leduc_holdem_cfr.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "leduc_holdem_cfr.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/mia1996/rlcard-tutoirial/blob/master/leduc_holdem_cfr.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "miBl4S8JARzX"
      },
      "source": [
        "\n",
        "\n",
        "# <a href='https://github.com/datamllab/rlcard'> <center> <img src='https://miro.medium.com/max/1000/1*_9abDpNTM9Cbsd2HEXYm9Q.png' width=500 class='center' /></a> \n",
        "\n",
        "## **Training CFR on Leduc Hold'em**\n",
        "In this tutorial, we will showcase a more advanced algorithm CFR, which uses `step` and `step_back` to traverse the game tree."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "DvSQRtFHfQde"
      },
      "source": [
        "First, we install RLcard and PyTorch."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zQ8CiXAJjQGi",
        "outputId": "f62f9be2-2b1e-4d34-a254-a5ddef2f5962",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "!pip install rlcard[torch]"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Collecting rlcard[torch]\n",
            "  Downloading rlcard-1.0.7.tar.gz (268 kB)\n",
            "\u001b[?25l\r\u001b[K     |█▏                              | 10 kB 13.5 MB/s eta 0:00:01\r\u001b[K     |██▍                             | 20 kB 14.3 MB/s eta 0:00:01\r\u001b[K     |███▋                            | 30 kB 9.7 MB/s eta 0:00:01\r\u001b[K     |████▉                           | 40 kB 5.0 MB/s eta 0:00:01\r\u001b[K     |██████                          | 51 kB 4.7 MB/s eta 0:00:01\r\u001b[K     |███████▎                        | 61 kB 5.5 MB/s eta 0:00:01\r\u001b[K     |████████▌                       | 71 kB 5.5 MB/s eta 0:00:01\r\u001b[K     |█████████▊                      | 81 kB 5.3 MB/s eta 0:00:01\r\u001b[K     |███████████                     | 92 kB 5.9 MB/s eta 0:00:01\r\u001b[K     |████████████▏                   | 102 kB 3.0 MB/s eta 0:00:01\r\u001b[K     |█████████████▍                  | 112 kB 3.0 MB/s eta 0:00:01\r\u001b[K     |██████████████▋                 | 122 kB 3.0 MB/s eta 0:00:01\r\u001b[K     |███████████████▉                | 133 kB 3.0 MB/s eta 0:00:01\r\u001b[K     |█████████████████               | 143 kB 3.0 MB/s eta 0:00:01\r\u001b[K     |██████████████████▎             | 153 kB 3.0 MB/s eta 0:00:01\r\u001b[K     |███████████████████▌            | 163 kB 3.0 MB/s eta 0:00:01\r\u001b[K     |████████████████████▊           | 174 kB 3.0 MB/s eta 0:00:01\r\u001b[K     |██████████████████████          | 184 kB 3.0 MB/s eta 0:00:01\r\u001b[K     |███████████████████████▏        | 194 kB 3.0 MB/s eta 0:00:01\r\u001b[K     |████████████████████████▍       | 204 kB 3.0 MB/s eta 0:00:01\r\u001b[K     |█████████████████████████▋      | 215 kB 3.0 MB/s eta 0:00:01\r\u001b[K     |██████████████████████████▉     | 225 kB 3.0 MB/s eta 0:00:01\r\u001b[K     |████████████████████████████    | 235 kB 3.0 MB/s eta 0:00:01\r\u001b[K     |█████████████████████████████▎  | 245 kB 3.0 MB/s eta 0:00:01\r\u001b[K     |██████████████████████████████▌ | 256 kB 3.0 MB/s eta 0:00:01\r\u001b[K     |███████████████████████████████▊| 266 kB 3.0 MB/s eta 0:00:01\r\u001b[K     |████████████████████████████████| 268 kB 3.0 MB/s \n",
            "\u001b[?25hRequirement already satisfied: numpy>=1.16.3 in /usr/local/lib/python3.7/dist-packages (from rlcard[torch]) (1.21.5)\n",
            "Requirement already satisfied: termcolor in /usr/local/lib/python3.7/dist-packages (from rlcard[torch]) (1.1.0)\n",
            "Requirement already satisfied: torch in /usr/local/lib/python3.7/dist-packages (from rlcard[torch]) (1.10.0+cu111)\n",
            "Collecting GitPython\n",
            "  Downloading GitPython-3.1.27-py3-none-any.whl (181 kB)\n",
            "\u001b[K     |████████████████████████████████| 181 kB 45.4 MB/s \n",
            "\u001b[?25hCollecting gitdb2\n",
            "  Downloading gitdb2-4.0.2-py3-none-any.whl (1.1 kB)\n",
            "Requirement already satisfied: matplotlib in /usr/local/lib/python3.7/dist-packages (from rlcard[torch]) (3.2.2)\n",
            "Collecting gitdb>=4.0.1\n",
            "  Downloading gitdb-4.0.9-py3-none-any.whl (63 kB)\n",
            "\u001b[K     |████████████████████████████████| 63 kB 1.4 MB/s \n",
            "\u001b[?25hCollecting smmap<6,>=3.0.1\n",
            "  Downloading smmap-5.0.0-py3-none-any.whl (24 kB)\n",
            "Requirement already satisfied: typing-extensions>=3.7.4.3 in /usr/local/lib/python3.7/dist-packages (from GitPython->rlcard[torch]) (3.10.0.2)\n",
            "Requirement already satisfied: kiwisolver>=1.0.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib->rlcard[torch]) (1.4.0)\n",
            "Requirement already satisfied: python-dateutil>=2.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib->rlcard[torch]) (2.8.2)\n",
            "Requirement already satisfied: cycler>=0.10 in /usr/local/lib/python3.7/dist-packages (from matplotlib->rlcard[torch]) (0.11.0)\n",
            "Requirement already satisfied: pyparsing!=2.0.4,!=2.1.2,!=2.1.6,>=2.0.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib->rlcard[torch]) (3.0.7)\n",
            "Requirement already satisfied: six>=1.5 in /usr/local/lib/python3.7/dist-packages (from python-dateutil>=2.1->matplotlib->rlcard[torch]) (1.15.0)\n",
            "Building wheels for collected packages: rlcard\n",
            "  Building wheel for rlcard (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
            "  Created wheel for rlcard: filename=rlcard-1.0.7-py3-none-any.whl size=325373 sha256=e59625c8a510c5eecf50ecc571cc57be45f37daf110a3cd1d59996d1ff298f64\n",
            "  Stored in directory: /root/.cache/pip/wheels/8a/90/bd/bc402a48ca90970c9a7c2c4387dcb885fdf6073ec231a605ad\n",
            "Successfully built rlcard\n",
            "Installing collected packages: smmap, gitdb, rlcard, GitPython, gitdb2\n",
            "Successfully installed GitPython-3.1.27 gitdb-4.0.9 gitdb2-4.0.2 rlcard-1.0.7 smmap-5.0.0\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Then we import all the classes and functions we need."
      ],
      "metadata": {
        "id": "0ToWVZr881JH"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "import rlcard\n",
        "from rlcard.agents import (\n",
        "    CFRAgent,\n",
        "    RandomAgent,\n",
        ")\n",
        "from rlcard.utils import (\n",
        "    tournament,\n",
        "    Logger,\n",
        "    plot_curve,\n",
        ")"
      ],
      "metadata": {
        "id": "pfmdahQ_86_W"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "We make two environments, where one allows `step_back` so that CFR can traverse the tree, and the other for evaluation only."
      ],
      "metadata": {
        "id": "dRXGan5w9BLT"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "env = rlcard.make(\n",
        "        'leduc-holdem',\n",
        "        config={\n",
        "            'allow_step_back': True,\n",
        "        }\n",
        "    )\n",
        "eval_env = rlcard.make(\n",
        "    'leduc-holdem',\n",
        ")"
      ],
      "metadata": {
        "id": "oN7k8icR89iB"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "We create the CFR agent."
      ],
      "metadata": {
        "id": "qd5fe5W_9Ukz"
      }
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "4_A_Br3Jj0xW"
      },
      "source": [
        "agent = CFRAgent(\n",
        "    env,\n",
        "    \"experiments/leduc_holdem_cfr_result/cfr_model\",\n",
        ")"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "Here, we save the trained model in the path `experiments/leduc_holdem_cfr_result/cfr_model`. Then we use a random agent as the opponent."
      ],
      "metadata": {
        "id": "yYvb4MDz9qH6"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "eval_env.set_agents([\n",
        "    agent,\n",
        "    RandomAgent(num_actions=env.num_actions),\n",
        "])"
      ],
      "metadata": {
        "id": "C4cq5mA-90W_"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "QOMi0W94fSH7"
      },
      "source": [
        "Now we start training for `1000` iterations, i.e., 1000 games."
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "with Logger(\"experiments/leduc_holdem_cfr_result\") as logger:\n",
        "    for episode in range(1000):\n",
        "        agent.train()\n",
        "        print('\\rIteration {}'.format(episode), end='')\n",
        "        # Evaluate the performance. Play with Random agents.\n",
        "        if episode % 50 == 0:\n",
        "            logger.log_performance(\n",
        "                env.timestep,\n",
        "                tournament(\n",
        "                    eval_env,\n",
        "                    10000,\n",
        "                )[0]\n",
        "            )\n",
        "\n",
        "    # Get the paths\n",
        "    csv_path, fig_path = logger.csv_path, logger.fig_path"
      ],
      "metadata": {
        "id": "MTbACbO4-D_c",
        "outputId": "f3841a04-9990-405c-a162-846f97abf816",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Iteration 0\n",
            "----------------------------------------\n",
            "  timestep     |  192\n",
            "  reward       |  5e-05\n",
            "----------------------------------------\n",
            "Iteration 50\n",
            "----------------------------------------\n",
            "  timestep     |  9792\n",
            "  reward       |  0.3098\n",
            "----------------------------------------\n",
            "Iteration 100\n",
            "----------------------------------------\n",
            "  timestep     |  19392\n",
            "  reward       |  0.43155\n",
            "----------------------------------------\n",
            "Iteration 150\n",
            "----------------------------------------\n",
            "  timestep     |  28992\n",
            "  reward       |  0.5922\n",
            "----------------------------------------\n",
            "Iteration 200\n",
            "----------------------------------------\n",
            "  timestep     |  38592\n",
            "  reward       |  0.5821\n",
            "----------------------------------------\n",
            "Iteration 250\n",
            "----------------------------------------\n",
            "  timestep     |  48192\n",
            "  reward       |  0.62535\n",
            "----------------------------------------\n",
            "Iteration 300\n",
            "----------------------------------------\n",
            "  timestep     |  57792\n",
            "  reward       |  0.70355\n",
            "----------------------------------------\n",
            "Iteration 350\n",
            "----------------------------------------\n",
            "  timestep     |  67392\n",
            "  reward       |  0.7055\n",
            "----------------------------------------\n",
            "Iteration 400\n",
            "----------------------------------------\n",
            "  timestep     |  76992\n",
            "  reward       |  0.7063\n",
            "----------------------------------------\n",
            "Iteration 450\n",
            "----------------------------------------\n",
            "  timestep     |  86592\n",
            "  reward       |  0.7075\n",
            "----------------------------------------\n",
            "Iteration 500\n",
            "----------------------------------------\n",
            "  timestep     |  96192\n",
            "  reward       |  0.70215\n",
            "----------------------------------------\n",
            "Iteration 550\n",
            "----------------------------------------\n",
            "  timestep     |  105792\n",
            "  reward       |  0.71865\n",
            "----------------------------------------\n",
            "Iteration 600\n",
            "----------------------------------------\n",
            "  timestep     |  115392\n",
            "  reward       |  0.74505\n",
            "----------------------------------------\n",
            "Iteration 650\n",
            "----------------------------------------\n",
            "  timestep     |  124992\n",
            "  reward       |  0.72085\n",
            "----------------------------------------\n",
            "Iteration 700\n",
            "----------------------------------------\n",
            "  timestep     |  134592\n",
            "  reward       |  0.7117\n",
            "----------------------------------------\n",
            "Iteration 750\n",
            "----------------------------------------\n",
            "  timestep     |  144192\n",
            "  reward       |  0.71705\n",
            "----------------------------------------\n",
            "Iteration 800\n",
            "----------------------------------------\n",
            "  timestep     |  153792\n",
            "  reward       |  0.7401\n",
            "----------------------------------------\n",
            "Iteration 850\n",
            "----------------------------------------\n",
            "  timestep     |  163392\n",
            "  reward       |  0.7587\n",
            "----------------------------------------\n",
            "Iteration 900\n",
            "----------------------------------------\n",
            "  timestep     |  172992\n",
            "  reward       |  0.75715\n",
            "----------------------------------------\n",
            "Iteration 950\n",
            "----------------------------------------\n",
            "  timestep     |  182592\n",
            "  reward       |  0.7278\n",
            "----------------------------------------\n",
            "Iteration 999\n",
            "Logs saved in experiments/leduc_holdem_cfr_result\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "We can plot the learning curve"
      ],
      "metadata": {
        "id": "d-qU4hS4-hI5"
      }
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "l_8Kuf47kghG",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 279
        },
        "outputId": "f6bc9709-e500-492c-9e88-348532ed5450"
      },
      "source": [
        "plot_curve(csv_path, fig_path, 'cfr')"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "1QRAbZ3vfwtW"
      },
      "source": [
        "Good job! Now you have your trained CFR agent on Leduc Hold'em!"
      ]
    }
  ]
}