Dev gzf exp (#1654)

zhifu gao

2024-04-24 861147c7308b91068ffa02724fdf74ee623a909e

 funasr/models/whisper/model.py

@@ -43,10 +43,13 @@
        
        self.encoder_output_size = self.model.dims.n_audio_state
        
    def forward(self, ):
    def forward(
        self,
    ):
        pass
    
    def inference(self,
    def inference(
        self,
                  data_in,
                  data_lengths=None,
                  key: list = None,
@@ -59,13 +62,17 @@

        if frontend is None and not hasattr(self, "frontend"):
            frontend_class = tables.frontend_classes.get("WhisperFrontend")
            frontend = frontend_class(n_mels=self.model.dims.n_mels, do_pad_trim=kwargs.get("do_pad_trim", True))
            frontend = frontend_class(
                n_mels=self.model.dims.n_mels, do_pad_trim=kwargs.get("do_pad_trim", True)
            )
            self.frontend = frontend
        else:
            frontend = frontend if frontend is not None else self.frontend

        meta_data = {}
        if isinstance(data_in, torch.Tensor) and kwargs.get("data_type", "sound") == "fbank":  # fbank
        if (
            isinstance(data_in, torch.Tensor) and kwargs.get("data_type", "sound") == "fbank"
        ):  # fbank
            speech, speech_lengths = data_in, data_lengths
            if len(speech.shape) < 3:
                speech = speech[None, :, :]
@@ -74,13 +81,18 @@
        else:
            # extract fbank feats
            time1 = time.perf_counter()
            audio_sample_list = load_audio_text_image_video(data_in, fs=frontend.fs if hasattr(frontend, "fs") else 16000, audio_fs=kwargs.get("fs", 16000),
            audio_sample_list = load_audio_text_image_video(
                data_in,
                fs=frontend.fs if hasattr(frontend, "fs") else 16000,
                audio_fs=kwargs.get("fs", 16000),
                                                            data_type=kwargs.get("data_type", "sound"),
                                                            tokenizer=tokenizer)
                tokenizer=tokenizer,
            )
            time2 = time.perf_counter()
            meta_data["load_data"] = f"{time2 - time1:0.3f}"
            speech, speech_lengths = extract_fbank(audio_sample_list, data_type=kwargs.get("data_type", "sound"),
                                                   frontend=frontend)
            speech, speech_lengths = extract_fbank(
                audio_sample_list, data_type=kwargs.get("data_type", "sound"), frontend=frontend
            )
            time3 = time.perf_counter()
            meta_data["extract_feat"] = f"{time3 - time2:0.3f}"
            frame_shift = frontend.frame_shift if hasattr(frontend, "frame_shift") else 10
@@ -104,4 +116,3 @@
        results.append(result_i)
    
        return results, meta_data