mxnet_ppo_test.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "mxnet_ppo_test.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/skywalker0803r/mxnet_course/blob/master/mxnet_ppo_test.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "9FiaX_ohLvWD",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "#!pip install mxnet\n",
        "import numpy as np \n",
        "import mxnet as mx \n",
        "from mxnet import gluon, autograd, nd\n",
        "import mxnet.ndarray as F\n",
        "from mxnet.gluon import nn\n",
        "from mxnet.gluon.loss import L2Loss\n",
        "import gym\n",
        "from mxnet import init\n",
        "import matplotlib.pyplot as plt"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "NET-eqwPgkq_",
        "colab_type": "text"
      },
      "source": [
        "# some function"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "uakkFrEad24J",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def nd_entropy(probs):\n",
        "  return -nd.sum(probs*nd.log(probs+1e-8),axis=1)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "6LIRe6hjgmjV",
        "colab_type": "text"
      },
      "source": [
        "# memory"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Y6KC6RytdufZ",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "class Memory:\n",
        "  def __init__(self):\n",
        "    self.actions = []\n",
        "    self.states = []\n",
        "    self.logprobs = []\n",
        "    self.rewards = []\n",
        "    self.is_terminals = []\n",
        "  def clear_memory(self):\n",
        "    del self.actions[:]\n",
        "    del self.states[:]\n",
        "    del self.logprobs[:]\n",
        "    del self.rewards[:]\n",
        "    del self.is_terminals[:]"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Npwuiud-gn71",
        "colab_type": "text"
      },
      "source": [
        "# model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "yFy1rzTIL5is",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "class ActorCritic_Discrete(gluon.Block):\n",
        "  def __init__(self,action_dim):\n",
        "    super(ActorCritic_Discrete, self).__init__()\n",
        "    self.action_dim = action_dim\n",
        "    with self.name_scope():\n",
        "      self.probs = nn.Sequential()\n",
        "      self.probs.add(nn.Dense(128,activation='relu'))\n",
        "      self.probs.add(nn.Dense(action_dim))\n",
        "      \n",
        "      self.value = nn.Sequential()\n",
        "      self.value.add(nn.Dense(128,activation='relu'))\n",
        "      self.value.add(nn.Dense(1))       \n",
        "  \n",
        "  def choose_action(self, state, memory):\n",
        "    state = nd.array(state)\n",
        "    probs = nd.softmax(self.probs(state))\n",
        "    action, log_prob = nd.sample_multinomial(probs, get_prob = True)\n",
        "    memory.states.append(list(nd.squeeze(state).asnumpy())) \n",
        "    memory.actions.append(action.asnumpy()[0])\n",
        "    memory.logprobs.append(log_prob.asnumpy()[0])\n",
        "    return action.asnumpy()[0]\n",
        "  \n",
        "  def evaluate(self, state, action):\n",
        "    state = nd.array(state)\n",
        "    action = nd.array(action)\n",
        "    probs = nd.softmax(self.probs(state))\n",
        "    action_logprobs = nd.pick(probs.log(),action,axis=1)\n",
        "    state_value = self.value(state)\n",
        "    return action_logprobs, nd.squeeze(state_value), nd_entropy(probs) "
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "xSSE1o9NNpHP",
        "colab_type": "text"
      },
      "source": [
        "# forward test"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "dKrA5ldMNstk",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 84
        },
        "outputId": "b62aa7e0-b428-4116-abe1-bb10c674bdf5"
      },
      "source": [
        "a = ActorCritic_Discrete(action_dim=2)\n",
        "a.initialize()\n",
        "m = Memory()\n",
        "#=============\n",
        "action = a.choose_action([[1,1,1,1]],m)\n",
        "print('action:',action)\n",
        "print('states:',m.states)\n",
        "print('actions:',m.actions)\n",
        "print('logprobs:',m.logprobs)"
      ],
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "action: 1\n",
            "states: [[1.0, 1.0, 1.0, 1.0]]\n",
            "actions: [1]\n",
            "logprobs: [-0.67801327]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "lN6047DPObDH",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 134
        },
        "outputId": "38a7e009-90eb-486b-8e6a-6be7e094e255"
      },
      "source": [
        "a = ActorCritic_Discrete(action_dim=2)\n",
        "a.initialize()\n",
        "m = Memory()\n",
        "#=============\n",
        "probs_log_action,value,entropy = a.evaluate([[1,1,1,1],\n",
        "                                             [1,1,1,1],\n",
        "                                             [1,1,1,1]],[0,1,0])\n",
        "print(probs_log_action,value,entropy)"
      ],
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "\n",
            "[-0.6901691  -0.69613415 -0.6901691 ]\n",
            "<NDArray 3 @cpu(0)> \n",
            "[-0.00307417 -0.00307417 -0.00307417]\n",
            "<NDArray 3 @cpu(0)> \n",
            "[0.6931428 0.6931428 0.6931428]\n",
            "<NDArray 3 @cpu(0)>\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "QxNAy1W-g5gm",
        "colab_type": "text"
      },
      "source": [
        "# ppo"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "AjBVLp_6YhaR",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "class PPO_Discrete(object):\n",
        "  def __init__(self, action_dim, lr, gamma):\n",
        "    \n",
        "    # initialize policy\n",
        "    self.policy = ActorCritic_Discrete(action_dim)\n",
        "    self.policy.probs.initialize(init=init.Xavier())\n",
        "    self.policy.value.initialize(init=init.Xavier())\n",
        "    \n",
        "    # initialize old policy\n",
        "    self.policy_old = ActorCritic_Discrete(action_dim)\n",
        "    self.policy_old.probs.initialize(init=init.Xavier())\n",
        "    self.policy_old.value.initialize(init=init.Xavier())\n",
        "\n",
        "    # initialize forward 4 net\n",
        "    random_x = nd.random_uniform(shape=(1,state_dim))\n",
        "    _ = self.policy.probs(random_x)\n",
        "    _ = self.policy.value(random_x)\n",
        "    _ = self.policy_old.probs(random_x)\n",
        "    _ = self.policy_old.value(random_x)\n",
        "    \n",
        "    # Copy policy params to old policy:\n",
        "    self.policy.probs.save_parameters(\"temp_probs.params\")\n",
        "    self.policy.value.save_parameters(\"temp_value.params\")\n",
        "    self.policy_old.probs.load_parameters(\"temp_probs.params\")\n",
        "    self.policy_old.value.load_parameters(\"temp_value.params\")\n",
        "    print('initialize copy params to old policy')\n",
        "    \n",
        "    self.lr = lr\n",
        "    self.gamma = gamma\n",
        "    self.trainer = gluon.Trainer(self.policy.collect_params(),'adam',{'learning_rate':self.lr})\n",
        "\n",
        "  def update(self, memory): \n",
        "    \n",
        "    # Monte Carlo estimate of state rewards:\n",
        "    rewards = []\n",
        "    discounted_reward = 0\n",
        "    for reward, is_terminal in zip(reversed(memory.rewards), reversed(memory.is_terminals)):\n",
        "      if is_terminal:\n",
        "        discounted_reward = 0\n",
        "      discounted_reward = reward + (self.gamma * discounted_reward)\n",
        "      rewards.insert(0, discounted_reward)\n",
        "    \n",
        "    # Normalizing the rewards:\n",
        "    rewards = np.array(rewards)\n",
        "    rewards = (rewards - rewards.mean()) / (rewards.std() + 1e-5)\n",
        "    \n",
        "    # sample batch data from memory\n",
        "    old_states = memory.states\n",
        "    old_actions = memory.actions\n",
        "    old_logprobs = memory.logprobs\n",
        "\n",
        "    # to ndarray\n",
        "    rewards = nd.array(rewards)\n",
        "    old_logprobs = nd.array(old_logprobs)\n",
        "\n",
        "    # Optimize policy for 4 epochs:\n",
        "    for _ in range(4):\n",
        "      with autograd.record():    \n",
        "        logprobs, state_values, dist_entropy = self.policy.evaluate(old_states, old_actions)\n",
        "        ratios = nd.exp(logprobs - old_logprobs)\n",
        "        advantages = rewards - state_values\n",
        "        surr1 = ratios * advantages\n",
        "        surr2 = nd.clip(ratios, 0.8, 1.2)*advantages\n",
        "        loss = -nd.minimum(surr1,surr2) + 0.5*(advantages)**2 - 0.01*dist_entropy\n",
        "        loss = nd.mean(loss)\n",
        "      # loss backward and trainer step\n",
        "      loss.backward()\n",
        "      self.trainer.step(1)\n",
        "    \n",
        "    # Copy policy params to old policy:\n",
        "    self.policy.probs.save_parameters(\"temp_probs.params\")\n",
        "    self.policy.value.save_parameters(\"temp_value.params\")\n",
        "    self.policy_old.probs.load_parameters(\"temp_probs.params\")\n",
        "    self.policy_old.value.load_parameters(\"temp_value.params\")\n",
        "    print('copy params to old policy')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "E2gVFfDErj2r",
        "colab_type": "text"
      },
      "source": [
        "# env and params"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3pKWwmrPrkCL",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "env_name = \"CartPole-v0\"\n",
        "env = gym.make(env_name)\n",
        "state_dim = env.observation_space.shape[0]\n",
        "action_dim = env.action_space.n\n",
        "render = False\n",
        "solved_reward = 200         \n",
        "log_interval = 10     \n",
        "update_timestep = 2000\n",
        "lr = 0.001\n",
        "gamma = 0.99"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "oNlWFrt0qOOH",
        "colab_type": "text"
      },
      "source": [
        "# main"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hgMq9T-lEkXq",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "6b330c69-741e-44e8-857e-9ff0d73e8dfc"
      },
      "source": [
        "memory = Memory()\n",
        "ppo = PPO_Discrete(action_dim,lr=0.001,gamma=0.99)"
      ],
      "execution_count": 9,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "initialize copy params to old policy\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3Y0fjmcimvg1",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "fb7ad27b-3cda-428b-be57-9cddc155b077"
      },
      "source": [
        "all_rewards = []\n",
        "avg_rewards = []\n",
        "time_step = 0    \n",
        "\n",
        "# run 50000 episode\n",
        "for i_episode in range(50000):\n",
        "  # reset env and R(total_reward)\n",
        "  state = env.reset()\n",
        "  R = 0\n",
        "  # game start \n",
        "  for t in range(1000):\n",
        "    time_step +=1\n",
        "    # transform state to nd_array\n",
        "    state = nd.array(np.expand_dims(state, 0))\n",
        "    # choose action\n",
        "    action = ppo.policy_old.choose_action(state, memory)\n",
        "    # do action on env\n",
        "    state, reward, done, _ = env.step(action)\n",
        "    \n",
        "    # append to memory\n",
        "    R += reward\n",
        "    memory.rewards.append(reward)\n",
        "    memory.is_terminals.append(done)\n",
        "    \n",
        "    # every update_timestep ppo learn\n",
        "    if time_step % update_timestep == 0:\n",
        "      ppo.update(memory)\n",
        "      memory.clear_memory()\n",
        "      time_step = 0\n",
        "    \n",
        "    # game over\n",
        "    if done:\n",
        "      break            \n",
        "  \n",
        "  # every episode record R\n",
        "  all_rewards.append(R)\n",
        "  avg_rewards.append(np.mean(all_rewards[-10:]))\n",
        "  \n",
        "  # if solved!\n",
        "  if (all_rewards[-1] >= solved_reward)&(avg_rewards[-1] >= solved_reward*0.9):\n",
        "    print(\"########## Solved! ##########\")\n",
        "    print(i_episode,all_rewards[-1],avg_rewards[-1])\n",
        "    break\n",
        "  \n",
        "  # print infomation\n",
        "  if i_episode % log_interval == 0:\n",
        "    print(i_episode,avg_rewards[-1])"
      ],
      "execution_count": 10,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "0 10.0\n",
            "10 17.2\n",
            "20 21.6\n",
            "30 19.4\n",
            "40 22.6\n",
            "50 21.6\n",
            "60 21.5\n",
            "70 25.8\n",
            "80 27.2\n",
            "copy params to old policy\n",
            "90 23.7\n",
            "100 28.2\n",
            "110 22.5\n",
            "120 26.0\n",
            "130 20.4\n",
            "140 28.0\n",
            "150 22.5\n",
            "160 24.5\n",
            "170 22.3\n",
            "copy params to old policy\n",
            "180 26.7\n",
            "190 26.6\n",
            "200 25.4\n",
            "210 23.7\n",
            "220 30.2\n",
            "230 26.6\n",
            "240 28.8\n",
            "copy params to old policy\n",
            "250 30.7\n",
            "260 35.1\n",
            "270 28.9\n",
            "280 30.9\n",
            "290 25.6\n",
            "300 39.4\n",
            "copy params to old policy\n",
            "310 28.0\n",
            "320 25.9\n",
            "330 31.6\n",
            "340 26.3\n",
            "350 36.8\n",
            "360 36.0\n",
            "370 30.4\n",
            "copy params to old policy\n",
            "380 37.7\n",
            "390 35.6\n",
            "400 25.1\n",
            "410 32.8\n",
            "420 38.4\n",
            "430 40.5\n",
            "copy params to old policy\n",
            "440 35.0\n",
            "450 42.9\n",
            "460 36.0\n",
            "470 37.0\n",
            "480 28.9\n",
            "copy params to old policy\n",
            "490 48.3\n",
            "500 49.1\n",
            "510 39.1\n",
            "520 36.5\n",
            "530 39.7\n",
            "copy params to old policy\n",
            "540 49.7\n",
            "550 48.3\n",
            "560 51.1\n",
            "570 42.9\n",
            "copy params to old policy\n",
            "580 47.2\n",
            "590 36.1\n",
            "600 44.2\n",
            "610 37.4\n",
            "620 45.7\n",
            "copy params to old policy\n",
            "630 60.5\n",
            "640 52.4\n",
            "650 38.4\n",
            "660 42.9\n",
            "copy params to old policy\n",
            "670 43.2\n",
            "680 54.3\n",
            "690 37.4\n",
            "700 70.9\n",
            "copy params to old policy\n",
            "710 60.9\n",
            "720 45.0\n",
            "730 47.5\n",
            "copy params to old policy\n",
            "740 67.7\n",
            "750 50.0\n",
            "760 56.0\n",
            "770 44.9\n",
            "copy params to old policy\n",
            "780 46.2\n",
            "790 58.1\n",
            "800 53.4\n",
            "810 61.8\n",
            "copy params to old policy\n",
            "820 65.7\n",
            "830 67.2\n",
            "840 54.0\n",
            "copy params to old policy\n",
            "850 58.4\n",
            "860 70.9\n",
            "870 68.2\n",
            "copy params to old policy\n",
            "880 57.1\n",
            "890 65.9\n",
            "900 56.2\n",
            "copy params to old policy\n",
            "910 54.7\n",
            "920 66.9\n",
            "930 73.6\n",
            "copy params to old policy\n",
            "940 75.1\n",
            "950 81.6\n",
            "960 48.5\n",
            "copy params to old policy\n",
            "970 78.1\n",
            "980 67.8\n",
            "990 100.8\n",
            "copy params to old policy\n",
            "1000 83.9\n",
            "1010 114.3\n",
            "copy params to old policy\n",
            "1020 110.3\n",
            "1030 88.1\n",
            "copy params to old policy\n",
            "1040 104.9\n",
            "copy params to old policy\n",
            "1050 114.7\n",
            "1060 140.4\n",
            "copy params to old policy\n",
            "1070 137.7\n",
            "copy params to old policy\n",
            "1080 110.0\n",
            "1090 101.5\n",
            "copy params to old policy\n",
            "1100 129.2\n",
            "1110 144.8\n",
            "copy params to old policy\n",
            "1120 141.5\n",
            "copy params to old policy\n",
            "1130 150.5\n",
            "copy params to old policy\n",
            "1140 148.4\n",
            "1150 174.7\n",
            "copy params to old policy\n",
            "########## Solved! ##########\n",
            "1152 200.0 181.2\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "tkHdkPFtUZOp",
        "colab_type": "text"
      },
      "source": [
        "# plot"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Wm8DVdTB2Lmy",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 279
        },
        "outputId": "da256c85-df8a-4805-9161-d47ba09d6151"
      },
      "source": [
        "plt.plot(all_rewards)\n",
        "plt.plot(avg_rewards)\n",
        "plt.xlabel('episode')\n",
        "plt.show()"
      ],
      "execution_count": 11,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYEAAAEGCAYAAACD7ClEAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nwwsd2k3XegAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-lNsvJstUo7l",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    }
  ]
}