提交

kongdeqiang

2026-03-13 28ccfbfc51068a663a80764e14074df5edf2b5ba

 funasr/models/uniasr/model.py

@@ -22,6 +22,7 @@
from funasr.utils.load_utils import load_audio_text_image_video, extract_fbank
from funasr.models.scama.utils import sequence_mask


@tables.register("model_classes", "UniASR")
class UniASR(torch.nn.Module):
    """
@@ -56,8 +57,8 @@
        ctc2: str = None,
        ctc2_conf: dict = None,
        ctc2_weight: float = 0.5,
        decoder_attention_chunk_type: str = 'chunk',
        decoder_attention_chunk_type2: str = 'chunk',
        decoder_attention_chunk_type: str = "chunk",
        decoder_attention_chunk_type2: str = "chunk",
        stride_conv=None,
        stride_conv_conf: dict = None,
        loss_weight_model1: float = 0.5,
@@ -71,7 +72,6 @@
        length_normalized_loss: bool = False,
        share_embedding: bool = False,
        **kwargs,
        
    ):
        super().__init__()

@@ -81,7 +81,7 @@
        if normalize is not None:
            normalize_class = tables.normalize_classes.get(normalize)
            normalize = normalize_class(**normalize_conf)
            

        encoder_class = tables.encoder_classes.get(encoder)
        encoder = encoder_class(input_size=input_size, **encoder_conf)
        encoder_output_size = encoder.output_size()
@@ -94,12 +94,14 @@
        )
        predictor_class = tables.predictor_classes.get(predictor)
        predictor = predictor_class(**predictor_conf)
        

        
        from funasr.models.transformer.utils.subsampling import Conv1dSubsampling
        stride_conv = Conv1dSubsampling(**stride_conv_conf, idim=input_size + encoder_output_size,
                                        odim=input_size + encoder_output_size)

        stride_conv = Conv1dSubsampling(
            **stride_conv_conf,
            idim=input_size + encoder_output_size,
            odim=input_size + encoder_output_size,
        )
        stride_conv_output_size = stride_conv.output_size()

        encoder_class = tables.encoder_classes.get(encoder2)
@@ -115,8 +117,6 @@
        predictor_class = tables.predictor_classes.get(predictor2)
        predictor2 = predictor_class(**predictor2_conf)


        
        self.blank_id = blank_id
        self.sos = sos
        self.eos = eos
@@ -127,7 +127,7 @@

        self.specaug = specaug
        self.normalize = normalize
        

        self.encoder = encoder

        self.error_calculator = None
@@ -142,16 +142,20 @@
            smoothing=lsm_weight,
            normalize_length=length_normalized_loss,
        )
        

        self.predictor = predictor
        self.predictor_weight = predictor_weight
        self.criterion_pre = mae_loss(normalize_length=length_normalized_loss)
        self.encoder1_encoder2_joint_training = kwargs.get("encoder1_encoder2_joint_training", True)
        

        if self.encoder.overlap_chunk_cls is not None:
            from funasr.models.scama.chunk_utilis import build_scama_mask_for_cross_attention_decoder
            self.build_scama_mask_for_cross_attention_decoder_fn = build_scama_mask_for_cross_attention_decoder
            from funasr.models.scama.chunk_utilis import (
                build_scama_mask_for_cross_attention_decoder,
            )

            self.build_scama_mask_for_cross_attention_decoder_fn = (
                build_scama_mask_for_cross_attention_decoder
            )
            self.decoder_attention_chunk_type = decoder_attention_chunk_type

        self.encoder2 = encoder2
@@ -164,8 +168,13 @@
        self.stride_conv = stride_conv
        self.loss_weight_model1 = loss_weight_model1
        if self.encoder2.overlap_chunk_cls is not None:
            from funasr.models.scama.chunk_utilis import build_scama_mask_for_cross_attention_decoder
            self.build_scama_mask_for_cross_attention_decoder_fn2 = build_scama_mask_for_cross_attention_decoder
            from funasr.models.scama.chunk_utilis import (
                build_scama_mask_for_cross_attention_decoder,
            )

            self.build_scama_mask_for_cross_attention_decoder_fn2 = (
                build_scama_mask_for_cross_attention_decoder
            )
            self.decoder_attention_chunk_type2 = decoder_attention_chunk_type2

        self.length_normalized_loss = length_normalized_loss
@@ -196,15 +205,15 @@

        batch_size = speech.shape[0]


        ind = self.encoder.overlap_chunk_cls.random_choice(self.training, decoding_ind)
        # 1. Encoder
        if self.enable_maas_finetune:
            with torch.no_grad():
                speech_raw, encoder_out, encoder_out_lens = self.encode(speech, speech_lengths, ind=ind)
                speech_raw, encoder_out, encoder_out_lens = self.encode(
                    speech, speech_lengths, ind=ind
                )
        else:
            speech_raw, encoder_out, encoder_out_lens = self.encode(speech, speech_lengths, ind=ind)


        loss_att, acc_att, cer_att, wer_att = None, None, None, None
        loss_ctc, cer_ctc = None, None
@@ -231,11 +240,10 @@
                    stats["wer"] = wer_att
                    stats["loss_pre"] = loss_pre.detach().cpu() if loss_pre is not None else None
            else:
                

                loss_att, acc_att, cer_att, wer_att, loss_pre = self._calc_att_predictor_loss(
                    encoder_out, encoder_out_lens, text, text_lengths
                )


                loss = loss_att + loss_pre * self.predictor_weight

@@ -254,20 +262,22 @@
            # encoder2
            if self.freeze_encoder2:
                with torch.no_grad():
                    encoder_out, encoder_out_lens = self.encode2(encoder_out, encoder_out_lens, speech_raw, speech_lengths, ind=ind)
                    encoder_out, encoder_out_lens = self.encode2(
                        encoder_out, encoder_out_lens, speech_raw, speech_lengths, ind=ind
                    )
            else:
                encoder_out, encoder_out_lens = self.encode2(encoder_out, encoder_out_lens, speech_raw, speech_lengths, ind=ind)
                encoder_out, encoder_out_lens = self.encode2(
                    encoder_out, encoder_out_lens, speech_raw, speech_lengths, ind=ind
                )

            intermediate_outs = None
            if isinstance(encoder_out, tuple):
                intermediate_outs = encoder_out[1]
                encoder_out = encoder_out[0]


            loss_att, acc_att, cer_att, wer_att, loss_pre = self._calc_att_predictor_loss2(
                encoder_out, encoder_out_lens, text, text_lengths
            )


            loss = loss_att + loss_pre * self.predictor2_weight

@@ -277,7 +287,7 @@
            stats["cer2"] = cer_att
            stats["wer2"] = wer_att
            stats["loss_pre2"] = loss_pre.detach().cpu() if loss_pre is not None else None
        

        loss2 = loss

        loss = loss1 * self.loss_weight_model1 + loss2 * (1 - self.loss_weight_model1)
@@ -287,7 +297,6 @@
        # force_gatherable: to-device and to-tensor if scalar for DataParallel
        if self.length_normalized_loss:
            batch_size = int((text_lengths + 1).sum())


        loss, stats, weight = force_gatherable((loss, stats, batch_size), loss.device)
        return loss, stats, weight
@@ -312,7 +321,10 @@
        return {"feats": feats, "feats_lengths": feats_lengths}

    def encode(
        self, speech: torch.Tensor, speech_lengths: torch.Tensor, **kwargs,
        self,
        speech: torch.Tensor,
        speech_lengths: torch.Tensor,
        **kwargs,
    ):
        """Frontend + Encoder. Note that this method is used by asr_inference.py
        Args:
@@ -324,13 +336,12 @@
            # Data augmentation
            if self.specaug is not None and self.training:
                speech, speech_lengths = self.specaug(speech, speech_lengths)
    

            # Normalization for feature: e.g. Global-CMVN, Utterance-CMVN
            if self.normalize is not None:
                speech, speech_lengths = self.normalize(speech, speech_lengths)
                
        speech_raw = speech.clone().to(speech.device)

        speech_raw = speech.clone().to(speech.device)

        # 4. Forward encoder
        encoder_out, encoder_out_lens, _ = self.encoder(speech, speech_lengths, ind=ind)
@@ -375,9 +386,7 @@
        if isinstance(encoder_out, tuple):
            encoder_out = encoder_out[0]


        return encoder_out, encoder_out_lens


    def nll(
        self,
@@ -472,9 +481,7 @@
        ys_in_lens = ys_pad_lens + 1

        # 1. Forward decoder
        decoder_out, _ = self.decoder(
            encoder_out, encoder_out_lens, ys_in_pad, ys_in_lens
        )
        decoder_out, _ = self.decoder(encoder_out, encoder_out_lens, ys_in_pad, ys_in_lens)

        # 2. Compute attention loss
        loss_att = self.criterion_att(decoder_out, ys_out_pad)
@@ -503,37 +510,49 @@
        ys_in_pad, ys_out_pad = add_sos_eos(ys_pad, self.sos, self.eos, self.ignore_id)
        ys_in_lens = ys_pad_lens + 1

        encoder_out_mask = sequence_mask(encoder_out_lens, maxlen=encoder_out.size(1), dtype=encoder_out.dtype,
                                         device=encoder_out.device)[:, None, :]
        encoder_out_mask = sequence_mask(
            encoder_out_lens,
            maxlen=encoder_out.size(1),
            dtype=encoder_out.dtype,
            device=encoder_out.device,
        )[:, None, :]
        mask_chunk_predictor = None
        if self.encoder.overlap_chunk_cls is not None:
            mask_chunk_predictor = self.encoder.overlap_chunk_cls.get_mask_chunk_predictor(None,
                                                                                           device=encoder_out.device,
                                                                                           batch_size=encoder_out.size(
                                                                                               0))
            mask_shfit_chunk = self.encoder.overlap_chunk_cls.get_mask_shfit_chunk(None, device=encoder_out.device,
                                                                                   batch_size=encoder_out.size(0))
            mask_chunk_predictor = self.encoder.overlap_chunk_cls.get_mask_chunk_predictor(
                None, device=encoder_out.device, batch_size=encoder_out.size(0)
            )
            mask_shfit_chunk = self.encoder.overlap_chunk_cls.get_mask_shfit_chunk(
                None, device=encoder_out.device, batch_size=encoder_out.size(0)
            )
            encoder_out = encoder_out * mask_shfit_chunk
        pre_acoustic_embeds, pre_token_length, pre_alphas, _ = self.predictor(encoder_out,
                                                                              ys_out_pad,
                                                                              encoder_out_mask,
                                                                              ignore_id=self.ignore_id,
                                                                              mask_chunk_predictor=mask_chunk_predictor,
                                                                              target_label_length=ys_in_lens,
                                                                              )
        predictor_alignments, predictor_alignments_len = self.predictor.gen_frame_alignments(pre_alphas,
                                                                                             encoder_out_lens)
        pre_acoustic_embeds, pre_token_length, pre_alphas, _ = self.predictor(
            encoder_out,
            ys_out_pad,
            encoder_out_mask,
            ignore_id=self.ignore_id,
            mask_chunk_predictor=mask_chunk_predictor,
            target_label_length=ys_in_lens,
        )
        predictor_alignments, predictor_alignments_len = self.predictor.gen_frame_alignments(
            pre_alphas, encoder_out_lens
        )

        scama_mask = None
        if self.encoder.overlap_chunk_cls is not None and self.decoder_attention_chunk_type == 'chunk':
        if (
            self.encoder.overlap_chunk_cls is not None
            and self.decoder_attention_chunk_type == "chunk"
        ):
            encoder_chunk_size = self.encoder.overlap_chunk_cls.chunk_size_pad_shift_cur
            attention_chunk_center_bias = 0
            attention_chunk_size = encoder_chunk_size
            decoder_att_look_back_factor = self.encoder.overlap_chunk_cls.decoder_att_look_back_factor_cur
            mask_shift_att_chunk_decoder = self.encoder.overlap_chunk_cls.get_mask_shift_att_chunk_decoder(None,
                                                                                                           device=encoder_out.device,
                                                                                                           batch_size=encoder_out.size(
                                                                                                               0))
            decoder_att_look_back_factor = (
                self.encoder.overlap_chunk_cls.decoder_att_look_back_factor_cur
            )
            mask_shift_att_chunk_decoder = (
                self.encoder.overlap_chunk_cls.get_mask_shift_att_chunk_decoder(
                    None, device=encoder_out.device, batch_size=encoder_out.size(0)
                )
            )
            scama_mask = self.build_scama_mask_for_cross_attention_decoder_fn(
                predictor_alignments=predictor_alignments,
                encoder_sequence_length=encoder_out_lens,
@@ -550,8 +569,9 @@
                is_training=self.training,
            )
        elif self.encoder.overlap_chunk_cls is not None:
            encoder_out, encoder_out_lens = self.encoder.overlap_chunk_cls.remove_chunk(encoder_out, encoder_out_lens,
                                                                                        chunk_outs=None)
            encoder_out, encoder_out_lens = self.encoder.overlap_chunk_cls.remove_chunk(
                encoder_out, encoder_out_lens, chunk_outs=None
            )
        # try:
        # 1. Forward decoder
        decoder_out, _ = self.decoder(
@@ -561,7 +581,6 @@
            ys_in_lens,
            chunk_mask=scama_mask,
            pre_acoustic_embeds=pre_acoustic_embeds,

        )

        # 2. Compute attention loss
@@ -592,37 +611,49 @@
        ys_in_pad, ys_out_pad = add_sos_eos(ys_pad, self.sos, self.eos, self.ignore_id)
        ys_in_lens = ys_pad_lens + 1

        encoder_out_mask = sequence_mask(encoder_out_lens, maxlen=encoder_out.size(1), dtype=encoder_out.dtype,
                                         device=encoder_out.device)[:, None, :]
        encoder_out_mask = sequence_mask(
            encoder_out_lens,
            maxlen=encoder_out.size(1),
            dtype=encoder_out.dtype,
            device=encoder_out.device,
        )[:, None, :]
        mask_chunk_predictor = None
        if self.encoder2.overlap_chunk_cls is not None:
            mask_chunk_predictor = self.encoder2.overlap_chunk_cls.get_mask_chunk_predictor(None,
                                                                                            device=encoder_out.device,
                                                                                            batch_size=encoder_out.size(
                                                                                                0))
            mask_shfit_chunk = self.encoder2.overlap_chunk_cls.get_mask_shfit_chunk(None, device=encoder_out.device,
                                                                                    batch_size=encoder_out.size(0))
            mask_chunk_predictor = self.encoder2.overlap_chunk_cls.get_mask_chunk_predictor(
                None, device=encoder_out.device, batch_size=encoder_out.size(0)
            )
            mask_shfit_chunk = self.encoder2.overlap_chunk_cls.get_mask_shfit_chunk(
                None, device=encoder_out.device, batch_size=encoder_out.size(0)
            )
            encoder_out = encoder_out * mask_shfit_chunk
        pre_acoustic_embeds, pre_token_length, pre_alphas, _ = self.predictor2(encoder_out,
                                                                               ys_out_pad,
                                                                               encoder_out_mask,
                                                                               ignore_id=self.ignore_id,
                                                                               mask_chunk_predictor=mask_chunk_predictor,
                                                                               target_label_length=ys_in_lens,
                                                                               )
        predictor_alignments, predictor_alignments_len = self.predictor2.gen_frame_alignments(pre_alphas,
                                                                                              encoder_out_lens)
        pre_acoustic_embeds, pre_token_length, pre_alphas, _ = self.predictor2(
            encoder_out,
            ys_out_pad,
            encoder_out_mask,
            ignore_id=self.ignore_id,
            mask_chunk_predictor=mask_chunk_predictor,
            target_label_length=ys_in_lens,
        )
        predictor_alignments, predictor_alignments_len = self.predictor2.gen_frame_alignments(
            pre_alphas, encoder_out_lens
        )

        scama_mask = None
        if self.encoder2.overlap_chunk_cls is not None and self.decoder_attention_chunk_type2 == 'chunk':
        if (
            self.encoder2.overlap_chunk_cls is not None
            and self.decoder_attention_chunk_type2 == "chunk"
        ):
            encoder_chunk_size = self.encoder2.overlap_chunk_cls.chunk_size_pad_shift_cur
            attention_chunk_center_bias = 0
            attention_chunk_size = encoder_chunk_size
            decoder_att_look_back_factor = self.encoder2.overlap_chunk_cls.decoder_att_look_back_factor_cur
            mask_shift_att_chunk_decoder = self.encoder2.overlap_chunk_cls.get_mask_shift_att_chunk_decoder(None,
                                                                                                            device=encoder_out.device,
                                                                                                            batch_size=encoder_out.size(
                                                                                                                0))
            decoder_att_look_back_factor = (
                self.encoder2.overlap_chunk_cls.decoder_att_look_back_factor_cur
            )
            mask_shift_att_chunk_decoder = (
                self.encoder2.overlap_chunk_cls.get_mask_shift_att_chunk_decoder(
                    None, device=encoder_out.device, batch_size=encoder_out.size(0)
                )
            )
            scama_mask = self.build_scama_mask_for_cross_attention_decoder_fn2(
                predictor_alignments=predictor_alignments,
                encoder_sequence_length=encoder_out_lens,
@@ -639,8 +670,9 @@
                is_training=self.training,
            )
        elif self.encoder2.overlap_chunk_cls is not None:
            encoder_out, encoder_out_lens = self.encoder2.overlap_chunk_cls.remove_chunk(encoder_out, encoder_out_lens,
                                                                                         chunk_outs=None)
            encoder_out, encoder_out_lens = self.encoder2.overlap_chunk_cls.remove_chunk(
                encoder_out, encoder_out_lens, chunk_outs=None
            )
        # try:
        # 1. Forward decoder
        decoder_out, _ = self.decoder2(
@@ -681,37 +713,49 @@
        # ys_in_lens = ys_pad_lens + 1
        ys_out_pad, ys_in_lens = None, None

        encoder_out_mask = sequence_mask(encoder_out_lens, maxlen=encoder_out.size(1), dtype=encoder_out.dtype,
                                         device=encoder_out.device)[:, None, :]
        encoder_out_mask = sequence_mask(
            encoder_out_lens,
            maxlen=encoder_out.size(1),
            dtype=encoder_out.dtype,
            device=encoder_out.device,
        )[:, None, :]
        mask_chunk_predictor = None
        if self.encoder.overlap_chunk_cls is not None:
            mask_chunk_predictor = self.encoder.overlap_chunk_cls.get_mask_chunk_predictor(None,
                                                                                           device=encoder_out.device,
                                                                                           batch_size=encoder_out.size(
                                                                                               0))
            mask_shfit_chunk = self.encoder.overlap_chunk_cls.get_mask_shfit_chunk(None, device=encoder_out.device,
                                                                                   batch_size=encoder_out.size(0))
            mask_chunk_predictor = self.encoder.overlap_chunk_cls.get_mask_chunk_predictor(
                None, device=encoder_out.device, batch_size=encoder_out.size(0)
            )
            mask_shfit_chunk = self.encoder.overlap_chunk_cls.get_mask_shfit_chunk(
                None, device=encoder_out.device, batch_size=encoder_out.size(0)
            )
            encoder_out = encoder_out * mask_shfit_chunk
        pre_acoustic_embeds, pre_token_length, pre_alphas, _ = self.predictor(encoder_out,
                                                                              ys_out_pad,
                                                                              encoder_out_mask,
                                                                              ignore_id=self.ignore_id,
                                                                              mask_chunk_predictor=mask_chunk_predictor,
                                                                              target_label_length=ys_in_lens,
                                                                              )
        predictor_alignments, predictor_alignments_len = self.predictor.gen_frame_alignments(pre_alphas,
                                                                                             encoder_out_lens)
        pre_acoustic_embeds, pre_token_length, pre_alphas, _ = self.predictor(
            encoder_out,
            ys_out_pad,
            encoder_out_mask,
            ignore_id=self.ignore_id,
            mask_chunk_predictor=mask_chunk_predictor,
            target_label_length=ys_in_lens,
        )
        predictor_alignments, predictor_alignments_len = self.predictor.gen_frame_alignments(
            pre_alphas, encoder_out_lens
        )

        scama_mask = None
        if self.encoder.overlap_chunk_cls is not None and self.decoder_attention_chunk_type == 'chunk':
        if (
            self.encoder.overlap_chunk_cls is not None
            and self.decoder_attention_chunk_type == "chunk"
        ):
            encoder_chunk_size = self.encoder.overlap_chunk_cls.chunk_size_pad_shift_cur
            attention_chunk_center_bias = 0
            attention_chunk_size = encoder_chunk_size
            decoder_att_look_back_factor = self.encoder.overlap_chunk_cls.decoder_att_look_back_factor_cur
            mask_shift_att_chunk_decoder = self.encoder.overlap_chunk_cls.get_mask_shift_att_chunk_decoder(None,
                                                                                                           device=encoder_out.device,
                                                                                                           batch_size=encoder_out.size(
                                                                                                               0))
            decoder_att_look_back_factor = (
                self.encoder.overlap_chunk_cls.decoder_att_look_back_factor_cur
            )
            mask_shift_att_chunk_decoder = (
                self.encoder.overlap_chunk_cls.get_mask_shift_att_chunk_decoder(
                    None, device=encoder_out.device, batch_size=encoder_out.size(0)
                )
            )
            scama_mask = self.build_scama_mask_for_cross_attention_decoder_fn(
                predictor_alignments=predictor_alignments,
                encoder_sequence_length=encoder_out_lens,
@@ -728,10 +772,17 @@
                is_training=self.training,
            )
        elif self.encoder.overlap_chunk_cls is not None:
            encoder_out, encoder_out_lens = self.encoder.overlap_chunk_cls.remove_chunk(encoder_out, encoder_out_lens,
                                                                                        chunk_outs=None)
            encoder_out, encoder_out_lens = self.encoder.overlap_chunk_cls.remove_chunk(
                encoder_out, encoder_out_lens, chunk_outs=None
            )

        return pre_acoustic_embeds, pre_token_length, predictor_alignments, predictor_alignments_len, scama_mask
        return (
            pre_acoustic_embeds,
            pre_token_length,
            predictor_alignments,
            predictor_alignments_len,
            scama_mask,
        )

    def calc_predictor_mask2(
        self,
@@ -744,37 +795,49 @@
        # ys_in_lens = ys_pad_lens + 1
        ys_out_pad, ys_in_lens = None, None

        encoder_out_mask = sequence_mask(encoder_out_lens, maxlen=encoder_out.size(1), dtype=encoder_out.dtype,
                                         device=encoder_out.device)[:, None, :]
        encoder_out_mask = sequence_mask(
            encoder_out_lens,
            maxlen=encoder_out.size(1),
            dtype=encoder_out.dtype,
            device=encoder_out.device,
        )[:, None, :]
        mask_chunk_predictor = None
        if self.encoder2.overlap_chunk_cls is not None:
            mask_chunk_predictor = self.encoder2.overlap_chunk_cls.get_mask_chunk_predictor(None,
                                                                                            device=encoder_out.device,
                                                                                            batch_size=encoder_out.size(
                                                                                                0))
            mask_shfit_chunk = self.encoder2.overlap_chunk_cls.get_mask_shfit_chunk(None, device=encoder_out.device,
                                                                                    batch_size=encoder_out.size(0))
            mask_chunk_predictor = self.encoder2.overlap_chunk_cls.get_mask_chunk_predictor(
                None, device=encoder_out.device, batch_size=encoder_out.size(0)
            )
            mask_shfit_chunk = self.encoder2.overlap_chunk_cls.get_mask_shfit_chunk(
                None, device=encoder_out.device, batch_size=encoder_out.size(0)
            )
            encoder_out = encoder_out * mask_shfit_chunk
        pre_acoustic_embeds, pre_token_length, pre_alphas, _ = self.predictor2(encoder_out,
                                                                               ys_out_pad,
                                                                               encoder_out_mask,
                                                                               ignore_id=self.ignore_id,
                                                                               mask_chunk_predictor=mask_chunk_predictor,
                                                                               target_label_length=ys_in_lens,
                                                                               )
        predictor_alignments, predictor_alignments_len = self.predictor2.gen_frame_alignments(pre_alphas,
                                                                                              encoder_out_lens)
        pre_acoustic_embeds, pre_token_length, pre_alphas, _ = self.predictor2(
            encoder_out,
            ys_out_pad,
            encoder_out_mask,
            ignore_id=self.ignore_id,
            mask_chunk_predictor=mask_chunk_predictor,
            target_label_length=ys_in_lens,
        )
        predictor_alignments, predictor_alignments_len = self.predictor2.gen_frame_alignments(
            pre_alphas, encoder_out_lens
        )

        scama_mask = None
        if self.encoder2.overlap_chunk_cls is not None and self.decoder_attention_chunk_type2 == 'chunk':
        if (
            self.encoder2.overlap_chunk_cls is not None
            and self.decoder_attention_chunk_type2 == "chunk"
        ):
            encoder_chunk_size = self.encoder2.overlap_chunk_cls.chunk_size_pad_shift_cur
            attention_chunk_center_bias = 0
            attention_chunk_size = encoder_chunk_size
            decoder_att_look_back_factor = self.encoder2.overlap_chunk_cls.decoder_att_look_back_factor_cur
            mask_shift_att_chunk_decoder = self.encoder2.overlap_chunk_cls.get_mask_shift_att_chunk_decoder(None,
                                                                                                            device=encoder_out.device,
                                                                                                            batch_size=encoder_out.size(
                                                                                                                0))
            decoder_att_look_back_factor = (
                self.encoder2.overlap_chunk_cls.decoder_att_look_back_factor_cur
            )
            mask_shift_att_chunk_decoder = (
                self.encoder2.overlap_chunk_cls.get_mask_shift_att_chunk_decoder(
                    None, device=encoder_out.device, batch_size=encoder_out.size(0)
                )
            )
            scama_mask = self.build_scama_mask_for_cross_attention_decoder_fn2(
                predictor_alignments=predictor_alignments,
                encoder_sequence_length=encoder_out_lens,
@@ -791,14 +854,22 @@
                is_training=self.training,
            )
        elif self.encoder2.overlap_chunk_cls is not None:
            encoder_out, encoder_out_lens = self.encoder2.overlap_chunk_cls.remove_chunk(encoder_out, encoder_out_lens,
                                                                                         chunk_outs=None)
            encoder_out, encoder_out_lens = self.encoder2.overlap_chunk_cls.remove_chunk(
                encoder_out, encoder_out_lens, chunk_outs=None
            )

        return pre_acoustic_embeds, pre_token_length, predictor_alignments, predictor_alignments_len, scama_mask
        return (
            pre_acoustic_embeds,
            pre_token_length,
            predictor_alignments,
            predictor_alignments_len,
            scama_mask,
        )

    def init_beam_search(self,
                         **kwargs,
                         ):
    def init_beam_search(
        self,
        **kwargs,
    ):
        from funasr.models.uniasr.beam_search import BeamSearchScama
        from funasr.models.transformer.scorers.ctc import CTCPrefixScorer
        from funasr.models.transformer.scorers.length_bonus import LengthBonus
@@ -810,23 +881,21 @@
            decoder = self.decoder2
        # 1. Build ASR model
        scorers = {}
    

        if self.ctc != None:
            ctc = CTCPrefixScorer(ctc=self.ctc, eos=self.eos)
            scorers.update(
                ctc=ctc
            )
            scorers.update(ctc=ctc)
        token_list = kwargs.get("token_list")
        scorers.update(
            decoder=decoder,
            length_bonus=LengthBonus(len(token_list)),
        )
    

        # 3. Build ngram model
        # ngram is not supported now
        ngram = None
        scorers["ngram"] = ngram
    

        weights = dict(
            decoder=1.0 - kwargs.get("decoding_ctc_weight", 0.0),
            ctc=kwargs.get("decoding_ctc_weight", 0.0),
@@ -844,17 +913,18 @@
            token_list=token_list,
            pre_beam_score_key=None if self.ctc_weight == 1.0 else "full",
        )
        

        self.beam_search = beam_search

    def inference(self,
                  data_in,
                  data_lengths=None,
                  key: list = None,
                  tokenizer=None,
                  frontend=None,
                  **kwargs,
                  ):
    def inference(
        self,
        data_in,
        data_lengths=None,
        key: list = None,
        tokenizer=None,
        frontend=None,
        **kwargs,
    ):

        decoding_model = kwargs.get("decoding_model", "normal")
        token_num_relax = kwargs.get("token_num_relax", 5)
@@ -868,14 +938,16 @@
            decoding_ind = 0
            decoding_mode = "model2"
        # init beamsearch
        

        if self.beam_search is None:
            logging.info("enable beam_search")
            self.init_beam_search(decoding_mode=decoding_mode, **kwargs)
            self.nbest = kwargs.get("nbest", 1)
    

        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, :, :]
@@ -884,17 +956,24 @@
        else:
            # extract fbank feats
            time1 = time.perf_counter()
            audio_sample_list = load_audio_text_image_video(data_in, fs=frontend.fs, audio_fs=kwargs.get("fs", 16000),
                                                            data_type=kwargs.get("data_type", "sound"),
                                                            tokenizer=tokenizer)
            audio_sample_list = load_audio_text_image_video(
                data_in,
                fs=frontend.fs,
                audio_fs=kwargs.get("fs", 16000),
                data_type=kwargs.get("data_type", "sound"),
                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}"
            meta_data["batch_data_time"] = speech_lengths.sum().item() * frontend.frame_shift * frontend.lfr_n / 1000
    
            meta_data["batch_data_time"] = (
                speech_lengths.sum().item() * frontend.frame_shift * frontend.lfr_n / 1000
            )

        speech = speech.to(device=kwargs["device"])
        speech_lengths = speech_lengths.to(device=kwargs["device"])
        speech_raw = speech.clone().to(device=kwargs["device"])
@@ -903,9 +982,10 @@
        if decoding_mode == "model1":
            predictor_outs = self.calc_predictor_mask(encoder_out, encoder_out_lens)
        else:
            encoder_out, encoder_out_lens = self.encode2(encoder_out, encoder_out_lens, speech_raw, speech_lengths, ind=decoding_ind)
            encoder_out, encoder_out_lens = self.encode2(
                encoder_out, encoder_out_lens, speech_raw, speech_lengths, ind=decoding_ind
            )
            predictor_outs = self.calc_predictor_mask2(encoder_out, encoder_out_lens)


        scama_mask = predictor_outs[4]
        pre_token_length = predictor_outs[1]
@@ -914,8 +994,13 @@
        minlen = max(0, pre_token_length.sum().item() - token_num_relax)
        # c. Passed the encoder result and the beam search
        nbest_hyps = self.beam_search(
            x=encoder_out[0], scama_mask=scama_mask, pre_acoustic_embeds=pre_acoustic_embeds, maxlenratio=0.0,
            minlenratio=0.0, maxlen=int(maxlen), minlen=int(minlen),
            x=encoder_out[0],
            scama_mask=scama_mask,
            pre_acoustic_embeds=pre_acoustic_embeds,
            maxlenratio=0.0,
            minlenratio=0.0,
            maxlen=int(maxlen),
            minlen=int(minlen),
        )

        nbest_hyps = nbest_hyps[: self.nbest]
@@ -933,15 +1018,13 @@
            # remove blank symbol id, which is assumed to be 0
            token_int = list(filter(lambda x: x != 0, token_int))


            # Change integer-ids to tokens
            token = tokenizer.ids2tokens(token_int)
            text_postprocessed = tokenizer.tokens2text(token)
            if not hasattr(tokenizer, "bpemodel"):
                text_postprocessed, _ = postprocess_utils.sentence_postprocess(token)
    

            result_i = {"key": key[0], "text": text_postprocessed}
            results.append(result_i)

        return results, meta_data
        return results, meta_data