Don't mix integers with floats in Theano.

scalars.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import os"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "os.environ['THEANO_FLAGS'] = 'floatX=float32,device=gpu1,nvcc.fastmath=True'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using gpu device 1: Tesla K40m (CNMeM is disabled)\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "import theano as th\n",
    "import theano.tensor as tt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "A = tt.matrix('A')\n",
    "B = tt.matrix('B')\n",
    "\n",
    "A_shape_0_ = tt.cast(A.shape[0], th.config.floatX)\n",
    "\n",
    "f_gpu = th.function([A, B], tt.dot(A, B) / A_shape_0_)\n",
    "f_cpu = th.function([A, B], tt.dot(A, B) / A.shape[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "HostFromGpu [@A] ''   8\n",
      " |GpuElemwise{TrueDiv}[(0, 0)] [@B] ''   7\n",
      "   |GpuDot22 [@C] ''   4\n",
      "   | |GpuFromHost [@D] ''   1\n",
      "   | | |A [@E]\n",
      "   | |GpuFromHost [@F] ''   2\n",
      "   |   |B [@G]\n",
      "   |GpuFromHost [@H] ''   6\n",
      "     |Elemwise{Cast{float32}} [@I] ''   5\n",
      "       |InplaceDimShuffle{x,x} [@J] ''   3\n",
      "         |Shape_i{0} [@K] ''   0\n",
      "           |A [@E]\n"
     ]
    }
   ],
   "source": [
    "th.printing.debugprint(f_gpu)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Elemwise{true_div,no_inplace} [@A] ''   6\n",
      " |HostFromGpu [@B] ''   5\n",
      " | |GpuDot22 [@C] ''   4\n",
      " |   |GpuFromHost [@D] ''   1\n",
      " |   | |A [@E]\n",
      " |   |GpuFromHost [@F] ''   2\n",
      " |     |B [@G]\n",
      " |InplaceDimShuffle{x,x} [@H] ''   3\n",
      "   |Shape_i{0} [@I] ''   0\n",
      "     |A [@E]\n"
     ]
    }
   ],
   "source": [
    "th.printing.debugprint(f_cpu)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "100 loops, best of 3: 3 ms per loop\n",
      "100 loops, best of 3: 4.21 ms per loop\n"
     ]
    }
   ],
   "source": [
    "A = np.asarray(np.random.randn(1000, 1000), dtype=th.config.floatX)\n",
    "B = np.asarray(np.random.randn(1000, 1000), dtype=th.config.floatX)\n",
    "\n",
    "%timeit f_gpu(A, B)\n",
    "%timeit f_cpu(A, B)"
   ]
  }
 ],
 "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": "3.4.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}