Proximal_Policy_Optimization(PPO).ipynb

{
  "cells": [
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "3w04_bU_4H9g"
      },
      "outputs": [],
      "source": [
        "import gym\n",
        "import torch\n",
        "import torch.nn as nn\n",
        "from torch.distributions import Categorical\n",
        "import matplotlib.pyplot as plt\n",
        "!pip install gym[classic_control]\n",
        "!pip install gym[box2d] #for lunarlander\n",
        "!apt update\n",
        "!apt install xvfb\n",
        "!pip install pyvirtualdisplay\n",
        "!pip install gym-notebook-wrapper\n",
        "import gnwrapper\n",
        "!nvidia-smi\n",
        "print(torch.cuda.is_available())"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "kWwb1Cd_gyJn",
        "outputId": "d24bbc66-7810-4f16-da6b-9f4ac6469117"
      },
      "outputs": [],
      "source": [
        "class Agent(nn.Module):\n",
        "    def __init__(self, input_dim, output_dim, width):\n",
        "        super().__init__()\n",
        "        self.P = PolicyNetwork(input_dim, output_dim, width)\n",
        "        self.P.to(device)\n",
        "        self.P.train()  \n",
        "        self.P_optimizer = torch.optim.Adam(self.P.parameters(), lr=0.0003)\n",
        "        self.V = ValueNetwork(input_dim, output_dim, width)\n",
        "        self.V.to(device)        \n",
        "        self.V.train()\n",
        "        self.V_optimizer = torch.optim.Adam(self.V.parameters(), lr=0.0003)  \n",
        "        self.trajectory = []      \n",
        "        self.env = gym.make(game_name)        \n",
        "\n",
        "    def self_play_old_P(self, max_timestep=1000000):\n",
        "        game_score = 0\n",
        "        state = self.env.reset() \n",
        "        for _ in range(max_timestep):               \n",
        "            old_state = state\n",
        "            output = self.P(torch.from_numpy(state).float().to(device)) \n",
        "            inferenced_v = self.V(torch.from_numpy(state).float().to(device))\n",
        "            prob_distribution = Categorical(output)\n",
        "            action = prob_distribution.sample() \n",
        "            state, r, done, _ = self.env.step(action.item()) \n",
        "            with torch.no_grad():\n",
        "                inferenced_v_from_next_s = self.V(torch.from_numpy(state).float().to(device))\n",
        "            if done==True:\n",
        "                inferenced_v_from_next_s = 0\n",
        "            self.trajectory.append((output, r, inferenced_v, inferenced_v_from_next_s, old_state ,action))\n",
        "            game_score += r \n",
        "            if done:\n",
        "                break\n",
        "        return game_score\n",
        "\n",
        "    def update_weights_PPO(self):\n",
        "        for i in range(5): #epoch \n",
        "            ## without minibatch loop now ##\n",
        "            A = 0 #GAE\n",
        "            lam = 0.9  \n",
        "            eps = 0.1      \n",
        "            self.V_optimizer.zero_grad()\n",
        "            self.P_optimizer.zero_grad()\n",
        "            for old_output, r, v, v_next, old_s, old_a in self.trajectory[::-1]: \n",
        "                with torch.no_grad():           \n",
        "                    delta = r + 0.99* v_next - v\n",
        "                    A += (0.99 * lam) * delta  \n",
        "                    V_target = A + v   \n",
        "\n",
        "                v = self.V(torch.from_numpy(old_s).float().to(device))                \n",
        "                V_loss = (V_target.item() - v)**2         \n",
        "                V_loss.backward()          \n",
        "            \n",
        "                output = self.P(torch.from_numpy(old_s).float().to(device))\n",
        "                upper = output.gather(0,old_a)\n",
        "                lower = old_output.gather(0, old_a)                            \n",
        "                probability_ratio = upper / lower.item() \n",
        "                \n",
        "                L_clip = torch.min(\n",
        "                    probability_ratio * A.item(),\n",
        "                    torch.clip(probability_ratio, 1-eps, 1+eps) * A.item()\n",
        "                )\n",
        "                P_loss = -1 * L_clip \n",
        "                P_loss.backward()             \n",
        "            self.V_optimizer.step()   \n",
        "            self.P_optimizer.step()             \n",
        "        self.trajectory.clear() \n",
        "        return \n",
        "   \n",
        "class PolicyNetwork(nn.Module):  \n",
        "    def __init__(self, input_dim, output_dim, width):\n",
        "        super().__init__()\n",
        "        self.layer1 = torch.nn.Linear(input_dim, width)\n",
        "        self.layer2 = torch.nn.Linear(width, width) \n",
        "        self.layer3 = torch.nn.Linear(width, output_dim) \n",
        "\n",
        "    def forward(self, x):\n",
        "        x = self.layer1(x)\n",
        "        x = torch.nn.functional.relu(x)\n",
        "        x = self.layer2(x)\n",
        "        x = torch.nn.functional.relu(x)\n",
        "        x = self.layer3(x)\n",
        "        x = torch.nn.functional.softmax(x, dim=0)\n",
        "        return x\n",
        "\n",
        "class ValueNetwork(nn.Module): \n",
        "    def __init__(self, input_dim, output_dim, width):\n",
        "        super().__init__()\n",
        "        self.layer1 = torch.nn.Linear(input_dim, width)\n",
        "        self.layer2 = torch.nn.Linear(width, width) \n",
        "        self.layer3 = torch.nn.Linear(width, 1) \n",
        "\n",
        "    def forward(self, x):\n",
        "        x = self.layer1(x)\n",
        "        x = torch.nn.functional.relu(x)\n",
        "        x = self.layer2(x)\n",
        "        x = torch.nn.functional.relu(x)\n",
        "        x = self.layer3(x)\n",
        "        return x\n",
        "\n",
        "device = torch.device('cuda:0') if torch.cuda.is_available() else torch.device('cpu')\n",
        "score_arr = []\n",
        "game_name = 'CartPole-v1' #'LunarLander-v2'   \n",
        "env = gym.make(game_name) \n",
        "agent = Agent(env.observation_space.shape[0], env.action_space.n, 128) \n",
        "print(agent)\n",
        "env.close()\n",
        "\n",
        "#Self play, weight update\n",
        "episode_nums = 300\n",
        "for i in range(episode_nums):    \n",
        "    game_score = agent.self_play_old_P()\n",
        "    agent.update_weights_PPO()    \n",
        "    score_arr.append(game_score)  \n",
        "    if i%50==0 : print('episode', i)    \n",
        "torch.save(agent.state_dict(), 'weights.pt') \n",
        "agent.env.close()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 283
        },
        "id": "_1Cz8LkVy4du",
        "outputId": "21efd51b-8ea2-4593-e0bf-5ba70203b498"
      },
      "outputs": [
        {
          "data": {
            "text/plain": [
              "<matplotlib.legend.Legend at 0x7f59a2d77d50>"
            ]
          },
          "execution_count": 3,
          "metadata": {},
          "output_type": "execute_result"
        },
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          },
          "output_type": "display_data"
        }
      ],
      "source": [
        "#Score\n",
        "plt.plot(score_arr, label ='score')\n",
        "plt.legend(loc='upper left')"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 982
        },
        "id": "KTGyfRFEzUHG",
        "outputId": "2551fe21-ce01-472b-a76b-0f3c8672210f"
      },
      "outputs": [],
      "source": [
        "#play video\n",
        "agent.load_state_dict(torch.load(\"weights.pt\"))\n",
        "env = gnwrapper.LoopAnimation(gym.make(game_name)) \n",
        "state = env.reset()\n",
        "for _ in range(200):\n",
        "    with torch.no_grad():\n",
        "        output = agent.P(torch.from_numpy(state).float().to(device))\n",
        "        prob_distribution = Categorical(output) \n",
        "        action = prob_distribution.sample() \n",
        "    env.render()\n",
        "    state, rew, done, _ = env.step(action.item())\n",
        "    if done:\n",
        "        state = env.reset()\n",
        "env.display()"
      ]
    }
  ],
  "metadata": {
    "accelerator": "GPU",
    "colab": {
      "provenance": []
    },
    "gpuClass": "standard",
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}