notebooks/ppo_atari.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "54ab4e32-3ce5-4249-af8a-f14f93e88d2f",
   "metadata": {},
   "source": [
    "# Discrete PPO\n",
    "This is the implementation of PPO + envpool for discrete action space using image inputs."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "72bf0cf6-6d79-4384-a20f-3d0d5d848291",
   "metadata": {},
   "source": [
    "## Setup arguments"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "949e1a15-c17b-498f-be92-cb52e4b750fe",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-10-12T18:10:59.206491Z",
     "iopub.status.busy": "2022-10-12T18:10:59.205560Z",
     "iopub.status.idle": "2022-10-12T18:10:59.229678Z",
     "shell.execute_reply": "2022-10-12T18:10:59.228945Z",
     "shell.execute_reply.started": "2022-10-12T18:10:59.206411Z"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The autoreload extension is already loaded. To reload it, use:\n",
      "  %reload_ext autoreload\n"
     ]
    }
   ],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "9f0521e7-34fc-49e5-84d7-326be26185a6",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-10-12T18:10:59.921103Z",
     "iopub.status.busy": "2022-10-12T18:10:59.920185Z",
     "iopub.status.idle": "2022-10-12T18:11:00.108447Z",
     "shell.execute_reply": "2022-10-12T18:11:00.107850Z",
     "shell.execute_reply.started": "2022-10-12T18:10:59.921022Z"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "0a8b141c64434d4187e8a3731c979d5f",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "IntText(value=1, description='seed', tooltip='seed of the experiment')"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "811735a1a3a84082a422930eceba2852",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Checkbox(value=True, description='torch-deterministic', indent=False, tooltip='if toggled, `torch.backends.cud…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "6124538a8a0c4be4b321e40ebc94320b",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Checkbox(value=True, description='cuda', indent=False, tooltip='if toggled, cuda will be enabled by default')"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "43c1f45c333c47439013f77e4ea604ca",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Checkbox(value=False, description='capture-video', indent=False, tooltip='whether to capture videos of the age…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "66904754c2474866ae90fff0b84c6d1b",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Text(value='Pong-v4', description='env-id', tooltip='the id of the environment')"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "8156d11b0bb14069aaa715653c42077a",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "IntText(value=10000000, description='total-timesteps', tooltip='total timesteps of the experiments')"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "185c533f0403440f80f1cefecb091767",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatText(value=0.00025, description='learning-rate', tooltip='the learning rate of the optimizer')"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "746618cc3a684b92bec1e476923a27f1",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "IntText(value=8, description='num-envs', tooltip='the number of parallel game environments')"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "b36cb84d04d3485199bb7dccde27a20f",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "IntText(value=128, description='num-steps', tooltip='the number of steps to run in each environment per policy…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "f2bff71015f848139cee8ccbf302ac5a",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Checkbox(value=True, description='anneal-lr', indent=False, tooltip='Toggle learning rate annealing for policy…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "fc917fd8aa5c4464ac42572724c9685d",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatText(value=0.99, description='gamma', tooltip='the discount factor gamma')"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "c57e480b9de347459e9fd2765eccc856",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatText(value=0.95, description='gae-lambda', tooltip='the lambda for the general advantage estimation')"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "4e920079ae8b456f91fc2ebd46f1eda8",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "IntText(value=4, description='num-minibatches', tooltip='The number of mini-batches')"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "3cb03e9fe39c4052be8b0c79a354033c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "IntText(value=4, description='update-epochs', tooltip='The K epochs to update the policy')"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "ccd3539046764f559cff28ad4214b0d6",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Checkbox(value=True, description='norm-adv', indent=False, tooltip='Toggles advantages normalization')"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "592fd3589e7a49aab44d041e5cdbad6c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatText(value=0.1, description='clip-coef', tooltip='the surrogate clipping coefficient')"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "0d0d092cacb24cf18b96223c4282310e",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Checkbox(value=True, description='clip-vloss', indent=False, tooltip='Toggles whether or not to use a clipped …"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "299d157d65764f77a27f6f2d1163d33e",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatText(value=0.01, description='ent-coef', tooltip='coefficient of the entropy')"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "8bac172505f24e8caed39711b305f819",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatText(value=0.5, description='vf-coef', tooltip='coefficient of the value function')"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "045c6109d16f4ae8af6010ee3876c33c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatText(value=0.5, description='max-grad-norm', tooltip='the maximum norm for the gradient clipping')"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "28bc12cf77da4a9a92ded38be4994759",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "FloatText(value=0.0, description='target-kl', tooltip='the target KL divergence threshold')"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import ipywidgets as widgets\n",
    "from IPython.display import display, clear_output, Image\n",
    "\n",
    "import time\n",
    "\n",
    "import numpy as np\n",
    "import torch\n",
    "\n",
    "from torch_rl.jupyter_utils import (\n",
    "    WidgetParser,\n",
    "    plot_metrics_notebook\n",
    ")\n",
    "\n",
    "from torch_rl.utils import (\n",
    "    seed_everything,\n",
    "    mp4_to_gif,\n",
    "    MetricStore,\n",
    "    SPS\n",
    ")\n",
    "\n",
    "\n",
    "parser = WidgetParser()\n",
    "parser.add_widget(widgets.IntText(\n",
    "    description='seed',\n",
    "    value=1,\n",
    "    tooltip=\"seed of the experiment\"\n",
    "))\n",
    "\n",
    "parser.add_widget(widgets.Checkbox(\n",
    "    description=\"torch-deterministic\",\n",
    "    value=True,\n",
    "    disabled=False,\n",
    "    indent=False,\n",
    "    tooltip=\"if toggled, `torch.backends.cudnn.deterministic=False`\"\n",
    "))\n",
    "\n",
    "parser.add_widget(widgets.Checkbox(\n",
    "    description=\"cuda\",\n",
    "    value=True,\n",
    "    disabled=False,\n",
    "    indent=False,\n",
    "    tooltip=\"if toggled, cuda will be enabled by default\"\n",
    "))\n",
    "\n",
    "parser.add_widget(widgets.Checkbox(\n",
    "    description=\"capture-video\",\n",
    "    value=False,\n",
    "    disabled=False,\n",
    "    indent=False,\n",
    "    tooltip=\"whether to capture videos of the agent performances (check out `videos` folder)\"\n",
    "))\n",
    "\n",
    "parser.add_widget(widgets.Text(\n",
    "    description=\"env-id\",\n",
    "    value=\"Pong-v4\",\n",
    "    tooltip=\"the id of the environment\"\n",
    "))\n",
    "\n",
    "parser.add_widget(widgets.IntText(\n",
    "    description='total-timesteps',\n",
    "    value=10000000,\n",
    "    tooltip=\"total timesteps of the experiments\"\n",
    "))\n",
    "\n",
    "parser.add_widget(widgets.FloatText(\n",
    "    description='learning-rate',\n",
    "    value=2.5e-4,\n",
    "    tooltip=\"the learning rate of the optimizer\"\n",
    "))\n",
    "\n",
    "parser.add_widget(widgets.IntText(\n",
    "    description='num-envs',\n",
    "    value=8,\n",
    "    tooltip=\"the number of parallel game environments\"\n",
    "))\n",
    "\n",
    "parser.add_widget(widgets.IntText(\n",
    "    description='num-steps',\n",
    "    value=128,\n",
    "    tooltip=\"the number of steps to run in each environment per policy rollout\"\n",
    "))\n",
    "\n",
    "parser.add_widget(widgets.Checkbox(\n",
    "    description=\"anneal-lr\",\n",
    "    value=True,\n",
    "    disabled=False,\n",
    "    indent=False,\n",
    "    tooltip=\"Toggle learning rate annealing for policy and value networks\"\n",
    "))\n",
    "\n",
    "parser.add_widget(widgets.FloatText(\n",
    "    description='gamma',\n",
    "    value=0.99,\n",
    "    tooltip=\"the discount factor gamma\"\n",
    "))\n",
    "\n",
    "parser.add_widget(widgets.FloatText(\n",
    "    description='gae-lambda',\n",
    "    value=0.95,\n",
    "    tooltip=\"the lambda for the general advantage estimation\"\n",
    "))\n",
    "\n",
    "parser.add_widget(widgets.IntText(\n",
    "    description='num-minibatches',\n",
    "    value=4,\n",
    "    tooltip=\"The number of mini-batches\"\n",
    "))\n",
    "\n",
    "parser.add_widget(widgets.IntText(\n",
    "    description='update-epochs',\n",
    "    value=4,\n",
    "    tooltip=\"The K epochs to update the policy\"\n",
    "))\n",
    "\n",
    "parser.add_widget(widgets.Checkbox(\n",
    "    description=\"norm-adv\",\n",
    "    value=True,\n",
    "    disabled=False,\n",
    "    indent=False,\n",
    "    tooltip=\"Toggles advantages normalization\"\n",
    "))\n",
    "\n",
    "parser.add_widget(widgets.FloatText(\n",
    "    description='clip-coef',\n",
    "    value=0.1,\n",
    "    tooltip=\"the surrogate clipping coefficient\"\n",
    "))\n",
    "\n",
    "parser.add_widget(widgets.Checkbox(\n",
    "    description=\"clip-vloss\",\n",
    "    value=True,\n",
    "    disabled=False,\n",
    "    indent=False,\n",
    "    tooltip=\"Toggles whether or not to use a clipped loss for the value function, as per the paper\"\n",
    "))\n",
    "\n",
    "parser.add_widget(widgets.FloatText(\n",
    "    description='ent-coef',\n",
    "    value=0.01,\n",
    "    tooltip=\"coefficient of the entropy\"\n",
    "))\n",
    "\n",
    "parser.add_widget(widgets.FloatText(\n",
    "    description='vf-coef',\n",
    "    value=0.5,\n",
    "    tooltip=\"coefficient of the value function\"\n",
    "))\n",
    "\n",
    "parser.add_widget(widgets.FloatText(\n",
    "    description='max-grad-norm',\n",
    "    value=0.5,\n",
    "    tooltip=\"the maximum norm for the gradient clipping\"\n",
    "))\n",
    "\n",
    "parser.add_widget(widgets.FloatText(\n",
    "    description='target-kl',\n",
    "    value=0,\n",
    "    tooltip=\"the target KL divergence threshold\"\n",
    "))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b9c0283d-5bd9-431e-9a7e-a0a83f5dfc6d",
   "metadata": {},
   "source": [
    "### Parse inputs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "e95c57e4-126e-4cac-9454-6af1863c29d0",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-10-12T18:12:06.330814Z",
     "iopub.status.busy": "2022-10-12T18:12:06.330496Z",
     "iopub.status.idle": "2022-10-12T18:12:06.352741Z",
     "shell.execute_reply": "2022-10-12T18:12:06.351991Z",
     "shell.execute_reply.started": "2022-10-12T18:12:06.330794Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'using device cuda'"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "args = parser.parse()\n",
    "args.batch_size = int(args.num_envs * args.num_steps)\n",
    "args.minibatch_size = int(args.batch_size // args.num_minibatches)\n",
    "\n",
    "run_name = f\"{args.env_id}__ppo__{args.seed}__{int(time.time())}\"\n",
    "device = torch.device(\"cuda\" if torch.cuda.is_available() and args.cuda else \"cpu\")\n",
    "seed_everything(args.seed, args.torch_deterministic)\n",
    "display(f\"using device {device}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0b53fde3-b2aa-45b4-8ac1-f28e6b21b68e",
   "metadata": {},
   "source": [
    "## Setup environment"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "e30e563d-4eb3-4626-941a-f156979190ac",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-10-12T18:12:06.354548Z",
     "iopub.status.busy": "2022-10-12T18:12:06.354224Z",
     "iopub.status.idle": "2022-10-12T18:12:06.376008Z",
     "shell.execute_reply": "2022-10-12T18:12:06.375192Z",
     "shell.execute_reply.started": "2022-10-12T18:12:06.354519Z"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "import gym\n",
    "from torch_rl.atari import (\n",
    "    NoopResetEnv,\n",
    "    MaxAndSkipEnv,\n",
    "    EpisodicLifeEnv,\n",
    "    ClipRewardEnv,\n",
    "    FireResetEnv\n",
    ")\n",
    "\n",
    "\n",
    "def make_env(env_id, seed, idx, capture_video, run_name):\n",
    "    def thunk():\n",
    "        env = gym.make(env_id)\n",
    "        env = gym.wrappers.RecordEpisodeStatistics(env)\n",
    "        if capture_video:\n",
    "            if idx == 0:\n",
    "                env = gym.wrappers.RecordVideo(env, f\"videos/{run_name}\")\n",
    "        env = NoopResetEnv(env, noop_max=30)\n",
    "        env = MaxAndSkipEnv(env, skip=4)\n",
    "        env = EpisodicLifeEnv(env)\n",
    "        if \"FIRE\" in env.unwrapped.get_action_meanings():\n",
    "            env = FireResetEnv(env)\n",
    "        env = ClipRewardEnv(env)\n",
    "        env = gym.wrappers.ResizeObservation(env, (84, 84))\n",
    "        env = gym.wrappers.GrayScaleObservation(env)\n",
    "        env = gym.wrappers.FrameStack(env, 4)\n",
    "        env.seed(seed)\n",
    "        env.action_space.seed(seed)\n",
    "        env.observation_space.seed(seed)\n",
    "        return env\n",
    "\n",
    "    return thunk\n",
    "\n",
    "\n",
    "def layer_init(layer, std=np.sqrt(2), bias_const=0.0):\n",
    "    torch.nn.init.orthogonal_(layer.weight, std)\n",
    "    torch.nn.init.constant_(layer.bias, bias_const)\n",
    "    return layer"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7e6ef50e-adfe-47c7-9745-6105462dd25c",
   "metadata": {},
   "source": [
    "### Ensure environment compatibility"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "39eec428-f195-40b3-8766-780b1574dd2b",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-10-12T18:12:06.377508Z",
     "iopub.status.busy": "2022-10-12T18:12:06.377155Z",
     "iopub.status.idle": "2022-10-12T18:12:08.588221Z",
     "shell.execute_reply": "2022-10-12T18:12:08.587290Z",
     "shell.execute_reply.started": "2022-10-12T18:12:06.377487Z"
    }
   },
   "outputs": [],
   "source": [
    "envs = gym.vector.SyncVectorEnv(\n",
    "        [make_env(args.env_id, args.seed + i, i, args.capture_video, run_name) for i in range(args.num_envs)]\n",
    ")\n",
    "assert isinstance(envs.single_action_space, gym.spaces.Discrete), \"only discrete action space is supported\""
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d564f66b-d7f8-4210-a875-6fe61ec1a11d",
   "metadata": {},
   "source": [
    "### Environment examples"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "a6f7cad5-0983-4d29-8cb8-ee5fb11b34f7",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-10-12T18:12:08.589370Z",
     "iopub.status.busy": "2022-10-12T18:12:08.589079Z",
     "iopub.status.idle": "2022-10-12T18:12:08.974719Z",
     "shell.execute_reply": "2022-10-12T18:12:08.974028Z",
     "shell.execute_reply.started": "2022-10-12T18:12:08.589352Z"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "tmp_env = make_env(args.env_id, args.seed, 0, False, None)()\n",
    "ob = tmp_env.reset()\n",
    "\n",
    "plt.title(args.env_id)\n",
    "plt.imshow(tmp_env.render(\"rgb_array\"))\n",
    "plt.axis('off')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "976f76e2-af45-4107-9610-4115dd23ef1a",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-10-12T18:12:08.976689Z",
     "iopub.status.busy": "2022-10-12T18:12:08.976422Z",
     "iopub.status.idle": "2022-10-12T18:12:09.111800Z",
     "shell.execute_reply": "2022-10-12T18:12:09.111086Z",
     "shell.execute_reply.started": "2022-10-12T18:12:08.976668Z"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "action = tmp_env.action_space.sample()\n",
    "ob, reward, done, info = tmp_env.step(action)\n",
    "plt.title(\"Agent observation (4 frames left to right)\")\n",
    "state_dim = tmp_env.observation_space.shape\n",
    "plt.imshow(np.array(tmp_env.frames).transpose([0,2,1]).reshape([-1,state_dim[1]]).transpose([1,0]))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "b012f3ae-facb-4055-a780-48c887b8d17f",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-10-12T18:12:09.112968Z",
     "iopub.status.busy": "2022-10-12T18:12:09.112730Z",
     "iopub.status.idle": "2022-10-12T18:12:09.134834Z",
     "shell.execute_reply": "2022-10-12T18:12:09.134163Z",
     "shell.execute_reply.started": "2022-10-12T18:12:09.112948Z"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0, (4, 84, 84))"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "action = tmp_env.action_space.sample()\n",
    "ob, reward, done, info = tmp_env.step(action)\n",
    "action, ob.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "76ed0dd4-3fa0-4373-ba59-43973f0fd3cc",
   "metadata": {},
   "source": [
    "## Modeling"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ae74d95f-7c18-45ad-b11e-9b917cc1ab48",
   "metadata": {},
   "source": [
    "### PPO training Wrapper"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "0cfbf718-8113-42f3-a808-aea7e10dd036",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-10-12T18:12:09.135941Z",
     "iopub.status.busy": "2022-10-12T18:12:09.135678Z",
     "iopub.status.idle": "2022-10-12T18:12:09.217806Z",
     "shell.execute_reply": "2022-10-12T18:12:09.217232Z",
     "shell.execute_reply.started": "2022-10-12T18:12:09.135921Z"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "from torch_rl.ppo import PPO, PPOTrainingWrapper\n",
    "agent = PPO(\n",
    "    envs\n",
    ").to(device)\n",
    "\n",
    "wrapper = PPOTrainingWrapper(\n",
    "    model = agent,\n",
    "    env=envs,\n",
    "    device=device\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "45c1cf3c-7a48-4a80-8ebe-4454f0ee6725",
   "metadata": {},
   "source": [
    "### Setup plots"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "882b44b4-0bee-4a02-a743-d1e9e7c5d3d3",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-10-12T18:12:09.221911Z",
     "iopub.status.busy": "2022-10-12T18:12:09.221126Z",
     "iopub.status.idle": "2022-10-12T18:12:09.241294Z",
     "shell.execute_reply": "2022-10-12T18:12:09.240463Z",
     "shell.execute_reply.started": "2022-10-12T18:12:09.221889Z"
    }
   },
   "outputs": [],
   "source": [
    "from torch_rl.utils import SPS\n",
    "\n",
    "\n",
    "def plot_metric(global_step, ms, sps, show_model_metrics=True):\n",
    "    env_metric_list = [\"episodic_return\", \"episodic_length\"]\n",
    "    model_metric_list = [\"value_loss\", \"policy_loss\"]\n",
    "    env_color_list = [\"blue\", \"black\"]\n",
    "    model_color_list = [\"orange\", \"red\"]\n",
    "    if show_model_metrics:\n",
    "        metric_list = env_metric_list + model_metric_list\n",
    "        color_list = env_color_list + model_color_list\n",
    "    else:\n",
    "        metric_list = env_metric_list\n",
    "        color_list = env_color_list\n",
    "\n",
    "    sps.step(global_step)\n",
    "    plot_metrics_notebook(ms, metric_list, color_list, sps=sps)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "89b7d082-ee8c-4343-97c9-d8a4cb9e3029",
   "metadata": {},
   "source": [
    "### PPO training loop"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "21b30a56-e50d-4752-9339-87a1b64e450c",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-10-12T18:12:09.242929Z",
     "iopub.status.busy": "2022-10-12T18:12:09.242664Z",
     "iopub.status.idle": "2022-10-12T23:36:11.585791Z",
     "shell.execute_reply": "2022-10-12T23:36:11.585070Z",
     "shell.execute_reply.started": "2022-10-12T18:12:09.242909Z"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "99% processed at 514 SPS. Remaining: 11 second(s) \n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 4 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import random\n",
    "import torch.optim as optim\n",
    "from collections import deque\n",
    "from IPython.display import display, clear_output\n",
    "\n",
    "\n",
    "optimizer = wrapper.get_optimizer(lr=args.learning_rate, eps=1e-5)\n",
    "ts = wrapper.get_storage(args.num_steps, args.num_envs)\n",
    "\n",
    "ms = MetricStore()\n",
    "sps = SPS(args.total_timesteps)\n",
    "\n",
    "global_step = 0\n",
    "start_time = time.time()\n",
    "next_obs = torch.Tensor(envs.reset()).to(device)\n",
    "next_done = torch.zeros(args.num_envs).to(device)\n",
    "num_updates = args.total_timesteps // args.batch_size\n",
    "\n",
    "for update in range(1, num_updates + 1):\n",
    "    if args.anneal_lr:\n",
    "        lrnow = wrapper.anneal_lr(args.learning_rate, update, num_updates)\n",
    "        optimizer.param_groups[0][\"lr\"] = lrnow\n",
    "\n",
    "    for step in range(0, args.num_steps):\n",
    "        global_step += 1 * args.num_envs\n",
    "        ts.obs[step] = next_obs\n",
    "        ts.dones[step] = next_done\n",
    "\n",
    "        with torch.no_grad():\n",
    "            action, logprob, _, value = agent.get_action_and_value(next_obs)\n",
    "            ts.values[step] = value.flatten()\n",
    "        ts.actions[step] = action\n",
    "        ts.logprobs[step] = logprob\n",
    "\n",
    "        next_obs, reward, done, info = envs.step(action.cpu().numpy())\n",
    "        ts.rewards[step] = torch.tensor(reward).to(device).view(-1)\n",
    "        next_obs, next_done = torch.Tensor(next_obs).to(device), torch.Tensor(done).to(device)\n",
    "\n",
    "        for idx, d in enumerate(done):\n",
    "            if d and info[idx][\"lives\"] == 0:\n",
    "                ms.log_metric(\"episodic_return\", info[idx][\"episode\"][\"r\"], global_step)\n",
    "                ms.log_metric(\"episodic_length\", info[idx][\"episode\"][\"l\"], global_step)\n",
    "\n",
    "    with torch.no_grad():\n",
    "        advantages = wrapper.get_advantages(args.num_steps, next_done, next_obs, args.gamma, args.gae_lambda, ts)\n",
    "        returns = advantages + ts.values\n",
    "\n",
    "    b_obs = ts.obs.reshape((-1,) + envs.single_observation_space.shape)\n",
    "    b_logprobs = ts.logprobs.reshape(-1)\n",
    "    b_actions = ts.actions.reshape((-1,) + envs.single_action_space.shape)\n",
    "    b_advantages = advantages.reshape(-1)\n",
    "    b_returns = returns.reshape(-1)\n",
    "    b_values = ts.values.reshape(-1)\n",
    "\n",
    "    # Optimizing the policy and value network\n",
    "    b_inds = np.arange(args.batch_size)\n",
    "    for epoch in range(args.update_epochs):\n",
    "        np.random.shuffle(b_inds)\n",
    "        for start in range(0, args.batch_size, args.minibatch_size):\n",
    "            end = start + args.minibatch_size\n",
    "            mb_inds = b_inds[start:end]\n",
    "\n",
    "            loss, approx_kl, losses = wrapper.train_step(\n",
    "                b_obs[mb_inds],\n",
    "                b_actions.long()[mb_inds],\n",
    "                b_logprobs[mb_inds],\n",
    "                b_advantages[mb_inds],\n",
    "                b_returns[mb_inds],\n",
    "                b_values[mb_inds],\n",
    "                args.norm_adv,\n",
    "                args.clip_coef,\n",
    "                args.clip_vloss,\n",
    "                args.ent_coef,\n",
    "                args.vf_coef\n",
    "            )\n",
    "\n",
    "            optimizer.zero_grad()\n",
    "            loss.backward()\n",
    "            wrapper.clip_grad_norm(args.max_grad_norm)\n",
    "            optimizer.step()\n",
    "\n",
    "        if args.target_kl is not None and args.target_kl > 0:\n",
    "            if approx_kl > args.target_kl:\n",
    "                break\n",
    "\n",
    "    y_pred, y_true = b_values.cpu().numpy(), b_returns.cpu().numpy()\n",
    "    var_y = np.var(y_true)\n",
    "    explained_var = np.nan if var_y == 0 else 1 - np.var(y_true - y_pred) / var_y\n",
    "\n",
    "    (pg_loss, v_loss, entropy_loss) = losses\n",
    "\n",
    "    ms.log_metric(\"learning_rate\", optimizer.param_groups[0][\"lr\"], global_step)\n",
    "    ms.log_metric(\"value_loss\", v_loss.item(), global_step)\n",
    "    ms.log_metric(\"policy_loss\", pg_loss.item(), global_step)\n",
    "    ms.log_metric(\"entropy\", entropy_loss.item(), global_step)\n",
    "    ms.log_metric(\"approx_kl\", approx_kl.item(), global_step)\n",
    "    ms.log_metric(\"explained_variance\", explained_var, global_step)\n",
    "\n",
    "    if update > 0 and update % 10 == 0:\n",
    "        global_step = update * args.batch_size\n",
    "        plot_metric(global_step, ms, sps)\n",
    "\n",
    "\n",
    "envs.close()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "33f37e7f-dc03-4ba1-9717-67aeabda381f",
   "metadata": {},
   "source": [
    "## Evaluation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "11189f2e-b627-4fd5-8669-f4a9f3b96cf8",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2022-10-12T23:36:11.587091Z",
     "iopub.status.busy": "2022-10-12T23:36:11.586725Z",
     "iopub.status.idle": "2022-10-12T23:36:11.949289Z",
     "shell.execute_reply": "2022-10-12T23:36:11.948350Z",
     "shell.execute_reply.started": "2022-10-12T23:36:11.587070Z"
    },
    "tags": []
   },
   "outputs": [
    {
     "ename": "TypeError",
     "evalue": "unsupported operand type(s) for /: 'LazyFrames' and 'float'",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mTypeError\u001b[0m                                 Traceback (most recent call last)",
      "Cell \u001b[0;32mIn [23], line 6\u001b[0m\n\u001b[1;32m      4\u001b[0m ob \u001b[38;5;241m=\u001b[39m test_env\u001b[38;5;241m.\u001b[39mreset()\n\u001b[1;32m      5\u001b[0m \u001b[38;5;28;01mwhile\u001b[39;00m \u001b[38;5;28;01mTrue\u001b[39;00m:\n\u001b[0;32m----> 6\u001b[0m     action, _, _, _ \u001b[38;5;241m=\u001b[39m \u001b[43magent\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mget_action_and_value\u001b[49m\u001b[43m(\u001b[49m\u001b[43mob\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m      7\u001b[0m     ob, reward, done, info \u001b[38;5;241m=\u001b[39m test_env\u001b[38;5;241m.\u001b[39mstep(action)\n\u001b[1;32m      8\u001b[0m     \u001b[38;5;28;01mif\u001b[39;00m done:\n",
      "File \u001b[0;32m/app/torch_rl/ppo.py:35\u001b[0m, in \u001b[0;36mPPO.get_action_and_value\u001b[0;34m(self, x, action)\u001b[0m\n\u001b[1;32m     34\u001b[0m \u001b[38;5;28;01mdef\u001b[39;00m \u001b[38;5;21mget_action_and_value\u001b[39m(\u001b[38;5;28mself\u001b[39m, x, action\u001b[38;5;241m=\u001b[39m\u001b[38;5;28;01mNone\u001b[39;00m):\n\u001b[0;32m---> 35\u001b[0m     hidden \u001b[38;5;241m=\u001b[39m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mnetwork(\u001b[43mx\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;241;43m/\u001b[39;49m\u001b[43m \u001b[49m\u001b[38;5;241;43m255.0\u001b[39;49m)\n\u001b[1;32m     36\u001b[0m     logits \u001b[38;5;241m=\u001b[39m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mactor(hidden)\n\u001b[1;32m     37\u001b[0m     probs \u001b[38;5;241m=\u001b[39m Categorical(logits\u001b[38;5;241m=\u001b[39mlogits)\n",
      "\u001b[0;31mTypeError\u001b[0m: unsupported operand type(s) for /: 'LazyFrames' and 'float'"
     ]
    }
   ],
   "source": [
    "run_test_name = f\"test-{run_name}\"\n",
    "test_env = make_env(args.env_id, args.seed, 0, True, run_test_name)()\n",
    "\n",
    "ob = test_env.reset()\n",
    "while True:\n",
    "    action, _, _, _ = agent.get_action_and_value(ob)\n",
    "    ob, reward, done, info = test_env.step(action)\n",
    "    if done:\n",
    "        print(f'episodic_return={info[\"episode\"][\"r\"]}, episodic_length={info[\"episode\"][\"l\"]}')\n",
    "        break\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5c048e70-62d2-4d0e-8db7-74420da247cb",
   "metadata": {},
   "source": [
    "### Display full evaluation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "2e1491d9-3a95-4dcb-922e-599296894526",
   "metadata": {
    "execution": {
     "iopub.status.busy": "2022-10-12T23:36:11.949975Z",
     "iopub.status.idle": "2022-10-12T23:36:11.950292Z",
     "shell.execute_reply": "2022-10-12T23:36:11.950160Z",
     "shell.execute_reply.started": "2022-10-12T23:36:11.950134Z"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "mp4_path = f\"videos/{run_test_name}/rl-video-episode-0.mp4\"\n",
    "gif_path = f\"images/ppo_atari_test.gif\"\n",
    "\n",
    "mp4_to_gif(mp4_path, gif_path)\n",
    "Image(url=gif_path, width=500, height=500)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.8.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}