Merge pull request #1825 from modelscope/dev_libt

Shi Xian

2024-06-18 6c467e6f0abfc6d20d0621fbbf67b4dbd81776cc

 runtime/python/onnxruntime/funasr_onnx/punc_bin.py

@@ -7,67 +7,72 @@
from typing import List, Union, Tuple
import numpy as np
import json
from .utils.utils import (ONNXRuntimeError,
                          OrtInferSession, get_logger,
                          read_yaml)
from .utils.utils import (TokenIDConverter, split_to_mini_sentence,code_mix_split_words,code_mix_split_words_jieba)
from .utils.utils import ONNXRuntimeError, OrtInferSession, get_logger, read_yaml
from .utils.utils import (
    TokenIDConverter,
    split_to_mini_sentence,
    code_mix_split_words,
    code_mix_split_words_jieba,
)

logging = get_logger()


class CT_Transformer():
class CT_Transformer:
    """
    Author: Speech Lab of DAMO Academy, Alibaba Group
    CT-Transformer: Controllable time-delay transformer for real-time punctuation prediction and disfluency detection
    https://arxiv.org/pdf/2003.01309.pdf
    """
    def __init__(self, model_dir: Union[str, Path] = None,
                 batch_size: int = 1,
                 device_id: Union[str, int] = "-1",
                 quantize: bool = False,
                 intra_op_num_threads: int = 4,
                 cache_dir: str = None,
                 ):
    

    def __init__(
        self,
        model_dir: Union[str, Path] = None,
        batch_size: int = 1,
        device_id: Union[str, int] = "-1",
        quantize: bool = False,
        intra_op_num_threads: int = 4,
        cache_dir: str = None,
        **kwargs
    ):

        if not Path(model_dir).exists():
            try:
                from modelscope.hub.snapshot_download import snapshot_download
            except:
                raise "You are exporting model from modelscope, please install modelscope and try it again. To install modelscope, you could:\n" \
                      "\npip3 install -U modelscope\n" \
                      "For the users in China, you could install with the command:\n" \
                      "\npip3 install -U modelscope -i https://mirror.sjtu.edu.cn/pypi/web/simple"
                raise "You are exporting model from modelscope, please install modelscope and try it again. To install modelscope, you could:\n" "\npip3 install -U modelscope\n" "For the users in China, you could install with the command:\n" "\npip3 install -U modelscope -i https://mirror.sjtu.edu.cn/pypi/web/simple"
            try:
                model_dir = snapshot_download(model_dir, cache_dir=cache_dir)
            except:
                raise "model_dir must be model_name in modelscope or local path downloaded from modelscope, but is {}".format(
                    model_dir)
    
        model_file = os.path.join(model_dir, 'model.onnx')
                    model_dir
                )

        model_file = os.path.join(model_dir, "model.onnx")
        if quantize:
            model_file = os.path.join(model_dir, 'model_quant.onnx')
            model_file = os.path.join(model_dir, "model_quant.onnx")
        if not os.path.exists(model_file):
            print(".onnx is not exist, begin to export onnx")
            print(".onnx does not exist, begin to export onnx")
            try:
                from funasr import AutoModel
            except:
                raise "You are exporting onnx, please install funasr and try it again. To install funasr, you could:\n" \
                      "\npip3 install -U funasr\n" \
                      "For the users in China, you could install with the command:\n" \
                      "\npip3 install -U funasr -i https://mirror.sjtu.edu.cn/pypi/web/simple"
                raise "You are exporting onnx, please install funasr and try it again. To install funasr, you could:\n" "\npip3 install -U funasr\n" "For the users in China, you could install with the command:\n" "\npip3 install -U funasr -i https://mirror.sjtu.edu.cn/pypi/web/simple"

            model = AutoModel(model=model_dir)
            model_dir = model.export(quantize=quantize)
            
        config_file = os.path.join(model_dir, 'config.yaml')
            model_dir = model.export(type="onnx", quantize=quantize, **kwargs)

        config_file = os.path.join(model_dir, "config.yaml")
        config = read_yaml(config_file)
        token_list = os.path.join(model_dir, 'tokens.json')
        with open(token_list, 'r', encoding='utf-8') as f:
        token_list = os.path.join(model_dir, "tokens.json")
        with open(token_list, "r", encoding="utf-8") as f:
            token_list = json.load(f)

        self.converter = TokenIDConverter(token_list)
        self.ort_infer = OrtInferSession(model_file, device_id, intra_op_num_threads=intra_op_num_threads)
        self.ort_infer = OrtInferSession(
            model_file, device_id, intra_op_num_threads=intra_op_num_threads
        )
        self.batch_size = 1
        self.punc_list = config["model_conf"]['punc_list']
        self.punc_list = config["model_conf"]["punc_list"]
        self.period = 0
        for i in range(len(self.punc_list)):
            if self.punc_list[i] == ",":
@@ -76,7 +81,7 @@
                self.punc_list[i] = "？"
            elif self.punc_list[i] == "。":
                self.period = i
        self.jieba_usr_dict_path = os.path.join(model_dir, 'jieba_usr_dict')
        self.jieba_usr_dict_path = os.path.join(model_dir, "jieba_usr_dict")
        if os.path.exists(self.jieba_usr_dict_path):
            self.seg_jieba = True
            self.code_mix_split_words_jieba = code_mix_split_words_jieba(self.jieba_usr_dict_path)
@@ -101,15 +106,15 @@
            mini_sentence = mini_sentences[mini_sentence_i]
            mini_sentence_id = mini_sentences_id[mini_sentence_i]
            mini_sentence = cache_sent + mini_sentence
            mini_sentence_id = np.array(cache_sent_id + mini_sentence_id, dtype='int32')
            mini_sentence_id = np.array(cache_sent_id + mini_sentence_id, dtype="int32")
            data = {
                "text": mini_sentence_id[None,:],
                "text_lengths": np.array([len(mini_sentence_id)], dtype='int32'),
                "text": mini_sentence_id[None, :],
                "text_lengths": np.array([len(mini_sentence_id)], dtype="int32"),
            }
            try:
                outputs = self.infer(data['text'], data['text_lengths'])
                outputs = self.infer(data["text"], data["text_lengths"])
                y = outputs[0]
                punctuations = np.argmax(y,axis=-1)[0]
                punctuations = np.argmax(y, axis=-1)[0]
                assert punctuations.size == len(mini_sentence)
            except ONNXRuntimeError:
                logging.warning("error")
@@ -119,26 +124,36 @@
                sentenceEnd = -1
                last_comma_index = -1
                for i in range(len(punctuations) - 2, 1, -1):
                    if self.punc_list[punctuations[i]] == "。" or self.punc_list[punctuations[i]] == "？":
                    if (
                        self.punc_list[punctuations[i]] == "。"
                        or self.punc_list[punctuations[i]] == "？"
                    ):
                        sentenceEnd = i
                        break
                    if last_comma_index < 0 and self.punc_list[punctuations[i]] == "，":
                        last_comma_index = i

                if sentenceEnd < 0 and len(mini_sentence) > cache_pop_trigger_limit and last_comma_index >= 0:
                if (
                    sentenceEnd < 0
                    and len(mini_sentence) > cache_pop_trigger_limit
                    and last_comma_index >= 0
                ):
                    # The sentence it too long, cut off at a comma.
                    sentenceEnd = last_comma_index
                    punctuations[sentenceEnd] = self.period
                cache_sent = mini_sentence[sentenceEnd + 1:]
                cache_sent_id = mini_sentence_id[sentenceEnd + 1:].tolist()
                mini_sentence = mini_sentence[0:sentenceEnd + 1]
                punctuations = punctuations[0:sentenceEnd + 1]
                cache_sent = mini_sentence[sentenceEnd + 1 :]
                cache_sent_id = mini_sentence_id[sentenceEnd + 1 :].tolist()
                mini_sentence = mini_sentence[0 : sentenceEnd + 1]
                punctuations = punctuations[0 : sentenceEnd + 1]

            new_mini_sentence_punc += [int(x) for x in punctuations]
            words_with_punc = []
            for i in range(len(mini_sentence)):
                if i > 0:
                    if len(mini_sentence[i][0].encode()) == 1 and len(mini_sentence[i - 1][0].encode()) == 1:
                    if (
                        len(mini_sentence[i][0].encode()) == 1
                        and len(mini_sentence[i - 1][0].encode()) == 1
                    ):
                        mini_sentence[i] = " " + mini_sentence[i]
                words_with_punc.append(mini_sentence[i])
                if self.punc_list[punctuations[i]] != "_":
@@ -156,8 +171,7 @@
                    new_mini_sentence_punc_out = new_mini_sentence_punc[:-1] + [self.period]
        return new_mini_sentence_out, new_mini_sentence_punc_out

    def infer(self, feats: np.ndarray,
              feats_len: np.ndarray) -> Tuple[np.ndarray, np.ndarray]:
    def infer(self, feats: np.ndarray, feats_len: np.ndarray) -> Tuple[np.ndarray, np.ndarray]:
        outputs = self.ort_infer([feats, feats_len])
        return outputs

@@ -168,14 +182,9 @@
    CT-Transformer: Controllable time-delay transformer for real-time punctuation prediction and disfluency detection
    https://arxiv.org/pdf/2003.01309.pdf
    """
    def __init__(self, model_dir: Union[str, Path] = None,
                 batch_size: int = 1,
                 device_id: Union[str, int] = "-1",
                 quantize: bool = False,
                 intra_op_num_threads: int = 4,
                 cache_dir: str = None
                 ):
        super().__init__(model_dir, batch_size, device_id, quantize, intra_op_num_threads, cache_dir=cache_dir)

    def __init__(self, *args, **kwargs):
        super().__init__(*args, **kwargs)

    def __call__(self, text: str, param_dict: map, split_size=20):
        cache_key = "cache"
@@ -195,7 +204,7 @@
        assert len(mini_sentences) == len(mini_sentences_id)

        cache_sent = []
        cache_sent_id = np.array([], dtype='int32')
        cache_sent_id = np.array([], dtype="int32")
        sentence_punc_list = []
        sentence_words_list = []
        cache_pop_trigger_limit = 200
@@ -204,19 +213,23 @@
            mini_sentence = mini_sentences[mini_sentence_i]
            mini_sentence_id = mini_sentences_id[mini_sentence_i]
            mini_sentence = cache_sent + mini_sentence
            mini_sentence_id = np.concatenate((cache_sent_id, mini_sentence_id), axis=0,dtype='int32')
            mini_sentence_id = np.concatenate(
                (cache_sent_id, mini_sentence_id), axis=0, dtype="int32"
            )
            text_length = len(mini_sentence_id)
            vad_mask = self.vad_mask(text_length, len(cache))[None, None, :, :].astype(np.float32)
            data = {
                "input": mini_sentence_id[None,:],
                "text_lengths": np.array([text_length], dtype='int32'),
                "input": mini_sentence_id[None, :],
                "text_lengths": np.array([text_length], dtype="int32"),
                "vad_mask": vad_mask,
                "sub_masks": vad_mask
                "sub_masks": vad_mask,
            }
            try:
                outputs = self.infer(data['input'], data['text_lengths'], data['vad_mask'], data["sub_masks"])
                outputs = self.infer(
                    data["input"], data["text_lengths"], data["vad_mask"], data["sub_masks"]
                )
                y = outputs[0]
                punctuations = np.argmax(y,axis=-1)[0]
                punctuations = np.argmax(y, axis=-1)[0]
                assert punctuations.size == len(mini_sentence)
            except ONNXRuntimeError:
                logging.warning("error")
@@ -226,20 +239,27 @@
                sentenceEnd = -1
                last_comma_index = -1
                for i in range(len(punctuations) - 2, 1, -1):
                    if self.punc_list[punctuations[i]] == "。" or self.punc_list[punctuations[i]] == "？":
                    if (
                        self.punc_list[punctuations[i]] == "。"
                        or self.punc_list[punctuations[i]] == "？"
                    ):
                        sentenceEnd = i
                        break
                    if last_comma_index < 0 and self.punc_list[punctuations[i]] == "，":
                        last_comma_index = i

                if sentenceEnd < 0 and len(mini_sentence) > cache_pop_trigger_limit and last_comma_index >= 0:
                if (
                    sentenceEnd < 0
                    and len(mini_sentence) > cache_pop_trigger_limit
                    and last_comma_index >= 0
                ):
                    # The sentence it too long, cut off at a comma.
                    sentenceEnd = last_comma_index
                    punctuations[sentenceEnd] = self.period
                cache_sent = mini_sentence[sentenceEnd + 1:]
                cache_sent_id = mini_sentence_id[sentenceEnd + 1:]
                mini_sentence = mini_sentence[0:sentenceEnd + 1]
                punctuations = punctuations[0:sentenceEnd + 1]
                cache_sent = mini_sentence[sentenceEnd + 1 :]
                cache_sent_id = mini_sentence_id[sentenceEnd + 1 :]
                mini_sentence = mini_sentence[0 : sentenceEnd + 1]
                punctuations = punctuations[0 : sentenceEnd + 1]

            punctuations_np = [int(x) for x in punctuations]
            new_mini_sentence_punc += punctuations_np
@@ -251,7 +271,10 @@
        sentence_punc_list_out = []
        for i in range(0, len(sentence_words_list)):
            if i > 0:
                if len(sentence_words_list[i][0].encode()) == 1 and len(sentence_words_list[i - 1][-1].encode()) == 1:
                if (
                    len(sentence_words_list[i][0].encode()) == 1
                    and len(sentence_words_list[i - 1][-1].encode()) == 1
                ):
                    sentence_words_list[i] = " " + sentence_words_list[i]
            if skip_num < len(cache):
                skip_num += 1
@@ -268,7 +291,7 @@
            if sentence_punc_list[i] == "。" or sentence_punc_list[i] == "？":
                sentenceEnd = i
                break
        cache_out = sentence_words_list[sentenceEnd + 1:]
        cache_out = sentence_words_list[sentenceEnd + 1 :]
        if sentence_out[-1] in self.punc_list:
            sentence_out = sentence_out[:-1]
            sentence_punc_list_out[-1] = "_"
@@ -286,15 +309,12 @@
        ret = np.ones((size, size), dtype=dtype)
        if vad_pos <= 0 or vad_pos >= size:
            return ret
        sub_corner = np.zeros(
            (vad_pos - 1, size - vad_pos), dtype=dtype)
        ret[0:vad_pos - 1, vad_pos:] = sub_corner
        sub_corner = np.zeros((vad_pos - 1, size - vad_pos), dtype=dtype)
        ret[0 : vad_pos - 1, vad_pos:] = sub_corner
        return ret

    def infer(self, feats: np.ndarray,
              feats_len: np.ndarray,
              vad_mask: np.ndarray,
              sub_masks: np.ndarray) -> Tuple[np.ndarray, np.ndarray]:
    def infer(
        self, feats: np.ndarray, feats_len: np.ndarray, vad_mask: np.ndarray, sub_masks: np.ndarray
    ) -> Tuple[np.ndarray, np.ndarray]:
        outputs = self.ort_infer([feats, feats_len, vad_mask, sub_masks])
        return outputs