/models/FireRed.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "0cde0648",
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "\n",
    "import argparse\n",
    "import sys\n",
    "from pathlib import Path\n",
    "\n",
    "import soundfile as sf\n",
    "import sherpa_onnx\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "def create_recognizer(provider=\"cpu\") -> sherpa_onnx.OfflineRecognizer:\n",
    "\n",
    "    model_dir = Path(\"./FireRed/sherpa-onnx-fire-red-asr-large-zh_en-2025-02-16\")\n",
    "    encoder = model_dir / 'encoder.int8.onnx'\n",
    "    decoder = model_dir / 'decoder.int8.onnx'\n",
    "    tokens = model_dir / \"tokens.txt\"\n",
    "    itn = Path(\"./ITN/itn_zh_number.fst\")\n",
    "    \n",
    "    return sherpa_onnx.OfflineRecognizer.from_fire_red_asr(\n",
    "        encoder=encoder.as_posix(),\n",
    "        decoder=decoder.as_posix(),\n",
    "        tokens = tokens.as_posix(),\n",
    "        debug=True,\n",
    "        num_threads=4,\n",
    "        provider=provider,\n",
    "        rule_fsts=itn.as_posix(),\n",
    "    )\n",
    "\n",
    "def decode_file(\n",
    "    recognizer: sherpa_onnx.OfflineRecognizer,\n",
    "    filename: str,\n",
    "):\n",
    "    \"\"\"Decode a single audio file.\"\"\"\n",
    "    audio, sample_rate = sf.read(filename, dtype=\"float32\", always_2d=True)\n",
    "    audio = audio[:, 0]  # only use the first channel\n",
    "\n",
    "    stream = recognizer.create_stream()\n",
    "    stream.accept_waveform(sample_rate, audio)\n",
    "    recognizer.decode_stream(stream)\n",
    "    result = stream.result\n",
    "    return result\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "b89b6afa",
   "metadata": {},
   "outputs": [],
   "source": [
    "recognizer_gpu = create_recognizer('cuda')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "fee418fb",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{\"lang\": \"\", \"emotion\": \"\", \"event\": \"\", \"text\": \"财联社1月7日电港股午间收盘恒生指数跌百分之1点01恒生科技指数跌百分之1点65大型科技股疲软阿里巴巴跌逾百分之4美团网易跌超百分之2\", \"timestamps\": [], \"durations\": [], \"tokens\":[\"财\", \"联\", \"社\", \"一\", \"月\", \"七\", \"日\", \"电\", \"港\", \"股\", \"午\", \"间\", \"收\", \"盘\", \"恒\", \"生\", \"指\", \"数\", \"跌\", \"百\", \"分\", \"之\", \"一\", \"点\", \"零\", \"一\", \"恒\", \"生\", \"科\", \"技\", \"指\", \"数\", \"跌\", \"百\", \"分\", \"之\", \"一\", \"点\", \"六\", \"五\", \"大\", \"型\", \"科\", \"技\", \"股\", \"疲\", \"软\", \"阿\", \"里\", \"巴\", \"巴\", \"跌\", \"逾\", \"百\", \"分\", \"之\", \"四\", \"美\", \"团\", \"网\", \"易\", \"跌\", \"超\", \"百\", \"分\", \"之\", \"二\"], \"ys_log_probs\": [], \"words\": []}\n",
      "财联社1月7日电港股午间收盘恒生指数跌百分之1点01恒生科技指数跌百分之1点65大型科技股疲软阿里巴巴跌逾百分之4美团网易跌超百分之2\n",
      "[]\n",
      "['财', '联', '社', '一', '月', '七', '日', '电', '港', '股', '午', '间', '收', '盘', '恒', '生', '指', '数', '跌', '百', '分', '之', '一', '点', '零', '一', '恒', '生', '科', '技', '指', '数', '跌', '百', '分', '之', '一', '点', '六', '五', '大', '型', '科', '技', '股', '疲', '软', '阿', '里', '巴', '巴', '跌', '逾', '百', '分', '之', '四', '美', '团', '网', '易', '跌', '超', '百', '分', '之', '二']\n"
     ]
    }
   ],
   "source": [
    "\n",
    "wav_path = f\"./demo4.mp3\"\n",
    "result = decode_file(recognizer_gpu, wav_path)\n",
    "print(result)\n",
    "print(result.text)\n",
    "print(result.timestamps)\n",
    "print(result.tokens)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "51339c12",
   "metadata": {},
   "outputs": [],
   "source": [
    "# recognizer_cpu = create_recognizer('cpu')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "65807bd6",
   "metadata": {},
   "outputs": [],
   "source": [
    "# wav_path = f\"../demo.mp3\"\n",
    "# result = decode_file(recognizer_cpu, wav_path)\n",
    "# print(result)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "capswriter",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10.19"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}