03_Optimize_Model.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Optimized Model for Deployment"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Predict with Python (SLOW)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from tensorflow.contrib import predictor\n",
    "import numpy as np\n",
    "\n",
    "saved_model_path = './linear_model/cpu/pipeline_tfserving/0'\n",
    "print(saved_model_path)\n",
    "\n",
    "input_shape = 1\n",
    "input_data = np.random.random_sample(input_shape)\n",
    "\n",
    "predict_fn = predictor.from_saved_model(saved_model_path)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "%%time\n",
    "predictions = predict_fn({'inputs': input_data})\n",
    "\n",
    "print('Prediction: %s' % predictions[\"outputs\"])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Optimize with TensorFlow Lite\n",
    "\n",
    "![PipelineAI + TensorFlow Lite](https://pipeline.ai/assets/img/toco-optimizer.png)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "import os\n",
    "\n",
    "from shutil import rmtree\n",
    "\n",
    "saved_model_path = './linear_model/cpu/pipeline_tfserving/0'\n",
    "print(saved_model_path)\n",
    "\n",
    "tflite_model_base_path = './linear_model/cpu/tflite/'\n",
    "\n",
    "os.makedirs(tflite_model_base_path, exist_ok=True)\n",
    "\n",
    "converter = tf.lite.TocoConverter.from_saved_model(saved_model_path)\n",
    "\n",
    "# TF 1.11+\n",
    "converter.post_training_quantize = True\n",
    "\n",
    "tflite_model = converter.convert()\n",
    "\n",
    "tflite_model_path = '%s/tflite_optimized_model.tflite' % tflite_model_base_path\n",
    "\n",
    "model_size = open(tflite_model_path, \"wb\").write(tflite_model)\n",
    "\n",
    "print('\\nModel size reduced to %s bytes' % model_size)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "%%bash -s \"$tflite_model_path\"\n",
    "echo \"ls -al $1\"\n",
    "echo \"\"\n",
    "ls -al $1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import tensorflow as tf\n",
    "\n",
    "# Load TFLite model and allocate tensors.\n",
    "interpreter = tf.lite.Interpreter(model_path=tflite_model_path)\n",
    "interpreter.allocate_tensors()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Get input and output tensors.\n",
    "input_details = interpreter.get_input_details()\n",
    "print('Input Tensor Details: %s' % input_details)\n",
    "\n",
    "output_details = interpreter.get_output_details()\n",
    "print('Output Tensor Details: %s' % output_details)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Test model on random input data.\n",
    "input_shape = input_details[0]['shape']\n",
    "input_data = np.array(np.random.random_sample(input_shape), dtype=np.float32)\n",
    "print('Input: %s' % input_data)\n",
    "interpreter.set_tensor(input_details[0]['index'], input_data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "%%time\n",
    "interpreter.invoke()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "output_data = interpreter.get_tensor(output_details[0]['index'])\n",
    "print('Prediction: %s' % output_data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "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": "3.6.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}