#!/usr/bin/env python3 import os import re from pathlib import Path from typing import List BASE_URL = "https://huggingface.co/csukuangfj/sherpa-onnx-apk/resolve/main/" from dataclasses import dataclass @dataclass class APK: major: int minor: int patch: int arch: str short_name: str def __init__(self, s): # sherpa-onnx-1.9.23-arm64-v8a-vad_asr-en-whisper_tiny.apk # sherpa-onnx-1.9.23-x86-vad_asr-en-whisper_tiny.apk s = str(s)[len("vad-asr/") :] split = s.split("-") self.major, self.minor, self.patch = list(map(int, split[2].split("."))) self.arch = split[3] self.lang = split[5] self.short_name = split[6] if "arm" in s: self.arch += "-" + split[4] self.lang = split[6] self.short_name = split[7] if "armeabi" in self.arch: self.arch = "y" + self.arch if "arm64" in self.arch: self.arch = "z" + self.arch if "small" in self.short_name: self.short_name = "zzz" + self.short_name def sort_by_apk(x): x = APK(x) return (x.major, x.minor, x.patch, x.arch, x.lang, x.short_name) def generate_url(files: List[str]) -> List[str]: ans = [] base = BASE_URL for f in files: ans.append(base + str(f)) return ans def get_all_files(d: str, suffix: str) -> List[str]: ans = sorted(Path(d).glob(suffix), key=sort_by_apk, reverse=True) return list(map(lambda x: BASE_URL + str(x), ans)) def to_file(filename: str, files: List[str]): content = r"""

APKs for VAD + non-streaming speech recognition

This page lists the VAD + non-streaming speech recognition APKs for sherpa-onnx, one of the deployment frameworks of the Next-gen Kaldi project.
The name of an APK has the following rule: where
You can download all supported models from https://github.com/k2-fsa/sherpa-onnx/releases/tag/asr-models

APK Comment VAD model Non-streaming ASR model
sherpa-onnx-x.y.z-arm64-v8a-vad_asr-zh-telespeech.apk 支持非常多种中文方言. It is converted from https://github.com/Tele-AI/TeleSpeech-ASR silero_vad.onnx sherpa-onnx-telespeech-ctc-int8-zh-2024-06-04.tar.bz2
sherpa-onnx-x.y.z-arm64-v8a-vad_asr-th-zipformer.apk It supports only Thai. It is converted from https://huggingface.co/yfyeung/icefall-asr-gigaspeech2-th-zipformer-2024-06-20/tree/main silero_vad.onnx sherpa-onnx-zipformer-thai-2024-06-20.tar.bz2
sherpa-onnx-x.y.z-arm64-v8a-vad_asr-ko-zipformer.apk It supports only Korean. It is converted from https://huggingface.co/johnBamma/icefall-asr-ksponspeech-zipformer-2024-06-24 silero_vad.onnx sherpa-onnx-zipformer-korean-2024-06-24.tar.bz2
sherpa-onnx-x.y.z-arm64-v8a-vad_asr-be_de_en_es_fr_hr_it_pl_ru_uk-fast_conformer_ctc_20k.apk It supports 10 languages: Belarusian, German, English, Spanish, French, Croatian, Italian, Polish, Russian, and Ukrainian. It is converted from STT Multilingual FastConformer Hybrid Transducer-CTC Large P&C from NVIDIA/NeMo. Note that only the CTC branch is used. It is trained on ~20000 hours of data. silero_vad.onnx sherpa-onnx-nemo-fast-conformer-transducer-be-de-en-es-fr-hr-it-pl-ru-uk-20k.tar.bz2
sherpa-onnx-x.y.z-arm64-v8a-vad_asr-en_des_es_fr-fast_conformer_ctc_14288.apk It supports 4 languages: German, English, Spanish, and French . It is converted from STT European FastConformer Hybrid Transducer-CTC Large P&C from NVIDIA/NeMo. Note that only the CTC branch is used. It is trained on 14288 hours of data. silero_vad.onnx sherpa-onnx-nemo-fast-conformer-transducer-en-de-es-fr-14288.tar.bz2
sherpa-onnx-x.y.z-arm64-v8a-vad_asr-es-fast_conformer_ctc_1424.apk It supports only Spanish. It is converted from STT Es FastConformer Hybrid Transducer-CTC Large P&C from NVIDIA/NeMo. Note that only the CTC branch is used. It is trained on 1424 hours of data. silero_vad.onnx sherpa-onnx-nemo-fast-conformer-transducer-es-1424.tar.bz2
sherpa-onnx-x.y.z-arm64-v8a-vad_asr-en-fast_conformer_ctc_24500.apk It supports only English. It is converted from STT En FastConformer Hybrid Transducer-CTC Large P&C from NVIDIA/NeMo. Note that only the CTC branch is used. It is trained on 8500 hours of data. silero_vad.onnx sherpa-onnx-nemo-fast-conformer-transducer-en-24500.tar.bz2
sherpa-onnx-x.y.z-arm64-v8a-vad_asr-zh-zipformer.apk It supports only Chinese. silero_vad.onnx icefall-asr-zipformer-wenetspeech-20230615
sherpa-onnx-x.y.z-arm64-v8a-vad_asr-zh-paraformer.apk It supports both Chinese and English. silero_vad.onnx sherpa-onnx-paraformer-zh-2023-03-28
sherpa-onnx-x.y.z-arm64-v8a-vad_asr-en-whisper_tiny.apk It supports only English. silero_vad.onnx sherpa-onnx-whisper-tiny.en


""" if "-cn" not in filename: content += """ For Chinese users, please visit this address, which replaces huggingface.co with hf-mirror.com

中国用户, 请访问这个地址

""" with open(filename, "w") as f: print(content, file=f) for x in files: name = x.rsplit("/", maxsplit=1)[-1] print(f'{name}
', file=f) def main(): apk = get_all_files("vad-asr", suffix="*.apk") to_file("./apk-vad-asr.html", apk) # for Chinese users apk2 = [] for a in apk: a = a.replace("huggingface.co", "hf-mirror.com") a = a.replace("resolve", "blob") apk2.append(a) to_file("./apk-vad-asr-cn.html", apk2) if __name__ == "__main__": main()