sampled_softmax.ipynb

slavic_softmax.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import tensorflow as tf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "config = tf.ConfigProto()\n",
    "config.log_device_placement = False\n",
    "config.gpu_options.allow_growth = True\n",
    "s = tf.InteractiveSession(config=config)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "intermediate_projection_size = 256\n",
    "max_output_sequence_length = 50\n",
    "dict_size = 10**5\n",
    "batch_size = 100"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### mocking your code"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "class self:\n",
    "    #mock\n",
    "    num_sampled = 100\n",
    "    intermediate_projection = tf.get_variable(\"logits\",\n",
    "                                              #assuming shape [batch, time, unit] or vice versa\n",
    "                                              initializer=tf.random_normal([batch_size,\n",
    "                                                                            max_output_sequence_length,\n",
    "                                                                            intermediate_projection_size]))\n",
    "    targets = tf.get_variable(\"answers\",\n",
    "                               initializer=tf.reshape(tf.range(batch_size*max_output_sequence_length),\n",
    "                                                      [batch_size,max_output_sequence_length])\n",
    "                               )\n",
    "#the old code\n",
    "with tf.variable_scope('output_projection'):\n",
    "    self.output_projection_weights = tf.get_variable('weights',\n",
    "                                                     shape=[intermediate_projection_size, dict_size],\n",
    "                                                     initializer=tf.uniform_unit_scaling_initializer())\n",
    "    self.output_projection_biases = tf.get_variable('biases',\n",
    "                                                    shape=[dict_size],\n",
    "                                                    initializer=tf.constant_initializer(0.0))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### main"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "\n",
    "with tf.name_scope('loss') as ns:\n",
    "    flat_intermediate_projection = tf.reshape(self.intermediate_projection,[-1,intermediate_projection_size])\n",
    "    flat_targets = tf.reshape(self.targets,[-1,1])\n",
    "    \n",
    "    flat_loss = tf.nn.sampled_softmax_loss(tf.transpose(self.output_projection_weights),\n",
    "                                           self.output_projection_biases,\n",
    "                                           flat_intermediate_projection,\n",
    "                                           flat_targets,\n",
    "                                           num_sampled = self.num_sampled,\n",
    "                                           num_classes = dict_size,\n",
    "                                          )\n",
    "    self.loss = tf.reshape(flat_loss,self.targets.get_shape())\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "100 loops, best of 3: 17.6 ms per loop\n"
     ]
    }
   ],
   "source": [
    "s.run(tf.initialize_all_variables())\n",
    "%timeit s.run(self.loss).shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python [Root]",
   "language": "python",
   "name": "Python [Root]"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}