joblib_gym_issue.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "# joblib and gym issue"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[2017-03-09 19:53:59,642] Making new env: Taxi-v2\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "+---------+\n",
      "|R: | : :G|\n",
      "| : : : : |\n",
      "| :\u001b[43m \u001b[0m: : : |\n",
      "| | : | : |\n",
      "|\u001b[35mY\u001b[0m| : |\u001b[34;1mB\u001b[0m: |\n",
      "+---------+\n",
      "\n"
     ]
    }
   ],
   "source": [
    "import gym\n",
    "import numpy as np, pandas as pd\n",
    "\n",
    "env = gym.make(\"Taxi-v2\")\n",
    "env.reset()\n",
    "env.render()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "n_states=500, n_actions=6\n"
     ]
    }
   ],
   "source": [
    "n_states = env.observation_space.n\n",
    "n_actions = env.action_space.n\n",
    "\n",
    "print(\"n_states=%i, n_actions=%i\"%(n_states,n_actions))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "policy = np.array([[ 1. / n_actions for a in range(n_actions)] for s in range(n_states)])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deletable": true,
    "editable": true
   },
   "source": [
    "# Play the game\n",
    "\n",
    "Just like before, but we also record all states and actions we took."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "def generate_session(t_max=10**6):\n",
    "    \"\"\"\n",
    "    Play game until end or for t_max ticks.\n",
    "    returns: list of states, list of actions and sum of rewards\n",
    "    \"\"\"\n",
    "    states, actions = [], []\n",
    "    total_reward = 0.\n",
    "    \n",
    "    s = env.reset()\n",
    "    \n",
    "    for t in range(t_max):\n",
    "        \n",
    "        a = np.random.choice(range(n_actions), size=1, p=policy[s])[0]        \n",
    "        new_s,r,done,info = env.step(a)\n",
    "        \n",
    "        total_reward += r\n",
    "        states.append(s)\n",
    "        actions.append(a)\n",
    "        \n",
    "        s = new_s\n",
    "        if done:\n",
    "            break\n",
    "    return states, actions, total_reward      "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true,
    "deletable": true,
    "editable": true
   },
   "outputs": [],
   "source": [
    "# attention, n_jobs = 25"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from sklearn.externals import joblib\n",
    "def parallel_sessions(n_samples):\n",
    "    return joblib.Parallel(n_jobs=25)(joblib.delayed(generate_session)() for _ in range(n_samples))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "n_samples = 25"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# to check if sessions are unique list is converted to set as follows:\n",
    "#     unique_data = [list(x) for x in set((tuple(x[0]), tuple(x[1])) for x in sessions)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      " Run 1\n",
      "sizes with joblib: all:25, unique:2\n",
      "sizes without joblib: all:25, unique:25\n",
      "\n",
      " Run 2\n",
      "sizes with joblib: all:25, unique:2\n",
      "sizes without joblib: all:25, unique:25\n",
      "\n",
      " Run 3\n",
      "sizes with joblib: all:25, unique:2\n",
      "sizes without joblib: all:25, unique:25\n",
      "\n",
      " Run 4\n",
      "sizes with joblib: all:25, unique:2\n",
      "sizes without joblib: all:25, unique:25\n",
      "\n",
      " Run 5\n",
      "sizes with joblib: all:25, unique:2\n",
      "sizes without joblib: all:25, unique:25\n",
      "\n",
      " Run 6\n",
      "sizes with joblib: all:25, unique:3\n",
      "sizes without joblib: all:25, unique:25\n",
      "\n",
      " Run 7\n",
      "sizes with joblib: all:25, unique:2\n",
      "sizes without joblib: all:25, unique:25\n",
      "\n",
      " Run 8\n",
      "sizes with joblib: all:25, unique:2\n",
      "sizes without joblib: all:25, unique:25\n",
      "\n",
      " Run 9\n",
      "sizes with joblib: all:25, unique:2\n",
      "sizes without joblib: all:25, unique:25\n",
      "\n",
      " Run 10\n",
      "sizes with joblib: all:25, unique:2\n",
      "sizes without joblib: all:25, unique:25\n"
     ]
    }
   ],
   "source": [
    "for i in range(10):\n",
    "    sessions = parallel_sessions(n_samples)#[generate_session() for _ in range(n_samples)]\n",
    "    sessions_2 = [generate_session() for _ in range(n_samples)]\n",
    "    unique_data = [list(x) for x in set((tuple(x[0]), tuple(x[1])) for x in sessions)]\n",
    "    unique_data_2 = [list(x) for x in set((tuple(x[0]), tuple(x[1])) for x in sessions_2)]\n",
    "    print('\\n Run {}'.format(i+1))\n",
    "    print('sizes with joblib: all:{}, unique:{}'.format(len(sessions), len(unique_data)))\n",
    "    print('sizes without joblib: all:{}, unique:{}'.format(len(sessions_2), len(unique_data_2)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "kernelspec": {
   "display_name": "Python (myenv)",
   "language": "python",
   "name": "myenv"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}