test

语帆

2024-02-21 62178770dccdbf5da42e831898ea32adeeacba45

test

 funasr/auto/auto_model.py

@@ -209,14 +209,12 @@
        kwargs.update(cfg)
        model = self.model if model is None else model
        model.eval()
        pdb.set_trace()

        batch_size = kwargs.get("batch_size", 1)
        # if kwargs.get("device", "cpu") == "cpu":
        #     batch_size = 1
        
        key_list, data_list = prepare_data_iterator(input, input_len=input_len, data_type=kwargs.get("data_type", None), key=key)
        pdb.set_trace()

        speed_stats = {}
        asr_result_list = []
@@ -225,14 +223,12 @@
        pbar = tqdm(colour="blue", total=num_samples, dynamic_ncols=True) if not disable_pbar else None
        time_speech_total = 0.0
        time_escape_total = 0.0
        pdb.set_trace()
        for beg_idx in range(0, num_samples, batch_size):
            pdb.set_trace()
            end_idx = min(num_samples, beg_idx + batch_size)
            data_batch = data_list[beg_idx:end_idx]
            key_batch = key_list[beg_idx:end_idx]
            batch = {"data_in": data_batch, "key": key_batch}
            pdb.set_trace()

            if (end_idx - beg_idx) == 1 and kwargs.get("data_type", None) == "fbank": # fbank
                batch["data_in"] = data_batch[0]
                batch["data_lengths"] = input_len

 funasr/models/contextual_paraformer/model.py

@@ -102,17 +102,16 @@
            text_lengths = text_lengths[:, 0]
        if len(speech_lengths.size()) > 1:
            speech_lengths = speech_lengths[:, 0]
        pdb.set_trace()

        batch_size = speech.shape[0]

        hotword_pad = kwargs.get("hotword_pad")
        hotword_lengths = kwargs.get("hotword_lengths")
        dha_pad = kwargs.get("dha_pad")
        pdb.set_trace()

        # 1. Encoder
        encoder_out, encoder_out_lens = self.encode(speech, speech_lengths)

        pdb.set_trace()
        loss_ctc, cer_ctc = None, None
        
        stats = dict()
@@ -127,12 +126,11 @@
            stats["loss_ctc"] = loss_ctc.detach() if loss_ctc is not None else None
            stats["cer_ctc"] = cer_ctc
        
        pdb.set_trace()
        # 2b. Attention decoder branch
        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
        )
        pdb.set_trace()

        # 3. CTC-Att loss definition
        if self.ctc_weight == 0.0:
            loss = loss_att + loss_pre * self.predictor_weight
@@ -170,26 +168,24 @@
    ):
        encoder_out_mask = (~make_pad_mask(encoder_out_lens, maxlen=encoder_out.size(1))[:, None, :]).to(
            encoder_out.device)
        pdb.set_trace()

        if self.predictor_bias == 1:
            _, ys_pad = add_sos_eos(ys_pad, self.sos, self.eos, self.ignore_id)
            ys_pad_lens = ys_pad_lens + self.predictor_bias
        pdb.set_trace()

        pre_acoustic_embeds, pre_token_length, _, _ = self.predictor(encoder_out, ys_pad, encoder_out_mask,
                                                                     ignore_id=self.ignore_id)
        pdb.set_trace()
        # -1. bias encoder
        if self.use_decoder_embedding:
            hw_embed = self.decoder.embed(hotword_pad)
        else:
            hw_embed = self.bias_embed(hotword_pad)
        pdb.set_trace()

        hw_embed, (_, _) = self.bias_encoder(hw_embed)
        pdb.set_trace()
        _ind = np.arange(0, hotword_pad.shape[0]).tolist()
        selected = hw_embed[_ind, [i - 1 for i in hotword_lengths.detach().cpu().tolist()]]
        contextual_info = selected.squeeze(0).repeat(ys_pad.shape[0], 1, 1).to(ys_pad.device)
        pdb.set_trace()

        # 0. sampler
        decoder_out_1st = None
        if self.sampling_ratio > 0.0:
@@ -201,7 +197,7 @@
            if self.step_cur < 2:
                logging.info("disable sampler in paraformer, sampling_ratio: {}".format(self.sampling_ratio))
            sematic_embeds = pre_acoustic_embeds
        pdb.set_trace()

        # 1. Forward decoder
        decoder_outs = self.decoder(
            encoder_out, encoder_out_lens, sematic_embeds, ys_pad_lens, contextual_info=contextual_info
@@ -217,7 +213,7 @@
            loss_ideal = None
        '''
        loss_ideal = None
        pdb.set_trace()

        if decoder_out_1st is None:
            decoder_out_1st = decoder_out
        # 2. Compute attention loss
@@ -294,11 +290,11 @@
                                                               enforce_sorted=False)
            _, (h_n, _) = self.bias_encoder(hw_embed)
            hw_embed = h_n.repeat(encoder_out.shape[0], 1, 1)
        pdb.set_trace()

        decoder_outs = self.decoder(
            encoder_out, encoder_out_lens, sematic_embeds, ys_pad_lens, contextual_info=hw_embed, clas_scale=clas_scale
        )
        pdb.set_trace()

        decoder_out = decoder_outs[0]
        decoder_out = torch.log_softmax(decoder_out, dim=-1)
        return decoder_out, ys_pad_lens
@@ -363,14 +359,11 @@
                                                                 clas_scale=kwargs.get("clas_scale", 1.0))
        decoder_out, ys_pad_lens = decoder_outs[0], decoder_outs[1]
        
        pdb.set_trace()
        results = []
        b, n, d = decoder_out.size()
        pdb.set_trace()
        for i in range(b):
            x = encoder_out[i, :encoder_out_lens[i], :]
            am_scores = decoder_out[i, :pre_token_length[i], :]
            pdb.set_trace()
            if self.beam_search is not None:
                nbest_hyps = self.beam_search(
                    x=x, am_scores=am_scores, maxlenratio=kwargs.get("maxlenratio", 0.0),

 funasr/models/seaco_paraformer/model.py

@@ -32,7 +32,7 @@
from funasr.models.transformer.utils.nets_utils import make_pad_mask, pad_list
from funasr.utils.load_utils import load_audio_text_image_video, extract_fbank


import pdb
if LooseVersion(torch.__version__) >= LooseVersion("1.6.0"):
    from torch.cuda.amp import autocast
else:
@@ -212,58 +212,87 @@
                               nfilter=50,
                               seaco_weight=1.0):
        # decoder forward
        pdb.set_trace()
        decoder_out, decoder_hidden, _ = self.decoder(encoder_out, encoder_out_lens, sematic_embeds, ys_pad_lens, return_hidden=True, return_both=True)
        pdb.set_trace()
        decoder_pred = torch.log_softmax(decoder_out, dim=-1)
        if hw_list is not None:
            pdb.set_trace()
            hw_lengths = [len(i) for i in hw_list]
            hw_list_ = [torch.Tensor(i).long() for i in hw_list]
            hw_list_pad = pad_list(hw_list_, 0).to(encoder_out.device)
            pdb.set_trace()
            selected = self._hotword_representation(hw_list_pad, torch.Tensor(hw_lengths).int().to(encoder_out.device))
            pdb.set_trace()
            contextual_info = selected.squeeze(0).repeat(encoder_out.shape[0], 1, 1).to(encoder_out.device)
            pdb.set_trace()
            num_hot_word = contextual_info.shape[1]
            _contextual_length = torch.Tensor([num_hot_word]).int().repeat(encoder_out.shape[0]).to(encoder_out.device)
            
            pdb.set_trace()
            # ASF Core
            if nfilter > 0 and nfilter < num_hot_word:
                for dec in self.seaco_decoder.decoders:
                    dec.reserve_attn = True
                pdb.set_trace()
                # cif_attended, _ = self.decoder2(contextual_info, _contextual_length, sematic_embeds, ys_pad_lens)
                dec_attended, _ = self.seaco_decoder(contextual_info, _contextual_length, decoder_hidden, ys_pad_lens)
                # cif_filter = torch.topk(self.decoder2.decoders[-1].attn_mat[0][0].sum(0).sum(0)[:-1], min(nfilter, num_hot_word-1))[1].tolist()
                pdb.set_trace()
                hotword_scores = self.seaco_decoder.decoders[-1].attn_mat[0][0].sum(0).sum(0)[:-1]
                # hotword_scores /= torch.sqrt(torch.tensor(hw_lengths)[:-1].float()).to(hotword_scores.device)
                pdb.set_trace()
                dec_filter = torch.topk(hotword_scores, min(nfilter, num_hot_word-1))[1].tolist()
                pdb.set_trace()
                add_filter = dec_filter
                pdb.set_trace()
                add_filter.append(len(hw_list_pad)-1)
                # filter hotword embedding
                pdb.set_trace()
                selected = selected[add_filter]
                # again
                pdb.set_trace()
                contextual_info = selected.squeeze(0).repeat(encoder_out.shape[0], 1, 1).to(encoder_out.device)
                pdb.set_trace()
                num_hot_word = contextual_info.shape[1]
                _contextual_length = torch.Tensor([num_hot_word]).int().repeat(encoder_out.shape[0]).to(encoder_out.device)
                pdb.set_trace()
                for dec in self.seaco_decoder.decoders:
                    dec.attn_mat = []
                    dec.reserve_attn = False
            
            pdb.set_trace()
            # SeACo Core
            cif_attended, _ = self.seaco_decoder(contextual_info, _contextual_length, sematic_embeds, ys_pad_lens)
            pdb.set_trace()
            dec_attended, _ = self.seaco_decoder(contextual_info, _contextual_length, decoder_hidden, ys_pad_lens)
            pdb.set_trace()
            merged = self._merge(cif_attended, dec_attended)
            pdb.set_trace()
            
            dha_output = self.hotword_output_layer(merged)  # remove the last token in loss calculation
            pdb.set_trace()
            dha_pred = torch.log_softmax(dha_output, dim=-1)
            pdb.set_trace()
            def _merge_res(dec_output, dha_output):
                pdb.set_trace()
                lmbd = torch.Tensor([seaco_weight] * dha_output.shape[0])
                pdb.set_trace()
                dha_ids = dha_output.max(-1)[-1]# [0]
                pdb.set_trace()
                dha_mask = (dha_ids == 8377).int().unsqueeze(-1)
                pdb.set_trace()
                a = (1 - lmbd) / lmbd
                b = 1 / lmbd
                pdb.set_trace()
                a, b = a.to(dec_output.device), b.to(dec_output.device)
                pdb.set_trace()
                dha_mask = (dha_mask + a.reshape(-1, 1, 1)) / b.reshape(-1, 1, 1)
                # logits = dec_output * dha_mask + dha_output[:,:,:-1] * (1-dha_mask)
                pdb.set_trace()
                logits = dec_output * dha_mask + dha_output[:,:,:] * (1-dha_mask)
                return logits

            merged_pred = _merge_res(decoder_pred, dha_pred)
            pdb.set_trace()
            # import pdb; pdb.set_trace()
            return merged_pred
        else:
@@ -318,7 +347,7 @@
            logging.info("enable beam_search")
            self.init_beam_search(**kwargs)
            self.nbest = kwargs.get("nbest", 1)
        
        pdb.set_trace()
        meta_data = {}
        
        # extract fbank feats
@@ -326,6 +355,7 @@
        audio_sample_list = load_audio_text_image_video(data_in, fs=frontend.fs, audio_fs=kwargs.get("fs", 16000))
        time2 = time.perf_counter()
        meta_data["load_data"] = f"{time2 - time1:0.3f}"
        pdb.set_trace()
        speech, speech_lengths = extract_fbank(audio_sample_list, data_type=kwargs.get("data_type", "sound"),
                                               frontend=frontend)
        time3 = time.perf_counter()
@@ -336,14 +366,18 @@
        speech = speech.to(device=kwargs["device"])
        speech_lengths = speech_lengths.to(device=kwargs["device"])
        
        pdb.set_trace()
        # hotword
        self.hotword_list = self.generate_hotwords_list(kwargs.get("hotword", None), tokenizer=tokenizer, frontend=frontend)
        
        pdb.set_trace()
        # Encoder
        encoder_out, encoder_out_lens = self.encode(speech, speech_lengths)
        if isinstance(encoder_out, tuple):
            encoder_out = encoder_out[0]
        

        pdb.set_trace()
        # predictor
        predictor_outs = self.calc_predictor(encoder_out, encoder_out_lens)
        pre_acoustic_embeds, pre_token_length, _, _ = predictor_outs[0], predictor_outs[1], \
@@ -352,15 +386,16 @@
        if torch.max(pre_token_length) < 1:
            return []


        pdb.set_trace()
        decoder_out = self._seaco_decode_with_ASF(encoder_out, encoder_out_lens,
                                                   pre_acoustic_embeds,
                                                   pre_token_length,
                                                   hw_list=self.hotword_list)
        pdb.set_trace()
        # decoder_out, _ = decoder_outs[0], decoder_outs[1]
        _, _, us_alphas, us_peaks = self.calc_predictor_timestamp(encoder_out, encoder_out_lens,
                                                                  pre_token_length)
        
        pdb.set_trace()
        results = []
        b, n, d = decoder_out.size()
        for i in range(b):