inference

游雁

2023-05-16 0271fbe4fd11114c259ec970b1b04938010a16e3

inference

 funasr/bin/asr_infer.py

@@ -32,6 +32,7 @@
from funasr.modules.beam_search.beam_search import Hypothesis
from funasr.modules.beam_search.beam_search_transducer import BeamSearchTransducer
from funasr.modules.beam_search.beam_search_transducer import Hypothesis as HypothesisTransducer
from funasr.modules.beam_search.beam_search_sa_asr import Hypothesis as HypothesisSAASR
from funasr.modules.scorers.ctc import CTCPrefixScorer
from funasr.modules.scorers.length_bonus import LengthBonus
from funasr.modules.subsampling import TooShortUttError
@@ -58,7 +59,7 @@
from funasr.utils.vad_utils import slice_padding_fbank
from funasr.tasks.vad import VADTask
from funasr.utils.timestamp_tools import time_stamp_sentence, ts_prediction_lfr6_standard

from funasr.tasks.asr import frontend_choices

class Speech2Text:
    """Speech2Text class
@@ -1599,3 +1600,251 @@
        
        return Speech2Text(**kwargs)


class Speech2TextSAASR:
    """Speech2Text class

    Examples:
        >>> import soundfile
        >>> speech2text = Speech2TextSAASR("asr_config.yml", "asr.pb")
        >>> audio, rate = soundfile.read("speech.wav")
        >>> speech2text(audio)
        [(text, token, token_int, hypothesis object), ...]

    """
    
    def __init__(
        self,
        asr_train_config: Union[Path, str] = None,
        asr_model_file: Union[Path, str] = None,
        cmvn_file: Union[Path, str] = None,
        lm_train_config: Union[Path, str] = None,
        lm_file: Union[Path, str] = None,
        token_type: str = None,
        bpemodel: str = None,
        device: str = "cpu",
        maxlenratio: float = 0.0,
        minlenratio: float = 0.0,
        batch_size: int = 1,
        dtype: str = "float32",
        beam_size: int = 20,
        ctc_weight: float = 0.5,
        lm_weight: float = 1.0,
        ngram_weight: float = 0.9,
        penalty: float = 0.0,
        nbest: int = 1,
        streaming: bool = False,
        frontend_conf: dict = None,
        **kwargs,
    ):
        assert check_argument_types()
        
        # 1. Build ASR model
        from funasr.modules.beam_search.beam_search_sa_asr import BeamSearch
        scorers = {}
        asr_model, asr_train_args = ASRTask.build_model_from_file(
            asr_train_config, asr_model_file, cmvn_file, device
        )
        frontend = None
        if asr_train_args.frontend is not None and asr_train_args.frontend_conf is not None:
            if asr_train_args.frontend == 'wav_frontend':
                frontend = WavFrontend(cmvn_file=cmvn_file, **asr_train_args.frontend_conf)
            else:
                frontend_class = frontend_choices.get_class(asr_train_args.frontend)
                frontend = frontend_class(**asr_train_args.frontend_conf).eval()
        
        logging.info("asr_model: {}".format(asr_model))
        logging.info("asr_train_args: {}".format(asr_train_args))
        asr_model.to(dtype=getattr(torch, dtype)).eval()
        
        decoder = asr_model.decoder
        
        ctc = CTCPrefixScorer(ctc=asr_model.ctc, eos=asr_model.eos)
        token_list = asr_model.token_list
        scorers.update(
            decoder=decoder,
            ctc=ctc,
            length_bonus=LengthBonus(len(token_list)),
        )
        
        # 2. Build Language model
        if lm_train_config is not None:
            lm, lm_train_args = LMTask.build_model_from_file(
                lm_train_config, lm_file, None, device
            )
            scorers["lm"] = lm.lm
        
        # 3. Build ngram model
        # ngram is not supported now
        ngram = None
        scorers["ngram"] = ngram
        
        # 4. Build BeamSearch object
        # transducer is not supported now
        beam_search_transducer = None
        
        weights = dict(
            decoder=1.0 - ctc_weight,
            ctc=ctc_weight,
            lm=lm_weight,
            ngram=ngram_weight,
            length_bonus=penalty,
        )
        beam_search = BeamSearch(
            beam_size=beam_size,
            weights=weights,
            scorers=scorers,
            sos=asr_model.sos,
            eos=asr_model.eos,
            vocab_size=len(token_list),
            token_list=token_list,
            pre_beam_score_key=None if ctc_weight == 1.0 else "full",
        )
        
        # 5. [Optional] Build Text converter: e.g. bpe-sym -> Text
        if token_type is None:
            token_type = asr_train_args.token_type
        if bpemodel is None:
            bpemodel = asr_train_args.bpemodel
        
        if token_type is None:
            tokenizer = None
        elif token_type == "bpe":
            if bpemodel is not None:
                tokenizer = build_tokenizer(token_type=token_type, bpemodel=bpemodel)
            else:
                tokenizer = None
        else:
            tokenizer = build_tokenizer(token_type=token_type)
        converter = TokenIDConverter(token_list=token_list)
        logging.info(f"Text tokenizer: {tokenizer}")
        
        self.asr_model = asr_model
        self.asr_train_args = asr_train_args
        self.converter = converter
        self.tokenizer = tokenizer
        self.beam_search = beam_search
        self.beam_search_transducer = beam_search_transducer
        self.maxlenratio = maxlenratio
        self.minlenratio = minlenratio
        self.device = device
        self.dtype = dtype
        self.nbest = nbest
        self.frontend = frontend
    
    @torch.no_grad()
    def __call__(
        self, speech: Union[torch.Tensor, np.ndarray], speech_lengths: Union[torch.Tensor, np.ndarray],
        profile: Union[torch.Tensor, np.ndarray], profile_lengths: Union[torch.Tensor, np.ndarray]
    ) -> List[
        Tuple[
            Optional[str],
            Optional[str],
            List[str],
            List[int],
            Union[HypothesisSAASR],
        ]
    ]:
        """Inference

        Args:
            speech: Input speech data
        Returns:
            text, text_id, token, token_int, hyp

        """
        assert check_argument_types()
        
        # Input as audio signal
        if isinstance(speech, np.ndarray):
            speech = torch.tensor(speech)
        
        if isinstance(profile, np.ndarray):
            profile = torch.tensor(profile)
        
        if self.frontend is not None:
            feats, feats_len = self.frontend.forward(speech, speech_lengths)
            feats = to_device(feats, device=self.device)
            feats_len = feats_len.int()
            self.asr_model.frontend = None
        else:
            feats = speech
            feats_len = speech_lengths
        lfr_factor = max(1, (feats.size()[-1] // 80) - 1)
        batch = {"speech": feats, "speech_lengths": feats_len}
        
        # a. To device
        batch = to_device(batch, device=self.device)
        
        # b. Forward Encoder
        asr_enc, _, spk_enc = self.asr_model.encode(**batch)
        if isinstance(asr_enc, tuple):
            asr_enc = asr_enc[0]
        if isinstance(spk_enc, tuple):
            spk_enc = spk_enc[0]
        assert len(asr_enc) == 1, len(asr_enc)
        assert len(spk_enc) == 1, len(spk_enc)
        
        # c. Passed the encoder result and the beam search
        nbest_hyps = self.beam_search(
            asr_enc[0], spk_enc[0], profile[0], maxlenratio=self.maxlenratio, minlenratio=self.minlenratio
        )
        
        nbest_hyps = nbest_hyps[: self.nbest]
        
        results = []
        for hyp in nbest_hyps:
            assert isinstance(hyp, (HypothesisSAASR)), type(hyp)
            
            # remove sos/eos and get results
            last_pos = -1
            if isinstance(hyp.yseq, list):
                token_int = hyp.yseq[1: last_pos]
            else:
                token_int = hyp.yseq[1: last_pos].tolist()
            
            spk_weigths = torch.stack(hyp.spk_weigths, dim=0)
            
            token_ori = self.converter.ids2tokens(token_int)
            text_ori = self.tokenizer.tokens2text(token_ori)
            
            text_ori_spklist = text_ori.split('$')
            cur_index = 0
            spk_choose = []
            for i in range(len(text_ori_spklist)):
                text_ori_split = text_ori_spklist[i]
                n = len(text_ori_split)
                spk_weights_local = spk_weigths[cur_index: cur_index + n]
                cur_index = cur_index + n + 1
                spk_weights_local = spk_weights_local.mean(dim=0)
                spk_choose_local = spk_weights_local.argmax(-1)
                spk_choose.append(spk_choose_local.item() + 1)
            
            # 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 = self.converter.ids2tokens(token_int)
            
            if self.tokenizer is not None:
                text = self.tokenizer.tokens2text(token)
            else:
                text = None
            
            text_spklist = text.split('$')
            assert len(spk_choose) == len(text_spklist)
            
            spk_list = []
            for i in range(len(text_spklist)):
                text_split = text_spklist[i]
                n = len(text_split)
                spk_list.append(str(spk_choose[i]) * n)
            
            text_id = '$'.join(spk_list)
            
            assert len(text) == len(text_id)
            
            results.append((text, text_id, token, token_int, hyp))
        
        assert check_return_type(results)
        return results

 funasr/bin/asr_inference_launch.py

@@ -77,6 +77,7 @@
from funasr.bin.punc_infer import Text2Punc
from funasr.bin.tp_infer import Speech2Timestamp
from funasr.bin.asr_infer import Speech2TextTransducer
from funasr.bin.asr_infer import Speech2TextSAASR

def inference_asr(
    maxlenratio: float,
@@ -1444,6 +1445,167 @@
    return _forward


def inference_sa_asr(
    maxlenratio: float,
    minlenratio: float,
    batch_size: int,
    beam_size: int,
    ngpu: int,
    ctc_weight: float,
    lm_weight: float,
    penalty: float,
    log_level: Union[int, str],
    # data_path_and_name_and_type,
    asr_train_config: Optional[str],
    asr_model_file: Optional[str],
    cmvn_file: Optional[str] = None,
    lm_train_config: Optional[str] = None,
    lm_file: Optional[str] = None,
    token_type: Optional[str] = None,
    key_file: Optional[str] = None,
    word_lm_train_config: Optional[str] = None,
    bpemodel: Optional[str] = None,
    allow_variable_data_keys: bool = False,
    streaming: bool = False,
    output_dir: Optional[str] = None,
    dtype: str = "float32",
    seed: int = 0,
    ngram_weight: float = 0.9,
    nbest: int = 1,
    num_workers: int = 1,
    mc: bool = False,
    param_dict: dict = None,
    **kwargs,
):
    assert check_argument_types()
    if batch_size > 1:
        raise NotImplementedError("batch decoding is not implemented")
    if word_lm_train_config is not None:
        raise NotImplementedError("Word LM is not implemented")
    if ngpu > 1:
        raise NotImplementedError("only single GPU decoding is supported")
    
    for handler in logging.root.handlers[:]:
        logging.root.removeHandler(handler)
    
    logging.basicConfig(
        level=log_level,
        format="%(asctime)s (%(module)s:%(lineno)d) %(levelname)s: %(message)s",
    )
    
    if ngpu >= 1 and torch.cuda.is_available():
        device = "cuda"
    else:
        device = "cpu"
    
    # 1. Set random-seed
    set_all_random_seed(seed)
    
    # 2. Build speech2text
    speech2text_kwargs = dict(
        asr_train_config=asr_train_config,
        asr_model_file=asr_model_file,
        cmvn_file=cmvn_file,
        lm_train_config=lm_train_config,
        lm_file=lm_file,
        token_type=token_type,
        bpemodel=bpemodel,
        device=device,
        maxlenratio=maxlenratio,
        minlenratio=minlenratio,
        dtype=dtype,
        beam_size=beam_size,
        ctc_weight=ctc_weight,
        lm_weight=lm_weight,
        ngram_weight=ngram_weight,
        penalty=penalty,
        nbest=nbest,
        streaming=streaming,
    )
    logging.info("speech2text_kwargs: {}".format(speech2text_kwargs))
    speech2text = Speech2TextSAASR(**speech2text_kwargs)
    
    def _forward(data_path_and_name_and_type,
                 raw_inputs: Union[np.ndarray, torch.Tensor] = None,
                 output_dir_v2: Optional[str] = None,
                 fs: dict = None,
                 param_dict: dict = None,
                 **kwargs,
                 ):
        # 3. Build data-iterator
        if data_path_and_name_and_type is None and raw_inputs is not None:
            if isinstance(raw_inputs, torch.Tensor):
                raw_inputs = raw_inputs.numpy()
            data_path_and_name_and_type = [raw_inputs, "speech", "waveform"]
        loader = ASRTask.build_streaming_iterator(
            data_path_and_name_and_type,
            dtype=dtype,
            fs=fs,
            mc=mc,
            batch_size=batch_size,
            key_file=key_file,
            num_workers=num_workers,
            preprocess_fn=ASRTask.build_preprocess_fn(speech2text.asr_train_args, False),
            collate_fn=ASRTask.build_collate_fn(speech2text.asr_train_args, False),
            allow_variable_data_keys=allow_variable_data_keys,
            inference=True,
        )
        
        finish_count = 0
        file_count = 1
        # 7 .Start for-loop
        # FIXME(kamo): The output format should be discussed about
        asr_result_list = []
        output_path = output_dir_v2 if output_dir_v2 is not None else output_dir
        if output_path is not None:
            writer = DatadirWriter(output_path)
        else:
            writer = None
        
        for keys, batch in loader:
            assert isinstance(batch, dict), type(batch)
            assert all(isinstance(s, str) for s in keys), keys
            _bs = len(next(iter(batch.values())))
            assert len(keys) == _bs, f"{len(keys)} != {_bs}"
            # batch = {k: v[0] for k, v in batch.items() if not k.endswith("_lengths")}
            # N-best list of (text, token, token_int, hyp_object)
            try:
                results = speech2text(**batch)
            except TooShortUttError as e:
                logging.warning(f"Utterance {keys} {e}")
                hyp = Hypothesis(score=0.0, scores={}, states={}, yseq=[])
                results = [[" ", ["sil"], [2], hyp]] * nbest
            
            # Only supporting batch_size==1
            key = keys[0]
            for n, (text, text_id, token, token_int, hyp) in zip(range(1, nbest + 1), results):
                # Create a directory: outdir/{n}best_recog
                if writer is not None:
                    ibest_writer = writer[f"{n}best_recog"]
                    
                    # Write the result to each file
                    ibest_writer["token"][key] = " ".join(token)
                    ibest_writer["token_int"][key] = " ".join(map(str, token_int))
                    ibest_writer["score"][key] = str(hyp.score)
                    ibest_writer["text_id"][key] = text_id
                
                if text is not None:
                    text_postprocessed, _ = postprocess_utils.sentence_postprocess(token)
                    item = {'key': key, 'value': text_postprocessed}
                    asr_result_list.append(item)
                    finish_count += 1
                    asr_utils.print_progress(finish_count / file_count)
                    if writer is not None:
                        ibest_writer["text"][key] = text
                
                logging.info("uttid: {}".format(key))
                logging.info("text predictions: {}".format(text))
                logging.info("text_id predictions: {}\n".format(text_id))
        return asr_result_list
    
    return _forward


def inference_launch(**kwargs):
    if 'mode' in kwargs:
        mode = kwargs['mode']
@@ -1464,6 +1626,8 @@
        return inference_mfcca(**kwargs)
    elif mode == "rnnt":
        return inference_transducer(**kwargs)
    elif mode == "sa_asr":
        return inference_sa_asr(**kwargs)
    else:
        logging.info("Unknown decoding mode: {}".format(mode))
        return None