Merge pull request #565 from alibaba-damo-academy/dev_zly2

hnluo

2023-05-31 94bee9029208762468399100654119b7004e769f

Merge pull request #565 from alibaba-damo-academy/dev_zly2

support in_cache

 funasr/bin/vad_infer.py

@@ -175,7 +175,8 @@
        batch_size = speech.shape[0]
        segments = [[]] * batch_size
        if self.frontend is not None:
            feats, feats_len = self.frontend.forward(speech, speech_lengths, is_final)
            reset = in_cache == dict()
            feats, feats_len = self.frontend.forward(speech, speech_lengths, is_final, reset)
            fbanks, _ = self.frontend.get_fbank()
        else:
            raise Exception("Need to extract feats first, please configure frontend configuration")

 funasr/models/e2e_vad.py

@@ -226,7 +226,6 @@
                                               self.vad_opts.frame_in_ms)
        self.encoder = encoder
        # init variables
        self.is_final = False
        self.data_buf_start_frame = 0
        self.frm_cnt = 0
        self.latest_confirmed_speech_frame = 0
@@ -257,7 +256,6 @@
        self.frontend = frontend

    def AllResetDetection(self):
        self.is_final = False
        self.data_buf_start_frame = 0
        self.frm_cnt = 0
        self.latest_confirmed_speech_frame = 0
@@ -473,6 +471,8 @@
    def forward(self, feats: torch.Tensor, waveform: torch.tensor, in_cache: Dict[str, torch.Tensor] = dict(),
                is_final: bool = False
                ) -> Tuple[List[List[List[int]]], Dict[str, torch.Tensor]]:
        if not in_cache:
            self.AllResetDetection()
        self.waveform = waveform  # compute decibel for each frame
        self.ComputeDecibel()
        self.ComputeScores(feats, in_cache)

 funasr/models/frontend/wav_frontend.py

@@ -395,8 +395,10 @@
        return feats_pad, feats_lens, lfr_splice_frame_idxs

    def forward(
            self, input: torch.Tensor, input_lengths: torch.Tensor, is_final: bool = False
        self, input: torch.Tensor, input_lengths: torch.Tensor, is_final: bool = False, reset: bool = False
    ) -> Tuple[torch.Tensor, torch.Tensor]:
        if reset:
            self.cache_reset()
        batch_size = input.shape[0]
        assert batch_size == 1, 'we support to extract feature online only when the batch size is equal to 1 now'
        waveforms, feats, feats_lengths = self.forward_fbank(input, input_lengths)  # input shape: B T D
@@ -500,4 +502,4 @@
        feats_pad = pad_sequence(feats,
                                 batch_first=True,
                                 padding_value=0.0)
        return feats_pad, feats_lens
        return feats_pad, feats_lens