rl_ppo_lunarlander.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "# PPO, https://spinningup.openai.com/en/latest/algorithms/ppo.html\n",
    "# Pretty stable, but could be better"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Populating the interactive namespace from numpy and matplotlib\n"
     ]
    }
   ],
   "source": [
    "%pylab inline\n",
    "import random\n",
    "import time\n",
    "import numpy as np\n",
    "\n",
    "import tensorflow as tf\n",
    "import tensorflow.keras.backend as K\n",
    "from tensorflow.keras.models import Model, clone_model\n",
    "from tensorflow.keras.optimizers import Adam, SGD\n",
    "from tensorflow.keras.layers import Input, Dense, Activation, Lambda\n",
    "\n",
    "import gym"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(Box(8,), Discrete(4), gym.spaces.discrete.Discrete)"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#env = gym.make(\"CartPole-v1\")\n",
    "env = gym.make(\"LunarLander-v2\")\n",
    "env.observation_space, env.action_space, type(env.action_space)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model_1\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_1 (InputLayer)         [(None, 8)]               0         \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 32)                288       \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 32)                1056      \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 32)                1056      \n",
      "_________________________________________________________________\n",
      "dense_3 (Dense)              (None, 4)                 132       \n",
      "_________________________________________________________________\n",
      "lambda (Lambda)              (None, 4)                 0         \n",
      "=================================================================\n",
      "Total params: 2,532\n",
      "Trainable params: 2,532\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "Model: \"model_2\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_4 (InputLayer)         [(None, 8)]               0         \n",
      "_________________________________________________________________\n",
      "dense_4 (Dense)              (None, 32)                288       \n",
      "_________________________________________________________________\n",
      "dense_5 (Dense)              (None, 32)                1056      \n",
      "_________________________________________________________________\n",
      "dense_6 (Dense)              (None, 32)                1056      \n",
      "_________________________________________________________________\n",
      "dense_7 (Dense)              (None, 1)                 33        \n",
      "=================================================================\n",
      "Total params: 2,433\n",
      "Trainable params: 2,433\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# policy function\n",
    "x = in_state = Input(env.observation_space.shape)\n",
    "x = Dense(32, activation='elu')(x)\n",
    "x = Dense(32, activation='elu')(x)\n",
    "x = Dense(32, activation='elu')(x)\n",
    "x = Dense(env.action_space.n)(x)\n",
    "action_dist = Lambda(lambda x: tf.nn.log_softmax(x, axis=-1))(x)\n",
    "p = Model(in_state, action_dist)\n",
    "pf = K.function(p.layers[0].input,\n",
    "                [p.layers[-1].output,\n",
    "                 tf.random.categorical(p.layers[-1].output, 1)[0]])\n",
    "\n",
    "in_advantage = Input((1,))\n",
    "in_old_prediction = Input((env.action_space.n,))\n",
    "def loss(y_true, y_pred): \n",
    "  advantage = tf.reshape(in_advantage, (-1,))\n",
    "  \n",
    "  # y_pred is the log probs of the actions\n",
    "  # y_true is the action mask\n",
    "  prob = tf.reduce_sum(y_true * y_pred, axis=-1)\n",
    "  old_prob = tf.reduce_sum(y_true * in_old_prediction, axis=-1)\n",
    "  ratio = tf.exp(prob - old_prob)  # hehe, they are log probs, so we subtract\n",
    "  \n",
    "  # this is the VPG objective\n",
    "  #ll = -(prob * advantage)\n",
    "  \n",
    "  # this is PPO objective\n",
    "  ll = -K.minimum(ratio*advantage, K.clip(ratio, 0.8, 1.2)*advantage)\n",
    "  return ll\n",
    "\n",
    "popt = Model([in_state, in_advantage, in_old_prediction], action_dist)\n",
    "popt.compile(Adam(5e-4), loss)\n",
    "popt.summary()\n",
    "\n",
    "# value function\n",
    "x = in1 = Input(env.observation_space.shape)\n",
    "x = Dense(32, activation='elu')(x)\n",
    "x = Dense(32, activation='elu')(x)\n",
    "x = Dense(32, activation='elu')(x)\n",
    "x = Dense(1)(x)\n",
    "v = Model(in1, x)\n",
    "v.compile(Adam(1e-3), 'mse')\n",
    "v.summary()\n",
    "\n",
    "vf = K.function(v.layers[0].input, v.layers[-1].output)\n",
    "\n",
    "gamma = 0.99\n",
    "lll = []"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "  0  ep_rew:  -154.44  loss:  85.10   vloss:  6250.24  counts:  8263/ 96 tm: 4.52 s\n",
      "  1  ep_rew:  -129.85  loss:  45.43   vloss:  1966.16  counts:  8238/102 tm: 3.19 s\n",
      "  2  ep_rew:  -141.88  loss:   5.01   vloss:  2239.95  counts:  8237/ 94 tm: 3.00 s\n",
      "  3  ep_rew:  -130.05  loss:  -4.49   vloss:   981.43  counts:  8192/102 tm: 2.91 s\n",
      "  4  ep_rew:  -136.80  loss:   3.62   vloss:  1238.68  counts:  8252/101 tm: 2.93 s\n",
      "  5  ep_rew:  -112.73  loss: -12.85   vloss:  1264.44  counts:  8271/ 97 tm: 3.01 s\n",
      "  6  ep_rew:  -132.80  loss:  12.59   vloss:   984.91  counts:  8210/ 98 tm: 2.90 s\n",
      "  7  ep_rew:  -120.64  loss:  -3.98   vloss:  1835.86  counts:  8222/ 96 tm: 2.98 s\n",
      "  8  ep_rew:  -115.60  loss:  -3.76   vloss:  1517.28  counts:  8262/ 91 tm: 3.04 s\n",
      "  9  ep_rew:  -106.82  loss:  -1.44   vloss:   838.22  counts:  8280/ 94 tm: 3.02 s\n",
      " 10  ep_rew:   -89.12  loss: -18.41   vloss:  1886.56  counts:  8255/ 77 tm: 3.73 s\n",
      " 11  ep_rew:   -90.95  loss:   8.20   vloss:  1633.67  counts:  8248/ 86 tm: 3.35 s\n",
      " 12  ep_rew:   -77.80  loss:  -8.44   vloss:   807.43  counts:  8256/ 80 tm: 3.71 s\n",
      " 13  ep_rew:   -57.32  loss:  -9.24   vloss:   631.23  counts:  8274/ 83 tm: 3.37 s\n",
      " 14  ep_rew:   -53.89  loss: -11.14   vloss:  1008.91  counts:  8269/ 63 tm: 4.86 s\n",
      " 15  ep_rew:   -42.72  loss:   4.88   vloss:   635.23  counts:  8259/ 73 tm: 3.28 s\n",
      " 16  ep_rew:   -33.70  loss:  -8.51   vloss:   848.04  counts:  8278/ 65 tm: 4.75 s\n",
      " 17  ep_rew:   -20.57  loss:  -7.42   vloss:   915.08  counts:  8251/ 57 tm: 5.59 s\n",
      " 18  ep_rew:   -14.92  loss:  -5.94   vloss:  1048.20  counts:  8249/ 47 tm: 6.17 s\n",
      " 19  ep_rew:    -4.67  loss:   0.60   vloss:   813.89  counts:  9026/ 41 tm: 8.13 s\n",
      " 20  ep_rew:    -4.01  loss:   2.00   vloss:   516.35  counts:  8719/ 46 tm: 6.59 s\n",
      " 21  ep_rew:   -13.02  loss:   2.02   vloss:   829.14  counts:  8693/ 35 tm: 7.64 s\n",
      " 22  ep_rew:    -9.23  loss:  -7.97   vloss:   706.81  counts:  8923/ 25 tm: 9.16 s\n",
      " 23  ep_rew:    -9.33  loss:   4.23   vloss:   617.43  counts:  8283/ 18 tm: 10.06 s\n",
      " 24  ep_rew:   -31.39  loss:  -2.00   vloss:   671.96  counts:  9073/ 15 tm: 12.03 s\n",
      " 25  ep_rew:   -18.51  loss:   5.44   vloss:   793.81  counts:  8232/ 14 tm: 10.92 s\n",
      " 26  ep_rew:   -42.43  loss:  -0.80   vloss:   782.01  counts:  8760/ 18 tm: 10.54 s\n",
      " 27  ep_rew:   -13.79  loss:  -1.89   vloss:   801.76  counts:  8487/ 16 tm: 11.70 s\n",
      " 28  ep_rew:    37.69  loss: -12.70   vloss:   240.46  counts:  8503/ 11 tm: 11.80 s\n",
      " 29  ep_rew:    14.04  loss:  -0.87   vloss:   472.80  counts:  8261/ 14 tm: 10.39 s\n",
      " 30  ep_rew:    16.45  loss:  -6.90   vloss:   308.73  counts:  8721/ 10 tm: 11.43 s\n",
      " 31  ep_rew:    44.00  loss:   1.45   vloss:   474.58  counts:  8540/ 17 tm: 10.59 s\n",
      " 32  ep_rew:    29.49  loss:   4.96   vloss:   493.54  counts:  9148/ 21 tm: 10.18 s\n",
      " 33  ep_rew:    26.67  loss:  -3.08   vloss:   576.46  counts:  8236/ 19 tm: 8.72 s\n",
      " 34  ep_rew:    57.33  loss:  -4.88   vloss:   336.48  counts:  9123/ 14 tm: 10.73 s\n",
      " 35  ep_rew:    85.11  loss:  -1.27   vloss:   235.50  counts:  8467/ 10 tm: 10.52 s\n",
      " 36  ep_rew:    61.13  loss:   4.28   vloss:   430.26  counts:  8354/ 13 tm: 9.99 s\n",
      " 37  ep_rew:   113.39  loss:  -7.99   vloss:   281.95  counts:  9089/ 10 tm: 11.03 s\n",
      " 38  ep_rew:    83.73  loss:   3.43   vloss:   290.49  counts:  8873/ 12 tm: 10.14 s\n",
      " 39  ep_rew:   103.24  loss:  -1.41   vloss:   173.40  counts:  8420/ 10 tm: 9.87 s\n",
      " 40  ep_rew:    59.29  loss:   5.74   vloss:   418.06  counts:  8270/ 12 tm: 10.21 s\n",
      " 41  ep_rew:   121.52  loss:  -5.50   vloss:   107.07  counts:  9000/  9 tm: 11.34 s\n",
      " 42  ep_rew:    71.62  loss:   2.85   vloss:   281.21  counts:  8725/ 11 tm: 10.66 s\n",
      " 43  ep_rew:   115.45  loss:  -3.75   vloss:   102.09  counts:  9000/  9 tm: 11.05 s\n",
      " 44  ep_rew:   118.28  loss:  -1.32   vloss:   121.64  counts:  8199/  9 tm: 10.17 s\n",
      " 45  ep_rew:    66.82  loss:   5.13   vloss:   250.21  counts:  8385/ 10 tm: 10.69 s\n",
      " 46  ep_rew:    75.54  loss:  -1.73   vloss:   300.84  counts:  8988/ 12 tm: 10.47 s\n",
      " 47  ep_rew:   110.95  loss:  -3.25   vloss:   150.57  counts:  8458/ 10 tm: 10.18 s\n",
      " 48  ep_rew:   116.29  loss:  -1.35   vloss:   165.06  counts:  9123/ 10 tm: 10.04 s\n",
      " 49  ep_rew:   118.22  loss:  -1.46   vloss:   155.42  counts:  8244/  9 tm: 9.54 s\n",
      " 50  ep_rew:   110.67  loss:  -0.93   vloss:   203.88  counts:  8422/ 10 tm: 8.43 s\n",
      " 51  ep_rew:   102.18  loss:   0.90   vloss:   220.39  counts:  8785/ 11 tm: 9.82 s\n",
      " 52  ep_rew:   116.17  loss:   1.21   vloss:   145.73  counts:  8512/ 10 tm: 9.02 s\n",
      " 53  ep_rew:   116.14  loss:  -1.87   vloss:   245.38  counts:  8636/ 11 tm: 8.90 s\n",
      " 54  ep_rew:   145.17  loss:  -2.53   vloss:   111.20  counts:  8235/  9 tm: 8.99 s\n",
      " 55  ep_rew:   117.78  loss:   4.96   vloss:   158.27  counts:  8231/  9 tm: 8.28 s\n",
      " 56  ep_rew:   120.96  loss:  -0.56   vloss:   141.93  counts:  8440/ 10 tm: 8.39 s\n",
      " 57  ep_rew:   129.52  loss:  -2.50   vloss:   124.23  counts:  9181/ 10 tm: 9.53 s\n",
      " 58  ep_rew:   132.95  loss:  -1.48   vloss:   117.58  counts:  8264/  9 tm: 8.93 s\n",
      " 59  ep_rew:   137.87  loss:  -0.01   vloss:    28.42  counts:  9000/  9 tm: 9.14 s\n",
      " 60  ep_rew:   150.81  loss:  -1.49   vloss:    36.14  counts:  9000/  9 tm: 9.63 s\n",
      " 61  ep_rew:   133.78  loss:   2.31   vloss:    95.63  counts:  8260/  9 tm: 8.44 s\n",
      " 62  ep_rew:   125.10  loss:   1.08   vloss:   186.97  counts:  8603/ 11 tm: 8.69 s\n",
      " 63  ep_rew:   141.99  loss:  -1.45   vloss:    87.85  counts:  8210/  9 tm: 8.75 s\n",
      " 64  ep_rew:   121.61  loss:   3.89   vloss:   229.23  counts:  8636/ 11 tm: 8.66 s\n",
      " 65  ep_rew:   122.97  loss:  -0.75   vloss:   218.79  counts:  8620/ 11 tm: 9.14 s\n",
      " 66  ep_rew:   142.49  loss:  -1.17   vloss:    19.25  counts:  9000/  9 tm: 9.45 s\n",
      " 67  ep_rew:   155.94  loss:  -1.02   vloss:    18.89  counts:  9000/  9 tm: 8.28 s\n",
      " 68  ep_rew:   170.66  loss:  -1.03   vloss:    20.81  counts:  9000/  9 tm: 9.12 s\n",
      " 69  ep_rew:   186.89  loss:  -3.67   vloss:   367.88  counts:  8749/ 12 tm: 8.69 s\n",
      " 70  ep_rew:   237.48  loss: -17.77   vloss:   552.37  counts:  8427/ 22 tm: 6.23 s\n",
      " 71  ep_rew:   240.16  loss:  -4.74   vloss:   555.18  counts:  8454/ 25 tm: 5.73 s\n",
      " 72  ep_rew:   259.14  loss:  -6.45   vloss:   285.17  counts:  8503/ 27 tm: 5.83 s\n",
      " 73  ep_rew:   232.60  loss:   6.97   vloss:   733.97  counts:  8294/ 29 tm: 5.23 s\n",
      " 74  ep_rew:   231.93  loss:   1.07   vloss:   921.59  counts:  8192/ 28 tm: 5.12 s\n",
      " 75  ep_rew:   257.29  loss:  -5.90   vloss:   296.07  counts:  8254/ 28 tm: 5.23 s\n",
      " 76  ep_rew:   253.99  loss:   4.30   vloss:   250.78  counts:  8429/ 27 tm: 5.57 s\n",
      " 77  ep_rew:   263.70  loss:  -1.76   vloss:   603.14  counts:  8230/ 25 tm: 5.61 s\n",
      " 78  ep_rew:   270.92  loss:  -8.91   vloss:   127.98  counts:  8358/ 29 tm: 5.39 s\n",
      " 79  ep_rew:   247.50  loss:   3.31   vloss:   344.59  counts:  8319/ 30 tm: 5.06 s\n",
      " 80  ep_rew:   258.56  loss:  -1.78   vloss:   449.86  counts:  8456/ 32 tm: 4.95 s\n",
      " 81  ep_rew:   265.75  loss:   9.26   vloss:   477.02  counts:  8270/ 26 tm: 5.11 s\n",
      " 82  ep_rew:   252.28  loss:  -4.97   vloss:   392.96  counts:  8260/ 32 tm: 4.85 s\n",
      " 83  ep_rew:   274.13  loss:  -0.48   vloss:   120.47  counts:  8204/ 30 tm: 4.91 s\n",
      " 84  ep_rew:   264.27  loss:   0.16   vloss:   245.64  counts:  8345/ 32 tm: 4.84 s\n",
      " 85  ep_rew:   268.85  loss:  -0.27   vloss:   229.00  counts:  8334/ 32 tm: 4.81 s\n",
      " 86  ep_rew:   261.91  loss:   2.25   vloss:   371.32  counts:  8359/ 30 tm: 5.04 s\n",
      " 87  ep_rew:   278.82  loss:  -5.33   vloss:    90.83  counts:  8376/ 33 tm: 4.78 s\n",
      " 88  ep_rew:   270.58  loss:   0.44   vloss:   251.73  counts:  8286/ 33 tm: 6.50 s\n",
      " 89  ep_rew:   259.43  loss:   8.46   vloss:   414.98  counts:  8343/ 30 tm: 6.10 s\n",
      " 90  ep_rew:   261.87  loss:  -0.61   vloss:   516.12  counts:  8346/ 33 tm: 4.81 s\n",
      " 91  ep_rew:   273.68  loss:  -2.85   vloss:   244.69  counts:  8365/ 32 tm: 4.84 s\n",
      " 92  ep_rew:   257.98  loss:   1.70   vloss:   340.42  counts:  8229/ 35 tm: 4.65 s\n",
      " 93  ep_rew:   270.89  loss:  -0.08   vloss:    70.11  counts:  8439/ 33 tm: 4.95 s\n",
      " 94  ep_rew:   260.94  loss:   2.81   vloss:   520.69  counts:  8360/ 33 tm: 4.89 s\n",
      " 95  ep_rew:   266.63  loss:  -0.69   vloss:   149.28  counts:  8293/ 34 tm: 4.72 s\n",
      " 96  ep_rew:   236.56  loss:   5.09   vloss:   590.68  counts:  8425/ 35 tm: 4.75 s\n",
      " 97  ep_rew:   248.33  loss:   0.31   vloss:   710.78  counts:  8280/ 34 tm: 4.66 s\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 98  ep_rew:   268.24  loss:  -5.55   vloss:   129.20  counts:  8317/ 33 tm: 4.73 s\n",
      " 99  ep_rew:   256.47  loss:   0.26   vloss:   282.02  counts:  8249/ 34 tm: 4.60 s\n"
     ]
    }
   ],
   "source": [
    "for epoch in range(100):\n",
    "  st = time.perf_counter()\n",
    "  \n",
    "  X,Y,V,P = [],[],[],[]\n",
    "  ll = []\n",
    "  while len(X) < 8192:\n",
    "    obs = env.reset()\n",
    "    rews = []\n",
    "\n",
    "    while True:\n",
    "      # get the action\n",
    "      pred, act = [x[0] for x in pf(obs[None])]\n",
    "      P.append(pred)\n",
    "      \n",
    "      # save this state action pair\n",
    "      X.append(np.copy(obs))\n",
    "      act_mask = np.zeros((env.action_space.n))\n",
    "      act_mask[act] = 1.0\n",
    "      Y.append(act_mask)\n",
    "\n",
    "      # take the action and save the reward\n",
    "      obs, rew, done, _ = env.step(act)\n",
    "      rews.append(rew)\n",
    "\n",
    "      if done:\n",
    "        ll.append(np.sum(rews))\n",
    "        # computes rewards to go with gamma discounting\n",
    "        for i in range(len(rews)-2, -1, -1):\n",
    "          rews[i] += rews[i+1] * gamma\n",
    "        V.extend(rews)\n",
    "        break\n",
    "  X,Y,V,P = [np.array(x) for x in [X,Y,V,P]]\n",
    "\n",
    "  # subtract value baseline to get advantage\n",
    "  A = V - vf(X)[:, 0]\n",
    "  \n",
    "  loss = popt.fit([X,A,P],Y, batch_size=1024, epochs=10, shuffle=True, verbose=0)\n",
    "  loss = loss.history['loss'][-1]\n",
    "  vloss = v.fit(X, V, batch_size=1024, epochs=10, shuffle=True, verbose=0)\n",
    "  vloss = vloss.history['loss'][-1]\n",
    "        \n",
    "  lll.append((epoch, np.mean(ll), loss, vloss, len(X), len(ll), time.perf_counter() - st))\n",
    "  print(\"%3d  ep_rew:%9.2f  loss:%7.2f   vloss:%9.2f  counts: %5d/%3d tm: %.2f s\" % lll[-1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x11d11b128>"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x576 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "figsize(16,8)\n",
    "plt.subplot(1,2,1)\n",
    "plot([x[1] for x in lll], label=\"Mean Episode Reward\")\n",
    "plot([x[2] for x in lll], label=\"Epoch Loss\")\n",
    "plt.legend()\n",
    "plt.subplot(1,2,2)\n",
    "plot([x[3] for x in lll], color='green', label=\"value Loss\")\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "ran 257 steps, got 293.665417 reward\n"
     ]
    }
   ],
   "source": [
    "obs = env.reset()\n",
    "rews = []\n",
    "while True:\n",
    "  env.render()\n",
    "  pred, act = [x[0] for x in pf(obs[None])]\n",
    "  obs, rew, done, _ = env.step(np.argmax(pred))\n",
    "  rews.append(rew)\n",
    "  time.sleep(0.05)\n",
    "  if done:\n",
    "    break\n",
    "print(\"ran %d steps, got %f reward\" % (len(rews), np.sum(rews)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}