Fsmn_vad支持多线程并发调用 (#2519)

王梦迪

2025-05-20 fe588bc508c0076bb007d6ed36c18ac8ecb341ac

 fun_text_processing/inverse_text_normalization/export_models.py

@@ -1,32 +1,19 @@
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

import os
from time import perf_counter
from argparse import ArgumentParser
from fun_text_processing.text_normalization.en.graph_utils import generator_main


def parse_args():
    parser = ArgumentParser()
    
    parser.add_argument(
        "--language", help="language", choices=['de', 'en', 'es', 'fr', 'id', 'ja', 'ko', 'pt', 'ru', 'vi', 'zh'], default="en", type=str
    )

    parser.add_argument(
        "--token_and_classify_and_verbalize", help="export the single token&classify and verbalize or combined", choices=['single', 'combine'],
        default="single", type=str
        "--language",
        help="language",
        choices=["de", "en", "es", "fr", "id", "ja", "ko", "pt", "ru", "vi", "zh"],
        default="en",
        type=str,
    )

    parser.add_argument(
        "--export_dir",
        help="path to export directory. Default to current directory.",
@@ -35,46 +22,95 @@
    )
    return parser.parse_args()

def get_grammars(lang: str='en'):
  if lang=='de':
    from fun_text_processing.inverse_text_normalization.de.taggers.tokenize_and_classify import ClassifyFst
    from fun_text_processing.inverse_text_normalization.de.verbalizers.verbalize_final import VerbalizeFinalFst
  elif lang=='en':
    from fun_text_processing.inverse_text_normalization.en.taggers.tokenize_and_classify import ClassifyFst
    from fun_text_processing.inverse_text_normalization.en.verbalizers.verbalize_final import VerbalizeFinalFst
  elif lang=='es':
    from fun_text_processing.inverse_text_normalization.es.taggers.tokenize_and_classify import ClassifyFst
    from fun_text_processing.inverse_text_normalization.es.verbalizers.verbalize_final import VerbalizeFinalFst
  elif lang=='fr':
    from fun_text_processing.inverse_text_normalization.fr.taggers.tokenize_and_classify import ClassifyFst
    from fun_text_processing.inverse_text_normalization.fr.verbalizers.verbalize_final import VerbalizeFinalFst
  elif lang=='id':
    from fun_text_processing.inverse_text_normalization.id.taggers.tokenize_and_classify import ClassifyFst
    from fun_text_processing.inverse_text_normalization.id.verbalizers.verbalize_final import VerbalizeFinalFst
  elif lang=='ja':
    from fun_text_processing.inverse_text_normalization.ja.taggers.tokenize_and_classify import ClassifyFst
    from fun_text_processing.inverse_text_normalization.ja.verbalizers.verbalize_final import VerbalizeFinalFst
  elif lang=='ko':
    from fun_text_processing.inverse_text_normalization.ko.taggers.tokenize_and_classify import ClassifyFst
    from fun_text_processing.inverse_text_normalization.ko.verbalizers.verbalize_final import VerbalizeFinalFst
  elif lang=='pt':
    from fun_text_processing.inverse_text_normalization.pt.taggers.tokenize_and_classify import ClassifyFst
    from fun_text_processing.inverse_text_normalization.pt.verbalizers.verbalize_final import VerbalizeFinalFst
  elif lang=='ru':
    from fun_text_processing.inverse_text_normalization.ru.taggers.tokenize_and_classify import ClassifyFst
    from fun_text_processing.inverse_text_normalization.ru.verbalizers.verbalize_final import VerbalizeFinalFst
  elif lang=='vi':
    from fun_text_processing.inverse_text_normalization.vi.taggers.tokenize_and_classify import ClassifyFst
    from fun_text_processing.inverse_text_normalization.vi.verbalizers.verbalize_final import VerbalizeFinalFst
  elif lang=='zh':
    from fun_text_processing.inverse_text_normalization.zh.taggers.tokenize_and_classify import ClassifyFst
    from fun_text_processing.inverse_text_normalization.zh.verbalizers.verbalize_final import VerbalizeFinalFst
  else:
    from fun_text_processing.inverse_text_normalization.en.taggers.tokenize_and_classify import ClassifyFst
    from fun_text_processing.inverse_text_normalization.en.verbalizers.verbalize_final import VerbalizeFinalFst

def get_grammars(lang: str = "en"):
    if lang == "de":
        from fun_text_processing.inverse_text_normalization.de.taggers.tokenize_and_classify import (
            ClassifyFst,
        )
        from fun_text_processing.inverse_text_normalization.de.verbalizers.verbalize_final import (
            VerbalizeFinalFst,
        )
    elif lang == "en":
        from fun_text_processing.inverse_text_normalization.en.taggers.tokenize_and_classify import (
            ClassifyFst,
        )
        from fun_text_processing.inverse_text_normalization.en.verbalizers.verbalize_final import (
            VerbalizeFinalFst,
        )
    elif lang == "es":
        from fun_text_processing.inverse_text_normalization.es.taggers.tokenize_and_classify import (
            ClassifyFst,
        )
        from fun_text_processing.inverse_text_normalization.es.verbalizers.verbalize_final import (
            VerbalizeFinalFst,
        )
    elif lang == "fr":
        from fun_text_processing.inverse_text_normalization.fr.taggers.tokenize_and_classify import (
            ClassifyFst,
        )
        from fun_text_processing.inverse_text_normalization.fr.verbalizers.verbalize_final import (
            VerbalizeFinalFst,
        )
    elif lang == "id":
        from fun_text_processing.inverse_text_normalization.id.taggers.tokenize_and_classify import (
            ClassifyFst,
        )
        from fun_text_processing.inverse_text_normalization.id.verbalizers.verbalize_final import (
            VerbalizeFinalFst,
        )
    elif lang == "ja":
        from fun_text_processing.inverse_text_normalization.ja.taggers.tokenize_and_classify import (
            ClassifyFst,
        )
        from fun_text_processing.inverse_text_normalization.ja.verbalizers.verbalize_final import (
            VerbalizeFinalFst,
        )
    elif lang == "ko":
        from fun_text_processing.inverse_text_normalization.ko.taggers.tokenize_and_classify import (
            ClassifyFst,
        )
        from fun_text_processing.inverse_text_normalization.ko.verbalizers.verbalize_final import (
            VerbalizeFinalFst,
        )
    elif lang == "pt":
        from fun_text_processing.inverse_text_normalization.pt.taggers.tokenize_and_classify import (
            ClassifyFst,
        )
        from fun_text_processing.inverse_text_normalization.pt.verbalizers.verbalize_final import (
            VerbalizeFinalFst,
        )
    elif lang == "ru":
        from fun_text_processing.inverse_text_normalization.ru.taggers.tokenize_and_classify import (
            ClassifyFst,
        )
        from fun_text_processing.inverse_text_normalization.ru.verbalizers.verbalize_final import (
            VerbalizeFinalFst,
        )
    elif lang == "vi":
        from fun_text_processing.inverse_text_normalization.vi.taggers.tokenize_and_classify import (
            ClassifyFst,
        )
        from fun_text_processing.inverse_text_normalization.vi.verbalizers.verbalize_final import (
            VerbalizeFinalFst,
        )
    elif lang == "zh":
        from fun_text_processing.inverse_text_normalization.zh.taggers.tokenize_and_classify import (
            ClassifyFst,
        )
        from fun_text_processing.inverse_text_normalization.zh.verbalizers.verbalize_final import (
            VerbalizeFinalFst,
        )
    else:
        from fun_text_processing.inverse_text_normalization.en.taggers.tokenize_and_classify import (
            ClassifyFst,
        )
        from fun_text_processing.inverse_text_normalization.en.verbalizers.verbalize_final import (
            VerbalizeFinalFst,
        )

  return ClassifyFst().fst, VerbalizeFinalFst().fst
    return ClassifyFst().fst, VerbalizeFinalFst().fst


if __name__ == "__main__":
    args = parse_args()
@@ -83,17 +119,9 @@
    os.makedirs(export_dir, exist_ok=True)
    tagger_far_file = os.path.join(export_dir, args.language + "_itn_tagger.far")
    verbalizer_far_file = os.path.join(export_dir, args.language + "_itn_verbalizer.far")
    tager_and_verbalizer_far_file = os.path.join(export_dir, args.language, "model.far" )

    start_time = perf_counter()
    tagger_fst, verbalizer_fst = get_grammars(args.language)

    if args.token_and_classify_and_verbalize == 'single':
        generator_main(tagger_far_file, {"tokenize_and_classify": tagger_fst})
        generator_main(verbalizer_far_file, {"verbalize": verbalizer_fst})
    elif args.token_and_classify_and_verbalize == 'combine':
        if not os.path.exists(os.path.join(export_dir, args.language)):
            os.makedirs(os.path.join(export_dir, args.language))
        generator_main(tager_and_verbalizer_far_file, {"tokenize_and_classify": tagger_fst, "verbalize": verbalizer_fst})
    print(f'Time to generate graph: {round(perf_counter() - start_time, 2)} sec')

    generator_main(tagger_far_file, {"tokenize_and_classify": tagger_fst})
    generator_main(verbalizer_far_file, {"verbalize": verbalizer_fst})
    print(f"Time to generate graph: {round(perf_counter() - start_time, 2)} sec")