update

haoneng.lhn

2023-03-29 d0d8684b964f06ab81279fa11a3725aaff01161c

 funasr/models/predictor/cif.py

@@ -200,6 +200,7 @@
        return acoustic_embeds, token_num, alphas, cif_peak



    def forward_chunk(self, hidden, cache=None):

        b, t, d = hidden.size()

        h = hidden

        context = h.transpose(1, 2)

        queries = self.pad(context)

@@ -220,10 +221,19 @@
            alphas = alphas * mask_chunk_predictor

      

        if cache is not None:

            if cache["is_final"]:

                alphas[:, cache["stride"] + cache["pad_left"] - 1] += 0.45

            if cache["cif_hidden"] is not None:

                hidden = torch.cat((cache["cif_hidden"], hidden), 1)

            if cache["cif_alphas"] is not None:

                alphas = torch.cat((cache["cif_alphas"], alphas), -1)



        #if cache["is_final"]:

        #    tail_threshold = torch.tensor([self.tail_threshold], dtype=alphas.dtype).to(alphas.device)

        #    tail_threshold = torch.reshape(tail_threshold, (1, 1))

        #    alphas = torch.cat([alphas, tail_threshold], dim=1)

        #    zeros_hidden = torch.zeros((b, 1, d), dtype=hidden.dtype).to(hidden.device)

        #    hidden = torch.cat([hidden, zeros_hidden], dim=1)



        token_num = alphas.sum(-1)

        acoustic_embeds, cif_peak = cif(hidden, alphas, self.threshold)

@@ -240,8 +250,9 @@
                pre_alphas_length = cache["cif_alphas"].size(-1)

                mask_chunk_peak_predictor[:, :pre_alphas_length] = 1.0

            mask_chunk_peak_predictor[:, pre_alphas_length + cache["pad_left"]:pre_alphas_length + cache["stride"] + cache["pad_left"]] = 1.0

            #if cache["is_final"]:

            #    mask_chunk_peak_predictor[:, -1] = 1.0

            



        if mask_chunk_peak_predictor is not None:

            cif_peak = cif_peak * mask_chunk_peak_predictor.squeeze(-1)