PaddleSpeech/.notebook/jit_infer.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "/home/ssd5/zhanghui/DeepSpeech2.x\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "'/home/ssd5/zhanghui/DeepSpeech2.x'"
      ]
     },
     "execution_count": 1,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "%cd ..\n",
    "%pwd"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-03-26 02:55:23,873 - WARNING - register user softmax to paddle, remove this when fixed!\n",
      "2021-03-26 02:55:23,875 - WARNING - register user sigmoid to paddle, remove this when fixed!\n",
      "2021-03-26 02:55:23,875 - WARNING - register user relu to paddle, remove this when fixed!\n",
      "2021-03-26 02:55:23,876 - WARNING - override cat of paddle if exists or register, remove this when fixed!\n",
      "2021-03-26 02:55:23,876 - WARNING - override eq of paddle.Tensor if exists or register, remove this when fixed!\n",
      "2021-03-26 02:55:23,877 - WARNING - override contiguous of paddle.Tensor if exists or register, remove this when fixed!\n",
      "2021-03-26 02:55:23,877 - WARNING - override size of paddle.Tensor (`to_static` do not process `size` property, maybe some `paddle` api dependent on it), remove this when fixed!\n",
      "2021-03-26 02:55:23,878 - WARNING - register user view to paddle.Tensor, remove this when fixed!\n",
      "2021-03-26 02:55:23,878 - WARNING - register user view_as to paddle.Tensor, remove this when fixed!\n",
      "2021-03-26 02:55:23,879 - WARNING - register user masked_fill to paddle.Tensor, remove this when fixed!\n",
      "2021-03-26 02:55:23,880 - WARNING - register user masked_fill_ to paddle.Tensor, remove this when fixed!\n",
      "2021-03-26 02:55:23,880 - WARNING - register user fill_ to paddle.Tensor, remove this when fixed!\n",
      "2021-03-26 02:55:23,881 - WARNING - register user repeat to paddle.Tensor, remove this when fixed!\n",
      "2021-03-26 02:55:23,881 - WARNING - register user softmax to paddle.Tensor, remove this when fixed!\n",
      "2021-03-26 02:55:23,882 - WARNING - register user sigmoid to paddle.Tensor, remove this when fixed!\n",
      "2021-03-26 02:55:23,882 - WARNING - register user relu to paddle.Tensor, remove this when fixed!\n",
      "2021-03-26 02:55:23,883 - WARNING - register user glu to paddle.nn.functional, remove this when fixed!\n",
      "2021-03-26 02:55:23,883 - WARNING - override ctc_loss of paddle.nn.functional if exists, remove this when fixed!\n",
      "2021-03-26 02:55:23,884 - WARNING - register user GLU to paddle.nn, remove this when fixed!\n",
      "2021-03-26 02:55:23,884 - WARNING - register user ConstantPad2d to paddle.nn, remove this when fixed!\n",
      "/home/ssd5/zhanghui/DeepSpeech2.x/tools/venv-dev/lib/python3.7/site-packages/scipy/fftpack/__init__.py:103: DeprecationWarning: The module numpy.dual is deprecated.  Instead of using dual, use the functions directly from numpy or scipy.\n",
      "  from numpy.dual import register_func\n",
      "/home/ssd5/zhanghui/DeepSpeech2.x/tools/venv-dev/lib/python3.7/site-packages/scipy/special/orthogonal.py:81: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  from numpy import (exp, inf, pi, sqrt, floor, sin, cos, around, int,\n"
     ]
    }
   ],
   "source": [
    "import os\n",
    "import time\n",
    "import argparse\n",
    "import functools\n",
    "import paddle\n",
    "import numpy as np\n",
    "\n",
    "from deepspeech.utils.socket_server import warm_up_test\n",
    "from deepspeech.utils.socket_server import AsrTCPServer\n",
    "from deepspeech.utils.socket_server import AsrRequestHandler\n",
    "\n",
    "from deepspeech.training.cli import default_argument_parser\n",
    "from deepspeech.exps.deepspeech2.config import get_cfg_defaults\n",
    "\n",
    "from deepspeech.frontend.utility import read_manifest\n",
    "from deepspeech.utils.utility import add_arguments, print_arguments\n",
    "\n",
    "from deepspeech.models.deepspeech2 import DeepSpeech2Model\n",
    "from deepspeech.models.deepspeech2 import DeepSpeech2InferModel\n",
    "from deepspeech.io.dataset import ManifestDataset\n",
    "\n",
    "\n",
    "\n",
    "from deepspeech.frontend.utility import read_manifest"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.0.0\n",
      "e7f28d6c0db54eb9c9a810612300b526687e56a6\n",
      "OFF\n",
      "OFF\n",
      "commit: e7f28d6c0db54eb9c9a810612300b526687e56a6\n",
      "None\n",
      "0\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/ssd5/zhanghui/DeepSpeech2.x/tools/venv-dev/lib/python3.7/site-packages/ipykernel/ipkernel.py:283: DeprecationWarning: `should_run_async` will not call `transform_cell` automatically in the future. Please pass the result to `transformed_cell` argument and any exception that happen during thetransform in `preprocessing_exc_tuple` in IPython 7.17 and above.\n",
      "  and should_run_async(code)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "['__builtins__',\n",
       " '__cached__',\n",
       " '__doc__',\n",
       " '__file__',\n",
       " '__loader__',\n",
       " '__name__',\n",
       " '__package__',\n",
       " '__spec__',\n",
       " 'commit',\n",
       " 'full_version',\n",
       " 'istaged',\n",
       " 'major',\n",
       " 'minor',\n",
       " 'mkl',\n",
       " 'patch',\n",
       " 'rc',\n",
       " 'show',\n",
       " 'with_mkl']"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print(paddle.__version__)\n",
    "print(paddle.version.commit)\n",
    "print(paddle.version.with_mkl)\n",
    "print(paddle.version.mkl())\n",
    "print(paddle.version.show())\n",
    "print(paddle.version.patch)\n",
    "dir(paddle.version)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "data:\n",
      "  augmentation_config: conf/augmentation.config\n",
      "  batch_size: 64\n",
      "  dev_manifest: data/manifest.dev\n",
      "  keep_transcription_text: False\n",
      "  max_duration: 27.0\n",
      "  max_freq: None\n",
      "  mean_std_filepath: examples/aishell/data/mean_std.npz\n",
      "  min_duration: 0.0\n",
      "  n_fft: None\n",
      "  num_workers: 0\n",
      "  random_seed: 0\n",
      "  shuffle_method: batch_shuffle\n",
      "  sortagrad: True\n",
      "  specgram_type: linear\n",
      "  stride_ms: 10.0\n",
      "  target_dB: -20\n",
      "  target_sample_rate: 16000\n",
      "  test_manifest: examples/aishell/data/manifest.test\n",
      "  train_manifest: data/manifest.train\n",
      "  use_dB_normalization: True\n",
      "  vocab_filepath: examples/aishell/data/vocab.txt\n",
      "  window_ms: 20.0\n",
      "decoding:\n",
      "  alpha: 2.6\n",
      "  batch_size: 128\n",
      "  beam_size: 300\n",
      "  beta: 5.0\n",
      "  cutoff_prob: 0.99\n",
      "  cutoff_top_n: 40\n",
      "  decoding_method: ctc_beam_search\n",
      "  error_rate_type: cer\n",
      "  lang_model_path: data/lm/zh_giga.no_cna_cmn.prune01244.klm\n",
      "  num_proc_bsearch: 10\n",
      "model:\n",
      "  num_conv_layers: 2\n",
      "  num_rnn_layers: 3\n",
      "  rnn_layer_size: 1024\n",
      "  share_rnn_weights: False\n",
      "  use_gru: True\n",
      "training:\n",
      "  global_grad_clip: 5.0\n",
      "  lr: 0.0005\n",
      "  lr_decay: 0.83\n",
      "  n_epoch: 30\n",
      "  weight_decay: 1e-06\n",
      "-----------  Configuration Arguments -----------\n",
      "checkpoint_path: examples/aishell/ckpt-loss2e-3-0.83-5/checkpoints/step-11725\n",
      "config: examples/aishell/conf/deepspeech2.yaml\n",
      "device: gpu\n",
      "dump_config: None\n",
      "export_path: None\n",
      "host_ip: localhost\n",
      "host_port: 8086\n",
      "model_dir: None\n",
      "model_file: examples/aishell/jit.model.pdmodel\n",
      "nprocs: 1\n",
      "opts: ['data.test_manifest', 'examples/aishell/data/manifest.test', 'data.mean_std_filepath', 'examples/aishell/data/mean_std.npz', 'data.vocab_filepath', 'examples/aishell/data/vocab.txt']\n",
      "output: None\n",
      "params_file: examples/aishell/jit.model.pdiparams\n",
      "speech_save_dir: demo_cache\n",
      "use_gpu: False\n",
      "warmup_manifest: examples/aishell/data/manifest.test\n",
      "------------------------------------------------\n"
     ]
    }
   ],
   "source": [
    "parser = default_argument_parser()\n",
    "add_arg = functools.partial(add_arguments, argparser=parser)\n",
    "add_arg('host_ip',          str,\n",
    "        'localhost',\n",
    "        \"Server's IP address.\")\n",
    "add_arg('host_port',        int,    8086,    \"Server's IP port.\")\n",
    "add_arg('speech_save_dir',  str,\n",
    "        'demo_cache',\n",
    "        \"Directory to save demo audios.\")\n",
    "add_arg('warmup_manifest',  \n",
    "        str, \n",
    "        \"examples/aishell/data/manifest.test\", \n",
    "        \"Filepath of manifest to warm up.\")\n",
    "add_arg(\n",
    "    \"--model_file\",\n",
    "    type=str,\n",
    "    default=\"examples/aishell/jit.model.pdmodel\",\n",
    "    help=\"Model filename, Specify this when your model is a combined model.\"\n",
    ")\n",
    "add_arg(\n",
    "    \"--params_file\",\n",
    "    type=str,\n",
    "    default=\"examples/aishell/jit.model.pdiparams\",\n",
    "    help=\n",
    "    \"Parameter filename, Specify this when your model is a combined model.\"\n",
    ")\n",
    "add_arg(\n",
    "    \"--model_dir\",\n",
    "    type=str,\n",
    "    default=None,\n",
    "    help=\n",
    "    \"Model dir, If you load a non-combined model, specify the directory of the model.\"\n",
    ")\n",
    "add_arg(\"--use_gpu\",type=bool,default=False, help=\"Whether use gpu.\")\n",
    "\n",
    "\n",
    "args = parser.parse_args(\n",
    "    \"--checkpoint_path examples/aishell/ckpt-loss2e-3-0.83-5/checkpoints/step-11725 --config examples/aishell/conf/deepspeech2.yaml --opts data.test_manifest examples/aishell/data/manifest.test data.mean_std_filepath examples/aishell/data/mean_std.npz  data.vocab_filepath examples/aishell/data/vocab.txt\".split()\n",
    ")\n",
    "\n",
    "\n",
    "config = get_cfg_defaults()\n",
    "if args.config:\n",
    "    config.merge_from_file(args.config)\n",
    "if args.opts:\n",
    "    config.merge_from_list(args.opts)\n",
    "config.freeze()\n",
    "print(config)\n",
    "\n",
    "args.warmup_manifest = config.data.test_manifest\n",
    "\n",
    "print_arguments(args)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "dataset = ManifestDataset(\n",
    "        config.data.test_manifest,\n",
    "        config.data.unit_type,\n",
    "        config.data.vocab_filepath,\n",
    "        config.data.mean_std_filepath,\n",
    "        augmentation_config=\"{}\",\n",
    "        max_duration=config.data.max_duration,\n",
    "        min_duration=config.data.min_duration,\n",
    "        stride_ms=config.data.stride_ms,\n",
    "        window_ms=config.data.window_ms,\n",
    "        n_fft=config.data.n_fft,\n",
    "        max_freq=config.data.max_freq,\n",
    "        target_sample_rate=config.data.target_sample_rate,\n",
    "        specgram_type=config.data.specgram_type,\n",
    "        feat_dim=config.data.feat_dim,\n",
    "        delta_delta=config.data.delat_delta,\n",
    "        use_dB_normalization=config.data.use_dB_normalization,\n",
    "        target_dB=config.data.target_dB,\n",
    "        random_seed=config.data.random_seed,\n",
    "        keep_transcription_text=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-03-26 02:55:57,930 - INFO - [checkpoint] Rank 0: loaded model from examples/aishell/ckpt-loss2e-3-0.83-5/checkpoints/step-11725.pdparams\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "layer summary:\n",
      "encoder.conv.conv_in.conv.weight|[32, 1, 41, 11]|14432\n",
      "encoder.conv.conv_in.bn.weight|[32]|32\n",
      "encoder.conv.conv_in.bn.bias|[32]|32\n",
      "encoder.conv.conv_in.bn._mean|[32]|32\n",
      "encoder.conv.conv_in.bn._variance|[32]|32\n",
      "encoder.conv.conv_stack.0.conv.weight|[32, 32, 21, 11]|236544\n",
      "encoder.conv.conv_stack.0.bn.weight|[32]|32\n",
      "encoder.conv.conv_stack.0.bn.bias|[32]|32\n",
      "encoder.conv.conv_stack.0.bn._mean|[32]|32\n",
      "encoder.conv.conv_stack.0.bn._variance|[32]|32\n",
      "encoder.rnn.rnn_stacks.0.fw_fc.weight|[1312, 3072]|4030464\n",
      "encoder.rnn.rnn_stacks.0.fw_bn.weight|[3072]|3072\n",
      "encoder.rnn.rnn_stacks.0.fw_bn.bias|[3072]|3072\n",
      "encoder.rnn.rnn_stacks.0.fw_bn._mean|[3072]|3072\n",
      "encoder.rnn.rnn_stacks.0.fw_bn._variance|[3072]|3072\n",
      "encoder.rnn.rnn_stacks.0.bw_fc.weight|[1312, 3072]|4030464\n",
      "encoder.rnn.rnn_stacks.0.bw_bn.weight|[3072]|3072\n",
      "encoder.rnn.rnn_stacks.0.bw_bn.bias|[3072]|3072\n",
      "encoder.rnn.rnn_stacks.0.bw_bn._mean|[3072]|3072\n",
      "encoder.rnn.rnn_stacks.0.bw_bn._variance|[3072]|3072\n",
      "encoder.rnn.rnn_stacks.0.fw_cell.weight_hh|[3072, 1024]|3145728\n",
      "encoder.rnn.rnn_stacks.0.fw_cell.bias_hh|[3072]|3072\n",
      "encoder.rnn.rnn_stacks.0.bw_cell.weight_hh|[3072, 1024]|3145728\n",
      "encoder.rnn.rnn_stacks.0.bw_cell.bias_hh|[3072]|3072\n",
      "encoder.rnn.rnn_stacks.0.fw_rnn.cell.weight_hh|[3072, 1024]|3145728\n",
      "encoder.rnn.rnn_stacks.0.fw_rnn.cell.bias_hh|[3072]|3072\n",
      "encoder.rnn.rnn_stacks.0.bw_rnn.cell.weight_hh|[3072, 1024]|3145728\n",
      "encoder.rnn.rnn_stacks.0.bw_rnn.cell.bias_hh|[3072]|3072\n",
      "encoder.rnn.rnn_stacks.1.fw_fc.weight|[2048, 3072]|6291456\n",
      "encoder.rnn.rnn_stacks.1.fw_bn.weight|[3072]|3072\n",
      "encoder.rnn.rnn_stacks.1.fw_bn.bias|[3072]|3072\n",
      "encoder.rnn.rnn_stacks.1.fw_bn._mean|[3072]|3072\n",
      "encoder.rnn.rnn_stacks.1.fw_bn._variance|[3072]|3072\n",
      "encoder.rnn.rnn_stacks.1.bw_fc.weight|[2048, 3072]|6291456\n",
      "encoder.rnn.rnn_stacks.1.bw_bn.weight|[3072]|3072\n",
      "encoder.rnn.rnn_stacks.1.bw_bn.bias|[3072]|3072\n",
      "encoder.rnn.rnn_stacks.1.bw_bn._mean|[3072]|3072\n",
      "encoder.rnn.rnn_stacks.1.bw_bn._variance|[3072]|3072\n",
      "encoder.rnn.rnn_stacks.1.fw_cell.weight_hh|[3072, 1024]|3145728\n",
      "encoder.rnn.rnn_stacks.1.fw_cell.bias_hh|[3072]|3072\n",
      "encoder.rnn.rnn_stacks.1.bw_cell.weight_hh|[3072, 1024]|3145728\n",
      "encoder.rnn.rnn_stacks.1.bw_cell.bias_hh|[3072]|3072\n",
      "encoder.rnn.rnn_stacks.1.fw_rnn.cell.weight_hh|[3072, 1024]|3145728\n",
      "encoder.rnn.rnn_stacks.1.fw_rnn.cell.bias_hh|[3072]|3072\n",
      "encoder.rnn.rnn_stacks.1.bw_rnn.cell.weight_hh|[3072, 1024]|3145728\n",
      "encoder.rnn.rnn_stacks.1.bw_rnn.cell.bias_hh|[3072]|3072\n",
      "encoder.rnn.rnn_stacks.2.fw_fc.weight|[2048, 3072]|6291456\n",
      "encoder.rnn.rnn_stacks.2.fw_bn.weight|[3072]|3072\n",
      "encoder.rnn.rnn_stacks.2.fw_bn.bias|[3072]|3072\n",
      "encoder.rnn.rnn_stacks.2.fw_bn._mean|[3072]|3072\n",
      "encoder.rnn.rnn_stacks.2.fw_bn._variance|[3072]|3072\n",
      "encoder.rnn.rnn_stacks.2.bw_fc.weight|[2048, 3072]|6291456\n",
      "encoder.rnn.rnn_stacks.2.bw_bn.weight|[3072]|3072\n",
      "encoder.rnn.rnn_stacks.2.bw_bn.bias|[3072]|3072\n",
      "encoder.rnn.rnn_stacks.2.bw_bn._mean|[3072]|3072\n",
      "encoder.rnn.rnn_stacks.2.bw_bn._variance|[3072]|3072\n",
      "encoder.rnn.rnn_stacks.2.fw_cell.weight_hh|[3072, 1024]|3145728\n",
      "encoder.rnn.rnn_stacks.2.fw_cell.bias_hh|[3072]|3072\n",
      "encoder.rnn.rnn_stacks.2.bw_cell.weight_hh|[3072, 1024]|3145728\n",
      "encoder.rnn.rnn_stacks.2.bw_cell.bias_hh|[3072]|3072\n",
      "encoder.rnn.rnn_stacks.2.fw_rnn.cell.weight_hh|[3072, 1024]|3145728\n",
      "encoder.rnn.rnn_stacks.2.fw_rnn.cell.bias_hh|[3072]|3072\n",
      "encoder.rnn.rnn_stacks.2.bw_rnn.cell.weight_hh|[3072, 1024]|3145728\n",
      "encoder.rnn.rnn_stacks.2.bw_rnn.cell.bias_hh|[3072]|3072\n",
      "decoder.ctc_lo.weight|[2048, 4300]|8806400\n",
      "decoder.ctc_lo.bias|[4300]|4300\n",
      "layer has 66 parameters, 80148012 elements.\n"
     ]
    }
   ],
   "source": [
    "model = DeepSpeech2InferModel.from_pretrained(dataset, config,\n",
    "                                             args.checkpoint_path)\n",
    "model.eval()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "examples/aishell/jit.model.pdmodel\n",
      "examples/aishell/jit.model.pdiparams\n",
      "0\n",
      "False\n"
     ]
    }
   ],
   "source": [
    "\n",
    "from paddle.inference import Config\n",
    "from paddle.inference import PrecisionType\n",
    "from paddle.inference import create_predictor\n",
    "\n",
    "args.use_gpu=False\n",
    "paddle.set_device('cpu')\n",
    "\n",
    "def init_predictor(args):\n",
    "    if args.model_dir is not None:\n",
    "        config = Config(args.model_dir)\n",
    "    else:\n",
    "        config = Config(args.model_file, args.params_file)\n",
    "\n",
    "    if args.use_gpu:\n",
    "        config.enable_use_gpu(memory_pool_init_size_mb=1000, device_id=0)\n",
    "#         config.enable_tensorrt_engine(precision_mode=PrecisionType.Float32,\n",
    "#                               use_calib_mode=True) # 开启TensorRT预测，精度为fp32，开启int8离线量化\n",
    "    else:\n",
    "        # If not specific mkldnn, you can set the blas thread.\n",
    "        # The thread num should not be greater than the number of cores in the CPU.\n",
    "        config.set_cpu_math_library_num_threads(1)\n",
    "        config.enable_mkldnn()\n",
    "        \n",
    "    config.enable_memory_optim()\n",
    "    config.switch_ir_optim(True)\n",
    "    \n",
    "    print(config.model_dir())\n",
    "    print(config.prog_file())\n",
    "    print(config.params_file())\n",
    "    print(config.gpu_device_id())\n",
    "    print(args.use_gpu)\n",
    "    predictor = create_predictor(config)\n",
    "    return predictor\n",
    "\n",
    "def run(predictor, audio, audio_len):\n",
    "    # copy img data to input tensor\n",
    "    input_names = predictor.get_input_names()\n",
    "    for i, name in enumerate(input_names):\n",
    "        print(\"input:\", i, name)\n",
    "        \n",
    "    audio_tensor = predictor.get_input_handle('audio')\n",
    "    audio_tensor.reshape(audio.shape)\n",
    "    audio_tensor.copy_from_cpu(audio.copy())\n",
    "    \n",
    "    audiolen_tensor = predictor.get_input_handle('audio_len')\n",
    "    audiolen_tensor.reshape(audio_len.shape)\n",
    "    audiolen_tensor.copy_from_cpu(audio_len.copy())\n",
    "\n",
    "    output_names = predictor.get_output_names()\n",
    "    for i, name in enumerate(output_names):\n",
    "        print(\"output:\", i, name)\n",
    "\n",
    "    # do the inference\n",
    "    predictor.run()\n",
    "\n",
    "    results = []\n",
    "    # get out data from output tensor\n",
    "    output_names = predictor.get_output_names()\n",
    "    for i, name in enumerate(output_names):\n",
    "        output_tensor = predictor.get_output_handle(name)\n",
    "        output_data = output_tensor.copy_to_cpu()\n",
    "        results.append(output_data)\n",
    "\n",
    "    return results\n",
    "\n",
    "\n",
    "predictor = init_predictor(args)\n",
    "\n",
    "def file_to_transcript(filename):\n",
    "        print(filename)\n",
    "        feature = dataset.process_utterance(filename, \"\")\n",
    "        audio = np.array([feature[0]]).astype('float32')  #[1, D, T]\n",
    "        audio_len = feature[0].shape[1]\n",
    "        audio_len = np.array([audio_len]).astype('int64')  # [1]\n",
    "        \n",
    "        \n",
    "        i_probs = run(predictor, audio, audio_len)\n",
    "        print('jit:', i_probs[0], type(i_probs[0]))\n",
    "        \n",
    "        audio = paddle.to_tensor(audio)\n",
    "        audio_len = paddle.to_tensor(audio_len)\n",
    "        print(audio.shape)\n",
    "        print(audio_len.shape)\n",
    "        \n",
    "        #eouts, eouts_len = model.encoder(audio, audio_len)\n",
    "        #probs = model.decoder.softmax(eouts)\n",
    "        probs = model.forward(audio, audio_len)\n",
    "        print('paddle:', probs.numpy())\n",
    "        \n",
    "        flag = np.allclose(i_probs[0], probs.numpy())\n",
    "        print(flag)\n",
    "        \n",
    "        return probs\n",
    "\n",
    "#         result_transcript = model.decode(\n",
    "#             audio,\n",
    "#             audio_len,\n",
    "#             vocab_list=dataset.vocab_list,\n",
    "#             decoding_method=config.decoding.decoding_method,\n",
    "#             lang_model_path=config.decoding.lang_model_path,\n",
    "#             beam_alpha=config.decoding.alpha,\n",
    "#             beam_beta=config.decoding.beta,\n",
    "#             beam_size=config.decoding.beam_size,\n",
    "#             cutoff_prob=config.decoding.cutoff_prob,\n",
    "#             cutoff_top_n=config.decoding.cutoff_top_n,\n",
    "#             num_processes=config.decoding.num_proc_bsearch)\n",
    "#         return result_transcript[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Warm-up Test Case %d: %s 0 /home/ssd5/zhanghui/DeepSpeech2.x/examples/aishell/../dataset/aishell/data_aishell/wav/test/S0764/BAC009S0764W0124.wav\n",
      "/home/ssd5/zhanghui/DeepSpeech2.x/examples/aishell/../dataset/aishell/data_aishell/wav/test/S0764/BAC009S0764W0124.wav\n",
      "input: 0 audio\n",
      "input: 1 audio_len\n",
      "output: 0 tmp_75\n",
      "jit: [[[8.91786298e-12 4.45648032e-12 3.67572750e-09 ... 8.91767563e-12\n",
      "   8.91573707e-12 4.64317296e-08]\n",
      "  [1.55950222e-15 2.62794089e-14 4.50423509e-12 ... 1.55944271e-15\n",
      "   1.55891342e-15 9.99992609e-01]\n",
      "  [1.24638127e-17 7.61802427e-16 2.93265812e-14 ... 1.24633371e-17\n",
      "   1.24587264e-17 1.00000000e+00]\n",
      "  ...\n",
      "  [4.37488240e-15 2.43676260e-12 1.98770514e-12 ... 4.37479896e-15\n",
      "   4.37354747e-15 1.00000000e+00]\n",
      "  [3.89334696e-13 1.66754856e-11 1.42900388e-11 ... 3.89329492e-13\n",
      "   3.89252270e-13 1.00000000e+00]\n",
      "  [1.00349985e-10 2.56293708e-10 2.91177582e-10 ... 1.00347876e-10\n",
      "   1.00334095e-10 9.99998808e-01]]] <class 'numpy.ndarray'>\n",
      "[1, 161, 522]\n",
      "[1]\n",
      "paddle: [[[8.91789680e-12 4.45649724e-12 3.67574149e-09 ... 8.91770945e-12\n",
      "   8.91577090e-12 4.64319072e-08]\n",
      "  [1.55950222e-15 2.62794089e-14 4.50423509e-12 ... 1.55944271e-15\n",
      "   1.55891342e-15 9.99992609e-01]\n",
      "  [1.24638599e-17 7.61805339e-16 2.93267472e-14 ... 1.24633842e-17\n",
      "   1.24587735e-17 1.00000000e+00]\n",
      "  ...\n",
      "  [4.37488240e-15 2.43676737e-12 1.98770514e-12 ... 4.37479896e-15\n",
      "   4.37354747e-15 1.00000000e+00]\n",
      "  [3.89336187e-13 1.66755481e-11 1.42900925e-11 ... 3.89330983e-13\n",
      "   3.89253761e-13 1.00000000e+00]\n",
      "  [1.00349985e-10 2.56293708e-10 2.91177582e-10 ... 1.00347876e-10\n",
      "   1.00334095e-10 9.99998808e-01]]]\n",
      "False\n"
     ]
    }
   ],
   "source": [
    "manifest = read_manifest(args.warmup_manifest)\n",
    "\n",
    "for idx, sample in enumerate(manifest[:1]):\n",
    "    print(\"Warm-up Test Case %d: %s\", idx, sample['audio_filepath'])\n",
    "    start_time = time.time()\n",
    "    transcript = file_to_transcript(sample['audio_filepath'])\n",
    "    finish_time = time.time()\n",
    "#     print(\"Response Time: %f, Transcript: %s\" %\n",
    "#           (finish_time - start_time, transcript))\n",
    "    break"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(1, 161, 522) (1,)\n",
      "input: 0 audio\n",
      "input: 1 audio_len\n",
      "output: 0 tmp_75\n",
      "jit: [[[8.91789680e-12 4.45649724e-12 3.67574149e-09 ... 8.91770945e-12\n",
      "   8.91577090e-12 4.64319072e-08]\n",
      "  [1.55950222e-15 2.62794089e-14 4.50423509e-12 ... 1.55944271e-15\n",
      "   1.55891342e-15 9.99992609e-01]\n",
      "  [1.24638599e-17 7.61805339e-16 2.93267472e-14 ... 1.24633842e-17\n",
      "   1.24587735e-17 1.00000000e+00]\n",
      "  ...\n",
      "  [4.37488240e-15 2.43676737e-12 1.98770514e-12 ... 4.37479896e-15\n",
      "   4.37354747e-15 1.00000000e+00]\n",
      "  [3.89336187e-13 1.66755481e-11 1.42900925e-11 ... 3.89330983e-13\n",
      "   3.89253761e-13 1.00000000e+00]\n",
      "  [1.00349985e-10 2.56293708e-10 2.91177582e-10 ... 1.00347876e-10\n",
      "   1.00334095e-10 9.99998808e-01]]]\n"
     ]
    }
   ],
   "source": [
    "def test(filename):\n",
    "    feature = dataset.process_utterance(filename, \"\")\n",
    "    audio = np.array([feature[0]]).astype('float32')  #[1, D, T]\n",
    "    audio_len = feature[0].shape[1]\n",
    "    audio_len = np.array([audio_len]).astype('int64')  # [1]\n",
    "    \n",
    "    print(audio.shape, audio_len.shape)\n",
    "\n",
    "    i_probs = run(predictor, audio, audio_len)\n",
    "    print('jit:', i_probs[0])\n",
    "    return i_probs\n",
    "    \n",
    "probs = test(sample['audio_filepath'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "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": "3.7.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}