update

speech_asr

2023-04-20 d6cc6896e4d55498d6d36331b5c661579906525f

 funasr/utils/build_asr_model.py

@@ -268,7 +268,7 @@
            token_list=token_list,
            **args.model_conf,
        )
    elif args.model == "paraformer":
    elif args.model in ["paraformer", "paraformer_bert", "bicif_paraformer", "contextual_paraformer"]:
        # predictor
        predictor_class = predictor_choices.get_class(args.predictor)
        predictor = predictor_class(**args.predictor_conf)
@@ -336,9 +336,18 @@
            stride_conv=stride_conv,
            **args.model_conf,
        )

    elif args.model == "timestamp_prediction":
        model_class = model_choices.get_class(args.model)
        model = model_class(
            frontend=frontend,
            encoder=encoder,
            token_list=token_list,
            **args.model_conf,
        )
    else:
        raise NotImplementedError("Not supported model: {}".format(args.model))

    if args.init is not None:
        initialize(model, args.init)

    return model