notebooks/Tensorflow_Inference.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "%%bash\n",
    "#wget http://download.tensorflow.org/models/resnet_v1_50_2016_08_28.tar.gz\n",
    "#tar -xvf resnet_v1_50_2016_08_28.tar.gz\n",
    "#rm resnet_v1_50_2016_08_28.tar.gz"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "import os\n",
    "import sys\n",
    "import numpy as np\n",
    "import tensorflow as tf\n",
    "# Upgrade dask before importing contrib!\n",
    "import tensorflow.contrib.slim\n",
    "from tensorflow.contrib.slim.nets import resnet_v1\n",
    "from common.params_inf import *\n",
    "from common.utils import *"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Force one-gpu\n",
    "os.environ[\"CUDA_VISIBLE_DEVICES\"] = \"0\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "OS:  linux\n",
      "Python:  3.5.2 |Anaconda custom (64-bit)| (default, Jul  2 2016, 17:53:06) \n",
      "[GCC 4.4.7 20120313 (Red Hat 4.4.7-1)]\n",
      "Numpy:  1.14.1\n",
      "GPU:  ['Tesla P100-PCIE-16GB', 'Tesla P100-PCIE-16GB']\n",
      "CUDA Version 8.0.61\n",
      "CuDNN Version  6.0.21\n"
     ]
    }
   ],
   "source": [
    "print(\"OS: \", sys.platform)\n",
    "print(\"Python: \", sys.version)\n",
    "print(\"Numpy: \", np.__version__)\n",
    "print(\"GPU: \", get_gpu_name())\n",
    "print(get_cuda_version())\n",
    "print(\"CuDNN Version \", get_cudnn_version())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "def predict_fn(classifier, data, batchsize):\n",
    "    \"\"\" Return features from classifier \"\"\"\n",
    "    out = np.zeros((len(data), RESNET_FEATURES), np.float32)\n",
    "    for idx, dta in yield_mb_X(data, batchsize):\n",
    "        pred = sess.run(classifier, feed_dict={input_tensor: dta}).squeeze()\n",
    "        out[idx*batchsize:(idx+1)*batchsize] = pred\n",
    "    return out"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(1280, 224, 224, 3) (1280, 3, 224, 224)\n"
     ]
    }
   ],
   "source": [
    "# Create batches of fake data\n",
    "fake_input_data_cl, fake_input_data_cf = give_fake_data(BATCH_SIZE*BATCHES_GPU)\n",
    "print(fake_input_data_cl.shape, fake_input_data_cf.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Restoring parameters from resnet_v1_50.ckpt\n"
     ]
    }
   ],
   "source": [
    "# Placeholders\n",
    "checkpoint_file = 'resnet_v1_50.ckpt'\n",
    "input_tensor = tf.placeholder(tf.float32, shape=(None,224,224,3), name='input_image')\n",
    "\n",
    "# Load the model\n",
    "sess = tf.Session()\n",
    "arg_scope = resnet_v1.resnet_arg_scope()\n",
    "with tensorflow.contrib.slim.arg_scope(arg_scope):\n",
    "    # Docstring ->\n",
    "    #     num_classes: Number of predicted classes for classification tasks. If None\n",
    "    #  we return the features before the logit layer.\n",
    "    logits, end_points = resnet_v1.resnet_v1_50(input_tensor, is_training=False)\n",
    "    \n",
    "saver = tf.train.Saver()\n",
    "saver.restore(sess, checkpoint_file)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "cold_start = predict_fn(logits, fake_input_data_cl, BATCH_SIZE)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 1.96 s, sys: 480 ms, total: 2.44 s\n",
      "Wall time: 2.26 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "features = predict_fn(logits, fake_input_data_cl, BATCH_SIZE)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Images per second 566.3716814159293\n"
     ]
    }
   ],
   "source": [
    "print(\"Images per second {}\".format((BATCH_SIZE*BATCHES_GPU)/2.26))"
   ]
  }
 ],
 "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": "3.5.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}