组织数据batch之batch()用法.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import tensorflow as tf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "files = tf.train.match_filenames_once(\"./output/data_tfrecords-*\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 创建文件队列shuffle参数设置为True，打乱文件"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "filename_queue = tf.train.string_input_producer(files, shuffle=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 实例化TFRecordReader类，准备读取TFRecord文件"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "reader = tf.TFRecordReader()\n",
    "_,serialized_example = reader.read(filename_queue)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 解析读取的样例"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "features = tf.parse_single_example(\n",
    "    serialized_example,\n",
    "    features={\n",
    "        \"image_raw\":tf.FixedLenFeature([],tf.string),\n",
    "        \"pixels\":tf.FixedLenFeature([],tf.int64),\n",
    "        \"label\":tf.FixedLenFeature([],tf.int64)\n",
    "    })\n",
    "images = tf.decode_raw(features[\"image_raw\"],tf.uint8)\n",
    "labels = tf.cast(features[\"label\"],tf.int32)\n",
    "pixels = tf.cast(features[\"pixels\"],tf.int32)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "####  设置每个batch中样例的个数"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "batch_size = 10"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 用于组合成batch的队列中最多可以缓存的样例的个数"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "capacity = 5000 + 3 * batch_size"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### set_shape()函数用来设置设置尺寸，这一步操作也可以使用image.resize_images()\n",
    "#### 函数来完成。设置尺寸的操作是必须的，在1.0.0版的TensorFlow中如果没有这一步\n",
    "#### 会在batch()函数的capacity参数处报错ValueError: All shapes must be fully\n",
    "#### defined: [TensorShape([Dimension(None)]), TensorShape([])]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "images.set_shape(784)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 使用batch()函数将样例组合成batch\n",
    "#### 函数原型batch(tensors,batch_size,num_threads,capacity,enqueue_many,shapes,\n",
    "####                         dynamic_pad,allow_smaller_final_batch,shared_name,name)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "image_batch, label_batch = tf.train.batch([images, labels],\n",
    "                                    batch_size=batch_size,capacity=capacity,)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 接下来就可以像之前的网络模型那样创建会话并开始训练了，与之前的训练过程相比，\n",
    "#### 这里的会话中增加了多线程处理的相关代码"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[0 0 0 ... 0 0 0]\n",
      " [0 0 0 ... 0 0 0]\n",
      " [0 0 0 ... 0 0 0]\n",
      " ...\n",
      " [0 0 0 ... 0 0 0]\n",
      " [0 0 0 ... 0 0 0]\n",
      " [0 0 0 ... 0 0 0]] [7 2 1 0 4 1 4 9 5 9]\n",
      "[[0 0 0 ... 0 0 0]\n",
      " [0 0 0 ... 0 0 0]\n",
      " [0 0 0 ... 0 0 0]\n",
      " ...\n",
      " [0 0 0 ... 0 0 0]\n",
      " [0 0 0 ... 0 0 0]\n",
      " [0 0 0 ... 0 0 0]] [0 6 9 0 1 5 9 7 3 4]\n",
      "[[0 0 0 ... 0 0 0]\n",
      " [0 0 0 ... 0 0 0]\n",
      " [0 0 0 ... 0 0 0]\n",
      " ...\n",
      " [0 0 0 ... 0 0 0]\n",
      " [0 0 0 ... 0 0 0]\n",
      " [0 0 0 ... 0 0 0]] [9 6 6 5 4 0 7 4 0 1]\n"
     ]
    }
   ],
   "source": [
    "with tf.Session() as sess:\n",
    "    tf.local_variables_initializer().run()\n",
    "    coord = tf.train.Coordinator()\n",
    "    threads = tf.train.start_queue_runners(sess=sess, coord=coord)\n",
    "\n",
    "    #一般在这个循环内开始训练，这里设定了训练轮数为3\n",
    "    #在每一轮训练的过程中都会执行一个组合样例为batch的操作并打印出来\n",
    "    for i in range(3):\n",
    "        xs,ys=sess.run([image_batch,label_batch])\n",
    "        print(xs,ys)\n",
    "\n",
    "    coord.request_stop()\n",
    "    coord.join(threads)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "kernelspec": {
   "display_name": "Python [conda root]",
   "language": "python",
   "name": "conda-root-py"
  },
  "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": 1
}