aishell example

游雁

2024-02-19 94de39dde2e616a01683c518023d0fab72b4e103

 funasr/auto/auto_model.py

@@ -20,7 +20,10 @@
from funasr.utils.load_utils import load_audio_text_image_video, extract_fbank
from funasr.utils.timestamp_tools import timestamp_sentence
from funasr.models.campplus.utils import sv_chunk, postprocess, distribute_spk
from funasr.models.campplus.cluster_backend import ClusterBackend
try:
    from funasr.models.campplus.cluster_backend import ClusterBackend
except:
    print("If you want to use the speaker diarization, please `pip install hdbscan`")


def prepare_data_iterator(data_in, input_len=None, data_type=None, key=None):
@@ -88,7 +91,8 @@
class AutoModel:
    
    def __init__(self, **kwargs):
        tables.print()
        if not kwargs.get("disable_log", False):
            tables.print()
        
        model, kwargs = self.build_model(**kwargs)
        
@@ -120,10 +124,6 @@
            if spk_mode not in ["default", "vad_segment", "punc_segment"]:
                logging.error("spk_mode should be one of default, vad_segment and punc_segment.")
            self.spk_mode = spk_mode
            self.preset_spk_num = kwargs.get("preset_spk_num", None)
            if self.preset_spk_num:
                logging.warning("Using preset speaker number: {}".format(self.preset_spk_num))
            logging.warning("Many to print when using speaker model...")
            
        self.kwargs = kwargs
        self.model = model
@@ -134,8 +134,6 @@
        self.spk_model = spk_model
        self.spk_kwargs = spk_kwargs
        self.model_path = kwargs.get("model_path")

  
        
    def build_model(self, **kwargs):
        assert "model" in kwargs
@@ -146,7 +144,7 @@
        set_all_random_seed(kwargs.get("seed", 0))
        
        device = kwargs.get("device", "cuda")
        if not torch.cuda.is_available() or kwargs.get("ngpu", 0):
        if not torch.cuda.is_available() or kwargs.get("ngpu", 1) == 0:
            device = "cpu"
            kwargs["batch_size"] = 1
        kwargs["device"] = device
@@ -176,7 +174,7 @@
        # build model
        model_class = tables.model_classes.get(kwargs["model"])
        model = model_class(**kwargs, **kwargs["model_conf"], vocab_size=vocab_size)
        model.eval()
        
        model.to(device)
        
        # init_param
@@ -200,8 +198,6 @@
        res = self.model(*args, kwargs)
        return res

        

    def generate(self, input, input_len=None, **cfg):
        if self.vad_model is None:
            return self.inference(input, input_len=input_len, **cfg)
@@ -213,6 +209,7 @@
        kwargs = self.kwargs if kwargs is None else kwargs
        kwargs.update(cfg)
        model = self.model if model is None else model
        model.eval()

        batch_size = kwargs.get("batch_size", 1)
        # if kwargs.get("device", "cpu") == "cpu":
@@ -224,7 +221,7 @@
        asr_result_list = []
        num_samples = len(data_list)
        disable_pbar = kwargs.get("disable_pbar", False)
        pbar = tqdm(colour="blue", total=num_samples+1, dynamic_ncols=True) if not disable_pbar else None
        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
        for beg_idx in range(0, num_samples, batch_size):
@@ -232,7 +229,7 @@
            data_batch = data_list[beg_idx:end_idx]
            key_batch = key_list[beg_idx:end_idx]
            batch = {"data_in": data_batch, "key": key_batch}
            if (end_idx - beg_idx) == 1 and isinstance(data_batch[0], torch.Tensor): # fbank
            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
        
@@ -261,7 +258,7 @@
            time_escape_total += time_escape

        if pbar:
            pbar.update(1)
            # pbar.update(1)
            pbar.set_description(f"rtf_avg: {time_escape_total/time_speech_total:0.3f}")
        torch.cuda.empty_cache()
        return asr_result_list
@@ -286,10 +283,10 @@
        
        key_list, data_list = prepare_data_iterator(input, input_len=input_len, data_type=kwargs.get("data_type", None))
        results_ret_list = []
        time_speech_total_all_samples = 0.0
        time_speech_total_all_samples = 1e-6

        beg_total = time.time()
        pbar_total = tqdm(colour="red", total=len(res) + 1, dynamic_ncols=True)
        pbar_total = tqdm(colour="red", total=len(res), dynamic_ncols=True)
        for i in range(len(res)):
            key = res[i]["key"]
            vadsegments = res[i]["value"]
@@ -314,11 +311,11 @@
            time_speech_total_per_sample = speech_lengths/16000
            time_speech_total_all_samples += time_speech_total_per_sample

            pbar_sample = tqdm(colour="blue", total=n + 1, dynamic_ncols=True)
            # pbar_sample = tqdm(colour="blue", total=n, dynamic_ncols=True)

            all_segments = []
            for j, _ in enumerate(range(0, n)):
                pbar_sample.update(1)
                # pbar_sample.update(1)
                batch_size_ms_cum += (sorted_data[j][0][1] - sorted_data[j][0][0])
                if j < n - 1 and (
                    batch_size_ms_cum + sorted_data[j + 1][0][1] - sorted_data[j + 1][0][0]) < batch_size and (
@@ -329,8 +326,6 @@
                speech_j, speech_lengths_j = slice_padding_audio_samples(speech, speech_lengths, sorted_data[beg_idx:end_idx])       
                results = self.inference(speech_j, input_len=None, model=model, kwargs=kwargs, disable_pbar=True, **cfg)
                if self.spk_model is not None:

                  
                    # compose vad segments: [[start_time_sec, end_time_sec, speech], [...]]
                    for _b in range(len(speech_j)):
                        vad_segments = [[sorted_data[beg_idx:end_idx][_b][0][0]/1000.0,
@@ -339,22 +334,20 @@
                        segments = sv_chunk(vad_segments)
                        all_segments.extend(segments)
                        speech_b = [i[2] for i in segments]
                        spk_res = self.inference(speech_b, input_len=None, model=self.spk_model, kwargs=kwargs, **cfg)
                        spk_res = self.inference(speech_b, input_len=None, model=self.spk_model, kwargs=kwargs, disable_pbar=True, **cfg)
                        results[_b]['spk_embedding'] = spk_res[0]['spk_embedding']
                beg_idx = end_idx
                if len(results) < 1:
                    continue
                results_sorted.extend(results)


            
            end_asr_total = time.time()
            time_escape_total_per_sample = end_asr_total - beg_asr_total
            pbar_sample.set_description(f"rtf_avg_per_sample: {time_escape_total_per_sample / time_speech_total_per_sample:0.3f}, "
                                 f"time_speech_total_per_sample: {time_speech_total_per_sample: 0.3f}, "
                                 f"time_escape_total_per_sample: {time_escape_total_per_sample:0.3f}")
            # end_asr_total = time.time()
            # time_escape_total_per_sample = end_asr_total - beg_asr_total
            # pbar_sample.update(1)
            # pbar_sample.set_description(f"rtf_avg_per_sample: {time_escape_total_per_sample / time_speech_total_per_sample:0.3f}, "
            #                      f"time_speech_total_per_sample: {time_speech_total_per_sample: 0.3f}, "
            #                      f"time_escape_total_per_sample: {time_escape_total_per_sample:0.3f}")
            

            restored_data = [0] * n
            for j in range(n):
                index = sorted_data[j][1]
@@ -377,7 +370,7 @@
                            result[k] = restored_data[j][k]
                        else:
                            result[k] = torch.cat([result[k], restored_data[j][k]], dim=0)
                    elif k == 'text':
                    elif 'text' in k:
                        if k not in result:
                            result[k] = restored_data[j][k]
                        else:
@@ -391,39 +384,51 @@
            # step.3 compute punc model
            if self.punc_model is not None:
                self.punc_kwargs.update(cfg)
                punc_res = self.inference(result["text"], model=self.punc_model, kwargs=self.punc_kwargs, **cfg)
                punc_res = self.inference(result["text"], model=self.punc_model, kwargs=self.punc_kwargs, disable_pbar=True, **cfg)
                import copy; raw_text = copy.copy(result["text"])
                result["text"] = punc_res[0]["text"]
                     
                
            # speaker embedding cluster after resorted
            if self.spk_model is not None:
            if self.spk_model is not None and kwargs.get('return_spk_res', True):
                all_segments = sorted(all_segments, key=lambda x: x[0])
                spk_embedding = result['spk_embedding']
                labels = self.cb_model(spk_embedding.cpu(), oracle_num=self.preset_spk_num)
                del result['spk_embedding']
                labels = self.cb_model(spk_embedding.cpu(), oracle_num=kwargs.get('preset_spk_num', None))
                # del result['spk_embedding']
                sv_output = postprocess(all_segments, None, labels, spk_embedding.cpu())
                if self.spk_mode == 'vad_segment':
                if self.spk_mode == 'vad_segment':  # recover sentence_list
                    sentence_list = []
                    for res, vadsegment in zip(restored_data, vadsegments):
                        sentence_list.append({"start": vadsegment[0],\
                                                "end": vadsegment[1],
                                                "sentence": res['raw_text'],
                                                "timestamp": res['timestamp']})
                else: # punc_segment
                elif self.spk_mode == 'punc_segment':
                    sentence_list = timestamp_sentence(punc_res[0]['punc_array'], \
                                                        result['timestamp'], \
                                                        result['text'])
                                                        result['raw_text'])
                distribute_spk(sentence_list, sv_output)
                result['sentence_info'] = sentence_list
            elif kwargs.get("sentence_timestamp", False):
                sentence_list = timestamp_sentence(punc_res[0]['punc_array'], \
                                                        result['timestamp'], \
                                                        result['raw_text'])
                result['sentence_info'] = sentence_list
            del result['spk_embedding']
                    
            result["key"] = key
            results_ret_list.append(result)
            end_asr_total = time.time()
            time_escape_total_per_sample = end_asr_total - beg_asr_total
            pbar_total.update(1)
            
        pbar_total.update(1)
        end_total = time.time()
        time_escape_total_all_samples = end_total - beg_total
        pbar_total.set_description(f"rtf_avg_all_samples: {time_escape_total_all_samples / time_speech_total_all_samples:0.3f}, "
                             f"time_speech_total_all_samples: {time_speech_total_all_samples: 0.3f}, "
                             f"time_escape_total_all_samples: {time_escape_total_all_samples:0.3f}")
            pbar_total.set_description(f"rtf_avg: {time_escape_total_per_sample / time_speech_total_per_sample:0.3f}, "
                                 f"time_speech: {time_speech_total_per_sample: 0.3f}, "
                                 f"time_escape: {time_escape_total_per_sample:0.3f}")


        # end_total = time.time()
        # time_escape_total_all_samples = end_total - beg_total
        # print(f"rtf_avg_all: {time_escape_total_all_samples / time_speech_total_all_samples:0.3f}, "
        #                      f"time_speech_all: {time_speech_total_all_samples: 0.3f}, "
        #                      f"time_escape_all: {time_escape_total_all_samples:0.3f}")
        return results_ret_list