comparing_models.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "e3036796",
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "from utils import evaluate_model_policy\n",
    "from trainer import get_trained_model\n",
    "from environment import StreetFighterEnv\n",
    "from trainer import _get_model\n",
    "from stable_baselines3 import PPO, A2C\n",
    "from actor_critic import A2CCNNPolicy\n",
    "from feature_extractors import CNNExtractorWithAttention, CNNExtractor\n",
    "from layers import ActorCriticLayer\n",
    "import matplotlib\n",
    "import matplotlib.pyplot as plt\n",
    "from tqdm import tqdm\n",
    "from driver import recorder\n",
    "matplotlib.rcParams['figure.figsize'] = (10, 10)\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "da7afa40",
   "metadata": {},
   "outputs": [],
   "source": [
    "# bias_capture_movement = \"models/chunli.state_True_True/best_model.zip\"\n",
    "# bias_no_capture_movement = \"models/chunli.state_True_False/best_model.zip\"\n",
    "# no_bias_capture_movement = \"models/chunli.state_False_True/best_model.zip\"\n",
    "# no_bias_no_capture_movement = \"models/chunli.state_False_False/best_model.zip\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "33bce196",
   "metadata": {},
   "outputs": [],
   "source": [
    "bias_capture_movement1 = \"experiments/final_models/biased_capture_movement/A2C_CHUNLI.zip\"\n",
    "bias_no_capture_movement1 = \"experiments/final_models/biased_no_capture_movement/A2C_CHUNLI.zip\"\n",
    "no_bias_capture_movement1 = \"experiments/final_models/unbiased_capture_movement/A2C_CHUNLI.zip\"\n",
    "no_bias_no_capture_movement1 = \"experiments/final_models/unbiased_no_capture_movement/A2C_CHUNLI.zip\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "e8536892",
   "metadata": {},
   "outputs": [],
   "source": [
    "rewards = []"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7905b231",
   "metadata": {},
   "source": [
    "# MODEL WITH ATTENTION"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "462b58f4",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "model_path: experiments/final_models/biased_capture_movement/A2C_CHUNLI.zip, capture_movement: True, state: chunli.state, model_name: A2C, render: False, record_dir: /tmp/record\n",
      "iterations:  4350\n",
      "total reward:  34377.0\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "34377.0"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "reward = recorder(model_path=bias_capture_movement1, capture_movement=True, state='chunli.state', \n",
    "         model_name=\"A2C\", render=False, record_dir='/tmp/record')\n",
    "rewards.append(reward)\n",
    "reward"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "872ec97b",
   "metadata": {},
   "source": [
    "# MODEL WITH ATTENTION WITHOUT MOVEMENT"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "8a34b075",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "model_path: experiments/final_models/biased_no_capture_movement/A2C_CHUNLI.zip, capture_movement: False, state: chunli.state, model_name: A2C, render: False, record_dir: /tmp/record\n",
      "iterations:  5211\n",
      "total reward:  32392.0\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "32392.0"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "reward = recorder(model_path=bias_no_capture_movement1, capture_movement=False, state='chunli.state', \n",
    "         model_name=\"A2C\", render=False, record_dir='/tmp/record')\n",
    "rewards.append(reward)\n",
    "reward"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6416119b",
   "metadata": {},
   "source": [
    "# MODEL WITHOUT ATTENTION WITH MOVEMENT"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "8393e859",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "model_path: experiments/final_models/unbiased_capture_movement/A2C_CHUNLI.zip, capture_movement: True, state: chunli.state, model_name: A2C, render: False, record_dir: /tmp/record\n",
      "iterations:  7670\n",
      "total reward:  29032.0\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "29032.0"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "reward = recorder(model_path=no_bias_capture_movement1, capture_movement=True, state='chunli.state', \n",
    "         model_name=\"A2C\", render=False, record_dir='/tmp/record')\n",
    "rewards.append(reward)\n",
    "reward"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e8445333",
   "metadata": {},
   "source": [
    "# MODEL WITHOUT ATTENTION WITHOUT MOVEMENT"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "509e3862",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "model_path: experiments/final_models/unbiased_no_capture_movement/A2C_CHUNLI.zip, capture_movement: False, state: chunli.state, model_name: A2C, render: False, record_dir: /tmp/record\n",
      "iterations:  6320\n",
      "total reward:  17023.0\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "17023.0"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "reward = recorder(model_path=no_bias_no_capture_movement1,  capture_movement=False, state='chunli.state', \n",
    "         model_name=\"A2C\", render=False, record_dir='/tmp/record')\n",
    "rewards.append(reward)\n",
    "reward"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "c9954f3d",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[34377.0, 32392.0, 29032.0, 17023.0]"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "rewards"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "5b4d3b3f",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "([0, 1, 2, 3],\n",
       " [Text(0, 0, ''), Text(0, 0, ''), Text(0, 0, ''), Text(0, 0, '')])"
      ]
     },
     "execution_count": 10,
     "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": [
    "fig, ax = plt.subplots()\n",
    "ax.bar(['attention_movement', 'attention_scene', 'cnn_movement', 'cnn_scene'], rewards)\n",
    "plt.xticks(rotation=30)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "f6ec107a",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}