High-level implementation of ActorCritic with minimal typing (General implementation)

minimal_AC_RL.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Based on AC code from https://github.com/kimmyungsup/Reinforcement-Learning-with-Tensorflow-2.0/blob/master/ActorCritic_tf20/a2c_tf20.py\n",
    "import gym\n",
    "from rl.ac_tf2 import ActorModel, CriticModel, ActorCriticTrain, ReplayBuff\n",
    "from rl.ac_tf2 import ac_step, ac_train, ac_report"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "env = gym.make('CartPole-v0')\n",
    "num_action = env.action_space.n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "e :  0  reward :  19.0  step :  20\n",
      "e :  100  reward :  27.0  step :  28\n",
      "e :  200  reward :  100.0  step :  101\n",
      "e :  300  reward :  48.0  step :  49\n",
      "e :  400  reward :  138.0  step :  139\n"
     ]
    }
   ],
   "source": [
    "actor_critic = ActorCriticTrain(num_action)\n",
    "\n",
    "t_end = 500\n",
    "epi = 500\n",
    "train_size = 20\n",
    "        \n",
    "buff = ReplayBuff() \n",
    "\n",
    "state = env.reset()\n",
    "for e in range(epi):\n",
    "    total_reward = 0\n",
    "    for t in range(t_end):        \n",
    "        state, total_reward, done = ac_step(env, actor_critic, buff, state, total_reward, t_end, t)\n",
    "        \n",
    "        ac_train(actor_critic, buff, train_size, done)\n",
    "\n",
    "        if done:\n",
    "            env.reset()\n",
    "            ac_report(actor_critic, total_reward, e, t)\n",
    "            break"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "tf2",
   "language": "python",
   "name": "tf2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}