', str, "raw", "speech feature type, e.g. raw(wav, flac), mat(ark), scp") add_arg('cmvn_path', str, 'examples/librispeech/data/mean_std.json', "Filepath of cmvn.") @@ -76,6 +76,7 @@ def main(): assert isinstance(feat_shape, (list, tuple)), type(feat_shape) if args.feat_type == 'raw': feat_shape.append(feat_dim) + line_json['filetype'] = 'sound' else: # kaldi raise NotImplementedError('no support kaldi feat now!') fout.write(json.dumps(line_json) + '\n')