update repo

嘉渊

2023-05-17 f5b35ba23deb160e7d1d0d2646adaf3081070f82

update repo

 egs/librispeech/conformer/run.sh

@@ -55,7 +55,7 @@

inference_config=conf/decode_asr_transformer.yaml
#inference_config=conf/decode_asr_transformer_beam60_ctc0.3.yaml
inference_asr_model=valid.acc.ave_10best.pth
inference_asr_model=valid.acc.ave_10best.pb

# you can set gpu num for decoding here
gpuid_list=$CUDA_VISIBLE_DEVICES  # set gpus for decoding, the same as training stage by default

 egs/librispeech_100h/conformer/run.sh

@@ -55,7 +55,7 @@

inference_config=conf/decode_asr_transformer.yaml
#inference_config=conf/decode_asr_transformer_beam60_ctc0.3.yaml
inference_asr_model=valid.acc.ave_10best.pth
inference_asr_model=valid.acc.ave_10best.pb

# you can set gpu num for decoding here
gpuid_list=$CUDA_VISIBLE_DEVICES  # set gpus for decoding, the same as training stage by default

 funasr/utils/prepare_data.py

@@ -207,10 +207,11 @@
    data_names = args.dataset_conf.get("data_names", "speech,text").split(",")
    data_types = args.dataset_conf.get("data_types", "sound,text").split(",")
    file_names = args.data_file_names.split(",")
    print("data_names: {}, data_types: {}, file_names: {}".format(data_names, data_types, file_names))
    assert len(data_names) == len(data_types) == len(file_names)
    if args.dataset_type == "small":
        args.train_shape_file = [os.path.join(args.data_dir, args.train_set, "{}_shape".format(data_names[0]))]
        args.valid_shape_file = [os.path.join(args.data_dir, args.valid_set, "{}}_shape".format(data_names[0]))]
        args.valid_shape_file = [os.path.join(args.data_dir, args.valid_set, "{}_shape".format(data_names[0]))]
        args.train_data_path_and_name_and_type, args.valid_data_path_and_name_and_type = [], []
        for file_name, data_name, data_type in zip(file_names, data_names, data_types):
            args.train_data_path_and_name_and_type.append(

 funasr/utils/prepare_data.py.bak

File was deleted