funasr2

游雁

2023-12-06 27f31cd42bb4e20dc19de0034fc0d80b449f1db1

 funasr/models/e2e_asr_contextual_paraformer.py

@@ -134,7 +134,7 @@
            text_lengths: torch.Tensor,
            hotword_pad: torch.Tensor,
            hotword_lengths: torch.Tensor,
            ideal_attn: torch.Tensor,
            dha_pad: torch.Tensor,
    ) -> Tuple[torch.Tensor, Dict[str, torch.Tensor], torch.Tensor]:
        """Frontend + Encoder + Decoder + Calc loss

@@ -207,7 +207,7 @@
        # 2b. Attention decoder branch
        if self.ctc_weight != 1.0:
            loss_att, acc_att, cer_att, wer_att, loss_pre, loss_ideal = self._calc_att_clas_loss(
                encoder_out, encoder_out_lens, text, text_lengths, hotword_pad, hotword_lengths, ideal_attn
                encoder_out, encoder_out_lens, text, text_lengths, hotword_pad, hotword_lengths
            )

        # 3. CTC-Att loss definition
@@ -242,7 +242,6 @@
            ys_pad_lens: torch.Tensor,
            hotword_pad: torch.Tensor,
            hotword_lengths: torch.Tensor,
            ideal_attn: torch.Tensor,
    ):
        encoder_out_mask = (~make_pad_mask(encoder_out_lens, maxlen=encoder_out.size(1))[:, None, :]).to(
            encoder_out.device)
@@ -350,6 +349,7 @@
            else:
                hw_embed = self.bias_embed(hw_list_pad)
            hw_embed, (h_n, _) = self.bias_encoder(hw_embed)
            hw_embed = h_n.repeat(encoder_out.shape[0], 1, 1)
        else:
            hw_lengths = [len(i) for i in hw_list]
            hw_list_pad = pad_list([torch.Tensor(i).long() for i in hw_list], 0).to(encoder_out.device)