DQN_final.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Deep Q Networks (DQN)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 548
    },
    "id": "lIYdn1woOS1n",
    "outputId": "4ac3019d-8647-4407-c64a-9c255256c16b"
   },
   "outputs": [],
   "source": [
    "# Reference: https://www.cs.toronto.edu/~vmnih/docs/dqn.pdf\n",
    "# https://github.com/eemlcommunity/PracticalSessions2020/blob/master/rl/EEML2020_RL_Tutorial.ipynb\n",
    "\n",
    "import torch\n",
    "import torch.nn as nn\n",
    "import torch.optim as optim\n",
    "import torch.nn.functional as F\n",
    "\n",
    "import argparse\n",
    "from distutils.util import strtobool\n",
    "import collections\n",
    "import numpy as np\n",
    "import gym\n",
    "from gym.wrappers import TimeLimit, Monitor\n",
    "from gym.spaces import Discrete, Box, MultiBinary, MultiDiscrete, Space\n",
    "import time\n",
    "import random\n",
    "import os\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "id": "MLLwE9TQ5NUU"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "376 357\n"
     ]
    }
   ],
   "source": [
    "device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')\n",
    "\n",
    "env = gym.make('Humanoid-v2')\n",
    "seed = 1234\n",
    "random.seed(seed)\n",
    "np.random.seed(seed)\n",
    "torch.manual_seed(seed)\n",
    "env.seed(seed)\n",
    "\n",
    "discrete_action_space_dim = 21\n",
    "input_dim = env.observation_space.shape[0]\n",
    "hidden_dim = 765\n",
    "output_dim = env.action_space.shape[0] * discrete_action_space_dim\n",
    "print (input_dim, output_dim)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "def init_weights(m):\n",
    "    if type(m) == nn.Linear:\n",
    "        torch.nn.init.kaiming_normal_(m.weight)\n",
    "        m.bias.data.fill_(0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "id": "k6w-tmg2421Q"
   },
   "outputs": [],
   "source": [
    "# modified from https://github.com/seungeunrho/minimalRL/blob/master/dqn.py#\n",
    "class ReplayBuffer():\n",
    "    def __init__(self, buffer_limit):\n",
    "        self.buffer = collections.deque(maxlen=buffer_limit)\n",
    "    \n",
    "    def put(self, transition):\n",
    "        #print (\"transition: \", len(transition))\n",
    "        self.buffer.append(transition)\n",
    "    \n",
    "    def sample(self, n):\n",
    "        mini_batch = random.sample(self.buffer, n)\n",
    "        s_lst, a_lst, r_lst, s_prime_lst, done_mask_lst = [], [], [], [], []\n",
    "        \n",
    "        for transition in mini_batch:\n",
    "            s, a, r, s_prime, done_mask = transition\n",
    "            s_lst.append(s)\n",
    "            a_lst.append(a)\n",
    "            r_lst.append(r)\n",
    "            s_prime_lst.append(s_prime)\n",
    "            done_mask_lst.append(done_mask)\n",
    "\n",
    "        return np.array(s_lst), np.array(a_lst), \\\n",
    "               np.array(r_lst), np.array(s_prime_lst), \\\n",
    "               np.array(done_mask_lst)\n",
    "    def flush(self):\n",
    "        self.buffer.clear()\n",
    "        \n",
    "    def len(self):\n",
    "        return (len(self.buffer))\n",
    "        \n",
    "\n",
    "# Multi Layer Percpetron architecture, to be used as a non-linear Q value function approximator\n",
    "class QNetwork(nn.Module):\n",
    "    def __init__(self, env):\n",
    "        super(QNetwork, self).__init__()\n",
    "        self.fc_1 = nn.Linear(input_dim, hidden_dim)\n",
    "        self.fc_1a = nn.Linear(hidden_dim, hidden_dim)\n",
    "        self.fc_2 = nn.Linear(hidden_dim, output_dim)\n",
    "        self.maxpool = nn.MaxPool1d(discrete_action_space_dim, stride=discrete_action_space_dim, return_indices=True)\n",
    "\n",
    "    def forward(self, x, device):\n",
    "        x = torch.FloatTensor(x).to(device)\n",
    "        x = torch.tanh(self.fc_1(x))\n",
    "        x = torch.tanh(self.fc_1a(x))\n",
    "        x = self.fc_2(x)\n",
    "        x1, indices = self.maxpool(x)\n",
    "        return x1, indices"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.04\n",
      "21\n"
     ]
    }
   ],
   "source": [
    "# Logic to discretise action space\n",
    "n = discrete_action_space_dim\n",
    "max_action = 0.4\n",
    "min_action = -0.4\n",
    "inc = (max_action - min_action)/ (n-1)\n",
    "print (inc)\n",
    "discrete_action_list = [min_action+inc*i for i in range(0, n)]\n",
    "print (len(discrete_action_list))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {
    "id": "x4KRqZ5I46Ti"
   },
   "outputs": [],
   "source": [
    "# Helper function to get action from action index output of the Neural Net\n",
    "def get_actions(action_indices):\n",
    "    ret_actions = np.zeros(env.action_space.shape[0])\n",
    "    for i in range(0, len(action_indices)):\n",
    "        index = action_indices[i] - i*discrete_action_space_dim\n",
    "        ret_actions[i] = discrete_action_list[index]\n",
    "    return ret_actions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "device(type='cuda')\n",
      "QNetwork(\n",
      "  (fc_1): Linear(in_features=376, out_features=765, bias=True)\n",
      "  (fc_1a): Linear(in_features=765, out_features=765, bias=True)\n",
      "  (fc_2): Linear(in_features=765, out_features=357, bias=True)\n",
      "  (maxpool): MaxPool1d(kernel_size=21, stride=21, padding=0, dilation=1, ceil_mode=False)\n",
      ")\n"
     ]
    }
   ],
   "source": [
    "buffer_size = 100_000\n",
    "lr = 9e-6\n",
    "\n",
    "\n",
    "rb = ReplayBuffer(buffer_size)\n",
    "q_network = QNetwork(env).to(device)\n",
    "target_network = QNetwork(env).to(device)\n",
    "q_network.apply(init_weights)\n",
    "target_network.load_state_dict(q_network.state_dict())\n",
    "\n",
    "optimizer = optim.RMSprop(q_network.parameters(), lr=lr)\n",
    "print(device.__repr__())\n",
    "print(q_network)\n",
    "exploration_fraction = 0.5\n",
    "target_network_frequency = 100\n",
    "batch_size = 64\n",
    "max_grad_norm = 1.0\n",
    "gamma = 0.99\n",
    "\n",
    "\n",
    "# Hyper params\n",
    "n_iters = 30\n",
    "steps_per_iter = 4000\n",
    "phase = 'train'\n",
    "max_step = 1000"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 84,
   "metadata": {
    "id": "Lod0C2yG4dYa"
   },
   "outputs": [],
   "source": [
    "# To run for each episode\n",
    "def agent_run(max_steps, global_step, epsilon, eval_mode):\n",
    "    obs = env.reset()\n",
    "    episode_reward = 0\n",
    "    episode = 0\n",
    "    step_count = 0\n",
    "\n",
    "    for steps in range(max_steps):\n",
    "\n",
    "        # Pick an action using epsilon-greedy policy\n",
    "        if random.random() < epsilon:\n",
    "            action = env.action_space.sample()\n",
    "        else:\n",
    "            q_vals, action_indices = q_network.forward(obs[None, :], device)\n",
    "            action = get_actions(action_indices.view((-1)))\n",
    "            \n",
    "        # Take the chosen action in the env and collect reward\n",
    "        next_obs, reward, done, _ = env.step(action)\n",
    "        episode_reward += reward\n",
    "\n",
    "         # Add the (s, a, s', r, done) tuple to the replay-buffer\n",
    "        rb.put((obs, action, reward, next_obs, done))\n",
    "\n",
    "        # If enough samples in the replay buffer, start training on a randomly sampled batch\n",
    "        if steps -  batch_size > 0:\n",
    "            s_obs, s_actions, s_rewards, s_next_obses, s_dones = rb.sample(batch_size)\n",
    "            with torch.no_grad():\n",
    "                q_next_val, action_next_indices = target_network.forward(s_next_obses, device)\n",
    "                target_max = q_next_val.sum()\n",
    "                # compute target\n",
    "                td_target = torch.Tensor(s_rewards).to(device) + gamma * target_max * (1 - torch.Tensor(s_dones).to(device))\n",
    "\n",
    "            old_q, old_action = q_network.forward(s_obs, device)\n",
    "            # Compute loss using target and the current q value\n",
    "            loss = F.smooth_l1_loss(td_target, torch.sum(old_q, 1))\n",
    "\n",
    "            # optimize the model\n",
    "            optimizer.zero_grad()\n",
    "            loss.backward()\n",
    "            nn.utils.clip_grad_norm_(list(q_network.parameters()), max_grad_norm)\n",
    "            optimizer.step()\n",
    "\n",
    "        # update the target network based on update frequency\n",
    "        if global_step % target_network_frequency == 0 and eval_mode==False:\n",
    "            target_network.load_state_dict(q_network.state_dict())\n",
    "\n",
    "        obs = next_obs\n",
    "        step_count+=1\n",
    "\n",
    "        # Keep interacting until agent reaches a terminal state.\n",
    "        if done:\n",
    "            break\n",
    "    return step_count, episode_reward\n",
    "      "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 85,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "---------------------------------------\n",
      "Iterations: 1\n",
      "Steps: 4005\n",
      "Episodes: 159\n",
      "EpisodeReturn: 177.78\n",
      "AverageReturn: 126.76\n",
      "EvalEpisodes: 10\n",
      "EvalEpisodeReturn: 156.76\n",
      "EvalAverageReturn: 141.78\n",
      "Time: 4\n",
      "---------------------------------------\n",
      "---------------------------------------\n",
      "Iterations: 2\n",
      "Steps: 8029\n",
      "Episodes: 307\n",
      "EpisodeReturn: 195.68\n",
      "AverageReturn: 131.86\n",
      "EvalEpisodes: 10\n",
      "EvalEpisodeReturn: 129.44\n",
      "EvalAverageReturn: 161.61\n",
      "Time: 9\n",
      "---------------------------------------\n",
      "---------------------------------------\n",
      "Iterations: 3\n",
      "Steps: 12032\n",
      "Episodes: 450\n",
      "EpisodeReturn: 261.2\n",
      "AverageReturn: 135.2\n",
      "EvalEpisodes: 10\n",
      "EvalEpisodeReturn: 101.38\n",
      "EvalAverageReturn: 158.5\n",
      "Time: 14\n",
      "---------------------------------------\n",
      "---------------------------------------\n",
      "Iterations: 4\n",
      "Steps: 16034\n",
      "Episodes: 591\n",
      "EpisodeReturn: 119.66\n",
      "AverageReturn: 137.43\n",
      "EvalEpisodes: 10\n",
      "EvalEpisodeReturn: 197.08\n",
      "EvalAverageReturn: 138.48\n",
      "Time: 19\n",
      "---------------------------------------\n",
      "---------------------------------------\n",
      "Iterations: 5\n",
      "Steps: 20056\n",
      "Episodes: 728\n",
      "EpisodeReturn: 238.21\n",
      "AverageReturn: 139.98\n",
      "EvalEpisodes: 10\n",
      "EvalEpisodeReturn: 149.95\n",
      "EvalAverageReturn: 189.75\n",
      "Time: 25\n",
      "---------------------------------------\n",
      "---------------------------------------\n",
      "Iterations: 6\n",
      "Steps: 24072\n",
      "Episodes: 864\n",
      "EpisodeReturn: 184.64\n",
      "AverageReturn: 141.99\n",
      "EvalEpisodes: 10\n",
      "EvalEpisodeReturn: 161.56\n",
      "EvalAverageReturn: 191.62\n",
      "Time: 31\n",
      "---------------------------------------\n",
      "---------------------------------------\n",
      "Iterations: 7\n",
      "Steps: 28085\n",
      "Episodes: 1001\n",
      "EpisodeReturn: 88.66\n",
      "AverageReturn: 143.22\n",
      "EvalEpisodes: 10\n",
      "EvalEpisodeReturn: 113.96\n",
      "EvalAverageReturn: 172.28\n",
      "Time: 37\n",
      "---------------------------------------\n",
      "---------------------------------------\n",
      "Iterations: 8\n",
      "Steps: 32124\n",
      "Episodes: 1129\n",
      "EpisodeReturn: 318.13\n",
      "AverageReturn: 145.53\n",
      "EvalEpisodes: 10\n",
      "EvalEpisodeReturn: 192.42\n",
      "EvalAverageReturn: 173.6\n",
      "Time: 43\n",
      "---------------------------------------\n",
      "---------------------------------------\n",
      "Iterations: 9\n",
      "Steps: 36135\n",
      "Episodes: 1254\n",
      "EpisodeReturn: 190.03\n",
      "AverageReturn: 147.55\n",
      "EvalEpisodes: 10\n",
      "EvalEpisodeReturn: 144.18\n",
      "EvalAverageReturn: 197.06\n",
      "Time: 49\n",
      "---------------------------------------\n",
      "---------------------------------------\n",
      "Iterations: 10\n",
      "Steps: 40168\n",
      "Episodes: 1368\n",
      "EpisodeReturn: 229.35\n",
      "AverageReturn: 150.43\n",
      "EvalEpisodes: 10\n",
      "EvalEpisodeReturn: 205.83\n",
      "EvalAverageReturn: 191.84\n",
      "Time: 55\n",
      "---------------------------------------\n",
      "---------------------------------------\n",
      "Iterations: 11\n",
      "Steps: 44204\n",
      "Episodes: 1482\n",
      "EpisodeReturn: 229.73\n",
      "AverageReturn: 152.92\n",
      "EvalEpisodes: 10\n",
      "EvalEpisodeReturn: 248.03\n",
      "EvalAverageReturn: 234.85\n",
      "Time: 61\n",
      "---------------------------------------\n",
      "---------------------------------------\n",
      "Iterations: 12\n",
      "Steps: 48223\n",
      "Episodes: 1575\n",
      "EpisodeReturn: 258.32\n",
      "AverageReturn: 157.22\n",
      "EvalEpisodes: 10\n",
      "EvalEpisodeReturn: 240.02\n",
      "EvalAverageReturn: 245.83\n",
      "Time: 68\n",
      "---------------------------------------\n",
      "---------------------------------------\n",
      "Iterations: 13\n",
      "Steps: 52235\n",
      "Episodes: 1671\n",
      "EpisodeReturn: 162.77\n",
      "AverageReturn: 160.66\n",
      "EvalEpisodes: 10\n",
      "EvalEpisodeReturn: 260.36\n",
      "EvalAverageReturn: 238.06\n",
      "Time: 74\n",
      "---------------------------------------\n",
      "---------------------------------------\n",
      "Iterations: 14\n",
      "Steps: 56242\n",
      "Episodes: 1760\n",
      "EpisodeReturn: 217.11\n",
      "AverageReturn: 164.41\n",
      "EvalEpisodes: 10\n",
      "EvalEpisodeReturn: 228.02\n",
      "EvalAverageReturn: 217.71\n",
      "Time: 81\n",
      "---------------------------------------\n",
      "---------------------------------------\n",
      "Iterations: 15\n",
      "Steps: 60264\n",
      "Episodes: 1848\n",
      "EpisodeReturn: 274.96\n",
      "AverageReturn: 167.92\n",
      "EvalEpisodes: 10\n",
      "EvalEpisodeReturn: 229.67\n",
      "EvalAverageReturn: 228.16\n",
      "Time: 88\n",
      "---------------------------------------\n",
      "---------------------------------------\n",
      "Iterations: 16\n",
      "Steps: 64294\n",
      "Episodes: 1930\n",
      "EpisodeReturn: 311.4\n",
      "AverageReturn: 171.63\n",
      "EvalEpisodes: 10\n",
      "EvalEpisodeReturn: 293.66\n",
      "EvalAverageReturn: 287.42\n",
      "Time: 96\n",
      "---------------------------------------\n",
      "---------------------------------------\n",
      "Iterations: 17\n",
      "Steps: 68341\n",
      "Episodes: 2006\n",
      "EpisodeReturn: 325.49\n",
      "AverageReturn: 175.66\n",
      "EvalEpisodes: 10\n",
      "EvalEpisodeReturn: 294.02\n",
      "EvalAverageReturn: 296.35\n",
      "Time: 103\n",
      "---------------------------------------\n",
      "---------------------------------------\n",
      "Iterations: 18\n",
      "Steps: 72361\n",
      "Episodes: 2082\n",
      "EpisodeReturn: 278.47\n",
      "AverageReturn: 179.32\n",
      "EvalEpisodes: 10\n",
      "EvalEpisodeReturn: 299.93\n",
      "EvalAverageReturn: 308.66\n",
      "Time: 111\n",
      "---------------------------------------\n",
      "---------------------------------------\n",
      "Iterations: 19\n",
      "Steps: 76377\n",
      "Episodes: 2155\n",
      "EpisodeReturn: 326.38\n",
      "AverageReturn: 182.99\n",
      "EvalEpisodes: 10\n",
      "EvalEpisodeReturn: 315.0\n",
      "EvalAverageReturn: 296.04\n",
      "Time: 119\n",
      "---------------------------------------\n",
      "---------------------------------------\n",
      "Iterations: 20\n",
      "Steps: 80415\n",
      "Episodes: 2228\n",
      "EpisodeReturn: 296.52\n",
      "AverageReturn: 186.48\n",
      "EvalEpisodes: 10\n",
      "EvalEpisodeReturn: 294.98\n",
      "EvalAverageReturn: 317.27\n",
      "Time: 127\n",
      "---------------------------------------\n",
      "---------------------------------------\n",
      "Iterations: 21\n",
      "Steps: 84453\n",
      "Episodes: 2302\n",
      "EpisodeReturn: 280.45\n",
      "AverageReturn: 189.68\n",
      "EvalEpisodes: 10\n",
      "EvalEpisodeReturn: 284.22\n",
      "EvalAverageReturn: 282.22\n",
      "Time: 135\n",
      "---------------------------------------\n",
      "---------------------------------------\n",
      "Iterations: 22\n",
      "Steps: 88461\n",
      "Episodes: 2375\n",
      "EpisodeReturn: 312.22\n",
      "AverageReturn: 192.63\n",
      "EvalEpisodes: 10\n",
      "EvalEpisodeReturn: 278.53\n",
      "EvalAverageReturn: 287.8\n",
      "Time: 143\n",
      "---------------------------------------\n",
      "---------------------------------------\n",
      "Iterations: 23\n",
      "Steps: 92484\n",
      "Episodes: 2446\n",
      "EpisodeReturn: 320.93\n",
      "AverageReturn: 195.61\n",
      "EvalEpisodes: 10\n",
      "EvalEpisodeReturn: 326.58\n",
      "EvalAverageReturn: 315.49\n",
      "Time: 151\n",
      "---------------------------------------\n",
      "---------------------------------------\n",
      "Iterations: 24\n",
      "Steps: 96517\n",
      "Episodes: 2515\n",
      "EpisodeReturn: 253.81\n",
      "AverageReturn: 198.58\n",
      "EvalEpisodes: 10\n",
      "EvalEpisodeReturn: 262.38\n",
      "EvalAverageReturn: 270.59\n",
      "Time: 159\n",
      "---------------------------------------\n",
      "---------------------------------------\n",
      "Iterations: 25\n",
      "Steps: 100521\n",
      "Episodes: 2583\n",
      "EpisodeReturn: 319.88\n",
      "AverageReturn: 201.37\n",
      "EvalEpisodes: 10\n",
      "EvalEpisodeReturn: 291.05\n",
      "EvalAverageReturn: 320.41\n",
      "Time: 167\n",
      "---------------------------------------\n",
      "---------------------------------------\n",
      "Iterations: 26\n",
      "Steps: 104554\n",
      "Episodes: 2643\n",
      "EpisodeReturn: 447.72\n",
      "AverageReturn: 204.62\n",
      "EvalEpisodes: 10\n",
      "EvalEpisodeReturn: 278.86\n",
      "EvalAverageReturn: 328.53\n",
      "Time: 176\n",
      "---------------------------------------\n",
      "---------------------------------------\n",
      "Iterations: 27\n",
      "Steps: 108598\n",
      "Episodes: 2703\n",
      "EpisodeReturn: 321.1\n",
      "AverageReturn: 207.81\n",
      "EvalEpisodes: 10\n",
      "EvalEpisodeReturn: 369.0\n",
      "EvalAverageReturn: 321.14\n",
      "Time: 185\n",
      "---------------------------------------\n",
      "---------------------------------------\n",
      "Iterations: 28\n",
      "Steps: 112600\n",
      "Episodes: 2766\n",
      "EpisodeReturn: 359.14\n",
      "AverageReturn: 210.56\n",
      "EvalEpisodes: 10\n",
      "EvalEpisodeReturn: 313.06\n",
      "EvalAverageReturn: 308.48\n",
      "Time: 193\n",
      "---------------------------------------\n",
      "---------------------------------------\n",
      "Iterations: 29\n",
      "Steps: 116662\n",
      "Episodes: 2822\n",
      "EpisodeReturn: 431.9\n",
      "AverageReturn: 213.7\n",
      "EvalEpisodes: 10\n",
      "EvalEpisodeReturn: 274.44\n",
      "EvalAverageReturn: 316.26\n",
      "Time: 203\n",
      "---------------------------------------\n",
      "---------------------------------------\n",
      "Iterations: 30\n",
      "Steps: 120719\n",
      "Episodes: 2879\n",
      "EpisodeReturn: 340.63\n",
      "AverageReturn: 216.68\n",
      "EvalEpisodes: 10\n",
      "EvalEpisodeReturn: 274.47\n",
      "EvalAverageReturn: 335.62\n",
      "Time: 212\n",
      "---------------------------------------\n"
     ]
    }
   ],
   "source": [
    "### This is the actual train/eval loop.\n",
    "# We run 10 episodes of evaluation for every 4000 steps (steps_per_iter) taken by the agent in the environment during training\n",
    "\n",
    "start_time = time.time()\n",
    "\n",
    "total_num_steps = 0\n",
    "train_sum_returns = 0.\n",
    "train_num_episodes = 0\n",
    "eval_mode = False\n",
    "\n",
    "train_avg_ret = []\n",
    "train_ep_ret = []\n",
    "eval_avg_ret = []\n",
    "eval_ep_ret = []\n",
    "epsilon = 0.9\n",
    "decay_rate = 0.9975\n",
    "\n",
    "for i in range(n_iters):\n",
    "        # Perform the training phase, during which the agent learns\n",
    "        if phase == 'train':\n",
    "            train_step_count = 0\n",
    "            while train_step_count <= steps_per_iter:\n",
    "                eval_mode = False\n",
    "                epsilon = max(epsilon*decay_rate, 0.1)\n",
    "                \n",
    "                # Run one episode\n",
    "                train_step_length, train_episode_return = agent_run(max_step, i, epsilon, eval_mode)\n",
    "                \n",
    "                total_num_steps += train_step_length\n",
    "                train_step_count += train_step_length\n",
    "                train_sum_returns += train_episode_return\n",
    "                train_num_episodes += 1\n",
    "\n",
    "                train_average_return = train_sum_returns / train_num_episodes if train_num_episodes > 0 else 0.0\n",
    "                \n",
    "                train_avg_ret.append(train_average_return)\n",
    "                train_ep_ret.append(train_episode_return)\n",
    "\n",
    "\n",
    "        # Perform the evaluation phase -- no learning\n",
    "        eval_sum_returns = 0.\n",
    "        eval_num_episodes = 0\n",
    "        eval_mode = True\n",
    "\n",
    "        for _ in range(10):\n",
    "            # Run one episode in eval mode\n",
    "            eval_step_length, eval_episode_return = agent_run(max_step, i, 0, eval_mode)\n",
    "            eval_sum_returns += eval_episode_return\n",
    "            eval_num_episodes += 1\n",
    "\n",
    "        eval_average_return = eval_sum_returns / eval_num_episodes if eval_num_episodes > 0 else 0.0\n",
    "        eval_avg_ret.append(eval_average_return)\n",
    "        eval_ep_ret.append(eval_episode_return)\n",
    "\n",
    "\n",
    "        if phase == 'train':\n",
    "            print('---------------------------------------')\n",
    "            print('Iterations:', i + 1)\n",
    "            print('Steps:', total_num_steps)\n",
    "            print('Episodes:', train_num_episodes)\n",
    "            print('EpisodeReturn:', round(train_episode_return, 2))\n",
    "            print('AverageReturn:', round(train_average_return, 2))\n",
    "            print('EvalEpisodes:', eval_num_episodes)\n",
    "            print('EvalEpisodeReturn:', round(eval_episode_return, 2))\n",
    "            print('EvalAverageReturn:', round(eval_average_return, 2))\n",
    "            print('Time:', int(time.time() - start_time))\n",
    "            print('---------------------------------------')\n",
    "        else:\n",
    "            print('---------------------------------------')\n",
    "            print('EvalEpisodes:', eval_num_episodes)\n",
    "            print('EvalEpisodeReturn:', round(eval_episode_return, 2))\n",
    "            print('EvalAverageReturn:', round(eval_average_return, 2))\n",
    "            print('Time:', int(time.time() - start_time))\n",
    "            print('---------------------------------------')\n",
    "            "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "metadata": {
    "id": "VEMSnKf9lsuB",
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'train episode rewards')"
      ]
     },
     "execution_count": 86,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.show()\n",
    "plt.plot(train_avg_ret[:])\n",
    "plt.xlabel(\"Episodes\")\n",
    "plt.ylabel(\"Rewards\")\n",
    "plt.title(\"train average rewards\")\n",
    "\n",
    "plt.show()\n",
    "plt.plot(train_ep_ret[:])\n",
    "plt.xlabel(\"Episodes\")\n",
    "plt.ylabel(\"Rewards\")\n",
    "plt.title(\"train episode rewards\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "metadata": {
    "id": "W6vLJACbE6Fu"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'test episode rewards')"
      ]
     },
     "execution_count": 88,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.show()\n",
    "plt.plot(eval_avg_ret)\n",
    "plt.xlabel(\"Episodes\")\n",
    "plt.ylabel(\"Rewards\")\n",
    "plt.title(\"test average rewards\")\n",
    "\n",
    "plt.show()\n",
    "plt.plot(eval_ep_ret)\n",
    "plt.xlabel(\"Episodes\")\n",
    "plt.ylabel(\"Rewards\")\n",
    "plt.title(\"test episode rewards\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {},
   "outputs": [],
   "source": [
    "flag = \"\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "12:11:2021_17:32:58\n"
     ]
    }
   ],
   "source": [
    "from datetime import datetime\n",
    "\n",
    "now = datetime.now() # current date and time\n",
    "date_time = str(now.strftime(\"%m:%d:%Y_%H:%M:%S\"))\n",
    "print (date_time)\n",
    "\n",
    "\n",
    "algo = \"DQN\"\n",
    "path = \"Results/\" + algo + date_time + flag\n",
    "np.save(path +\"_train_avg\", train_avg_ret)\n",
    "np.save(path + \"_test_avg\", eval_avg_ret)\n",
    "np.save(path +\"_train_ep\", train_ep_ret)\n",
    "np.save(path + \"_test_ep\", eval_ep_ret)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 91,
   "metadata": {},
   "outputs": [],
   "source": [
    "torch.save(q_network, path+\"model\")\n",
    "torch.save(target_network, path+\"model\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "collapsed_sections": [],
   "machine_shape": "hm",
   "name": "Copy of dqn_working.ipynb",
   "provenance": []
  },
  "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": "3.7.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}