Spaces:
Runtime error
Runtime error
| # Copyright 2022-2023 Xiaomi Corp. (authors: Fangjun Kuang) | |
| # | |
| # See LICENSE for clarification regarding multiple authors | |
| # | |
| # 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 huggingface_hub import hf_hub_download | |
| os.system( | |
| "cp -v /home/user/.local/lib/python3.8/site-packages/k2/lib/*.so /home/user/.local/lib/python3.8/site-packages/sherpa/lib/" | |
| ) | |
| import sherpa # noqa | |
| def _get_nn_model_filename( | |
| repo_id: str, | |
| filename: str, | |
| subfolder: str = "exp", | |
| ) -> str: | |
| nn_model_filename = hf_hub_download( | |
| repo_id=repo_id, | |
| filename=filename, | |
| subfolder=subfolder, | |
| ) | |
| return nn_model_filename | |
| def _get_token_filename( | |
| repo_id: str, | |
| filename: str = "tokens.txt", | |
| subfolder: str = "data/lang_char", | |
| ) -> str: | |
| token_filename = hf_hub_download( | |
| repo_id=repo_id, | |
| filename=filename, | |
| subfolder=subfolder, | |
| ) | |
| return token_filename | |
| def get_english_model_2022_12_19(repo_id: str): | |
| encoder = _get_nn_model_filename(repo_id=repo_id, filename="encoder_jit_trace.pt") | |
| decoder = _get_nn_model_filename(repo_id=repo_id, filename="decoder_jit_trace.pt") | |
| joiner = _get_nn_model_filename(repo_id=repo_id, filename="joiner_jit_trace.pt") | |
| tokens = _get_token_filename(repo_id=repo_id, subfolder="data/lang_bpe_500") | |
| feat_config = sherpa.FeatureConfig() | |
| feat_config.fbank_opts.frame_opts.samp_freq = 16000 | |
| feat_config.fbank_opts.mel_opts.num_bins = 80 | |
| feat_config.fbank_opts.frame_opts.dither = 0 | |
| config = sherpa.OnlineRecognizerConfig( | |
| nn_model="", | |
| encoder_model=encoder, | |
| decoder_model=decoder, | |
| joiner_model=joiner, | |
| tokens=tokens, | |
| use_gpu=False, | |
| feat_config=feat_config, | |
| decoding_method="greedy_search", | |
| chunk_size=32, | |
| ) | |
| recognizer = sherpa.OnlineRecognizer(config) | |
| return recognizer | |
| def get_chinese_english_mixed_model_conv_emformer_transducer_stateless2_zh( | |
| repo_id: str, | |
| ) -> sherpa.OnlineRecognizer: | |
| nn_model = _get_nn_model_filename( | |
| repo_id=repo_id, filename="cpu_jit-epoch-11-avg-1.pt" | |
| ) | |
| tokens = _get_token_filename(repo_id=repo_id, subfolder="data/lang_char_bpe") | |
| feat_config = sherpa.FeatureConfig() | |
| feat_config.fbank_opts.frame_opts.samp_freq = 16000 | |
| feat_config.fbank_opts.mel_opts.num_bins = 80 | |
| feat_config.fbank_opts.frame_opts.dither = 0 | |
| config = sherpa.OnlineRecognizerConfig( | |
| nn_model=nn_model, | |
| tokens=tokens, | |
| use_gpu=False, | |
| feat_config=feat_config, | |
| decoding_method="greedy_search", | |
| ) | |
| recognizer = sherpa.OnlineRecognizer(config) | |
| return recognizer | |
| def get_chinese_english_mixed_model_k2fsa_zipformer_chinese_english_mixed( | |
| repo_id: str, | |
| ) -> sherpa.OnlineRecognizer: | |
| encoder = _get_nn_model_filename(repo_id=repo_id, filename="encoder_jit_trace.pt") | |
| decoder = _get_nn_model_filename(repo_id=repo_id, filename="decoder_jit_trace.pt") | |
| joiner = _get_nn_model_filename(repo_id=repo_id, filename="joiner_jit_trace.pt") | |
| tokens = _get_token_filename(repo_id=repo_id, subfolder="data/lang_char_bpe") | |
| feat_config = sherpa.FeatureConfig() | |
| feat_config.fbank_opts.frame_opts.samp_freq = 16000 | |
| feat_config.fbank_opts.mel_opts.num_bins = 80 | |
| feat_config.fbank_opts.frame_opts.dither = 0 | |
| config = sherpa.OnlineRecognizerConfig( | |
| nn_model="", | |
| encoder_model=encoder, | |
| decoder_model=decoder, | |
| joiner_model=joiner, | |
| tokens=tokens, | |
| use_gpu=False, | |
| feat_config=feat_config, | |
| decoding_method="greedy_search", | |
| chunk_size=32, | |
| ) | |
| recognizer = sherpa.OnlineRecognizer(config) | |
| return recognizer | |
| def create_recognizer(repo_id: str) -> sherpa.OnlineRecognizer: | |
| if repo_id in english_models: | |
| return english_models[repo_id](repo_id) | |
| elif repo_id in chinese_english_mixed_models: | |
| return chinese_english_mixed_models[repo_id](repo_id) | |
| else: | |
| raise ValueError(f"Unsupported repo_id: {repo_id}") | |
| english_models = { | |
| # https://huggingface.co/Zengwei/icefall-asr-librispeech-pruned-transducer-stateless7-streaming-2022-12-29 | |
| # "Zengwei/icefall-asr-librispeech-pruned-transducer-stateless7-streaming-2022-12-29": get_english_model_2022_12_19 | |
| "desh2608/icefall-asr-librispeech-pruned-transducer-stateless7-streaming-small" : get_english_model_2022_12_19 | |
| } | |
| chinese_english_mixed_models = { | |
| # https://huggingface.co/pfluo/k2fsa-zipformer-chinese-english-mixed | |
| "pfluo/k2fsa-zipformer-chinese-english-mixed": get_chinese_english_mixed_model_k2fsa_zipformer_chinese_english_mixed, | |
| # https://huggingface.co/ptrnull/icefall-asr-conv-emformer-transducer-stateless2-zh | |
| "ptrnull/icefall-asr-conv-emformer-transducer-stateless2-zh": get_chinese_english_mixed_model_conv_emformer_transducer_stateless2_zh, | |
| } | |
| all_models = { | |
| **english_models, | |
| **chinese_english_mixed_models, | |
| } | |
| language_to_models = { | |
| "English": list(english_models.keys()), | |
| "Chinese+English": list(chinese_english_mixed_models.keys()), | |
| } | |