data-generator/index.html

<!--
 @license
 Copyright 2019 Google LLC. All Rights Reserved.
 Licensed under the Apache License, Version 2.0 (the "License");
 you may not use this file except in compliance with the License.
 You may obtain a copy of the License at
  http://www.apache.org/licenses/LICENSE-2.0
  Unless required by applicable law or agreed to in writing, software
 distributed under the License is distributed on an "AS IS" BASIS,
 WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License.
 =============================================================================
-->

<html>

<head>
  <link rel="stylesheet" href="../shared/tfjs-examples.css" />
  <style>
    #train {
      margin-top: 10px;
    }

    label {
      display: inline-block;
      width: 250px;
      padding: 6px 0 6px 0;
    }

    .canvases {
      display: inline-block;
    }


    .status {
      border: none;
      font-size: 20px;
      text-align: center;
      resize: none;
    }

    .query {
      font-size: 20px;
      text-align: left;
      resize: none;
    }

    .textarea {
      font-size: 20px;
      text-align: center;
      resize: none;
    }

    .textoutput {
      font-size: 20px;
      text-align: left;
      resize: none;
    }

    .container {
      width: 80%;
      margin-left: auto;
      margin-right: auto;
      text-align: center;
    }

    .divTable {
      display: table;
      width: 100%;
    }

    .divTableRow {
      display: table-row;
    }

    .divTableHeading {
      background-color: #EEE;
      display: table-header-group;
    }

    .divTableCell,
    .divTableHead {
      border: 1px solid #999999;
      display: table-cell;
      padding: 3px 10px;
    }

    .divTableCellKey {
      border: 1px solid #999999;
      display: table-cell;
      padding: 3px 10px;
      font-weight: bold;
    }

    .divTableHeading {
      background-color: #EEE;
      display: table-header-group;
      font-weight: bold;
    }

    .divTableFoot {
      background-color: #EEE;
      display: table-footer-group;
      font-weight: bold;
    }

    .divTableBody {
      display: table-row-group;
    }
  </style>
</head>

<body>
  <div class="container">
    <section class='title-area'>
      <h3>TensorFlow.js: tfjs-data</h3>
      Building a tf.data.Dataset from a generator function.
    </section>
  </div>
  <p class='section-head'>How to use this demo.</p>

  This demo is meant to illustrate how to train a model using a tf.data.Dataset
  constructed from a generator using <pre>tf.data.generator()</pre>.
  In this case, we will predict the likelihood of
  winning a game. The game itself is described below, and implemented in game.
  js.The game exports a function we will use as a "generator" function to
  create a dataset. We will then fit a model to this dataset using
  <pre>model.fitDataset()</pre>, with the goal of predicting
  the "win" state from features calculated from Player 1's hand.  The output
  logit value will be larger for hands which the model considers more likely
  to win.

  <p class='section-head'>How to use this app.</p>
  <ol>
    <li> Familiarize yourself with the rules of the game below.</li>
    <li> Simulate the game by pressing the <i>simulate-game</i> button.</li>
    <li> Select how many cards each player gets using the drop-down menu.</li>
    <li> See how games are generated, wins are calculated, and notice that the
      simulation count increases.</li>
    <li> Familiarize yourself with the tensor-valued features calculated from
      player 1's hand.</li>
    <li> See what batches of data looks like by clicking dataset-to-array.</li>
    <li> Before training a model, select a batchSize, the number of
      batchesPerEpoch, and how many epochs to train on.</li>
    <li> When you are ready, train a model on the generated features by
      clicking <i>'train-model-using-fit-dataset'</i>.</li>
    <li> Once the model is trained, make predictions by entering values
      for the user's hand, and clicking predict.</li>
  </ol>
  <div class="divTable">
    <div class="divTableRow">
      <!-- Left hand pane -->
      <div class="divTableCell" style='width:50%'>
        <p></p>
        <p class='section-head'>Game Simulation</p>
        Click "Simulate Game" to run one play of the game. Three numbers will be
        randomly selected for each player. The "win" status will indicate
        whether player one's 'win' status according to the following rules..
        <p></p>
        Rules:
        <ul>
          <li>The player with the largest group of same-valued cards wins. E.g.,
            if player 1 has three-of-a-kind, and player 2 only has a pair,
            player 1 wins.</li>
          <li>If both players have the same sized maximal group, then the player
            with the group with the largest face
            value wins. E.g., A pair of 5s beats a pair of 4s.</li>
          <li>If neither player even has a pair, the player with the highest
            single card wins.</li>
          <li>Ties are settled randomly, 50/50.</li>
        </ul>
        Number of cards per hand
        <select id='select-cards-per-hand'>
          <option value="3">3</option>
          <option value="4">4</option>
          <option value="5">5</option>
          <option value="6">6</option>
          <option value="7">7</option>
          <option value="8">8</option>
        </select>
        <p></p>
        <button id='simulate-game'>simulate-game</button>
        <h3>Simulation Results (Simulations so far =
          <div style=display:inline id=num-simulations-so-far>0</div>)
        </h3>
        <p></p>
        <div class='divTable'>
          <div class='divTableRow' id='sim-features-row'>
            <div id='player1Cell' , class='divTableCell' style='width:40%'>
              <b>player 1</b>
              <div class='divTable'>
                <div class='divTableRow' id='player1-row'>
                </div>
              </div>
            </div>
            <div id='player2Cell' , class='divTableCell' style='width:40%'>
              <b>opponent</b>
              <div class='divTable'>
                <div class='divTableRow' id='player2-row'>
                </div>
              </div>
            </div>
            <div class='divTableCell' style='width:10%'><b>win?</b>
              <div class='divTable'>
                <div class='divTableRow' id='result-row'>
                </div>
              </div>
            </div>
          </div>
        </div>
        <p></p>
        <div class='divTable'>
          <div class='divTableRow'>
            <div class='divTableCell' style='width:100%'>
              <b>Game to features and label.</b> Note that the features fed into
              the model only include values visible to player
              1, since we want to predict whether player 1 will win.
              <p></p>
            </div>
          </div>
        </div>
        <div class='divTable'>
          <div class='divTableRow'>
            <div class='divTableCell' style='width:20%'><b>Features:</b> </div>
            <div class='divTableCell' style='width:80%' id='sim-features'>
              Click "sample" to run
            </div>
          </div>
          <div class='divTableRow'>
            <div class='divTableCell' style='width:20%'><b>Label:</b> </div>
            <div class='divTableCell' style='width:80%' id='sim-label'></div>
          </div>
        </div>
        <div id="generated-sample-data-message"></div>
        <p class='section-head'>Data Pipeline</p>
        <h4>tf.dataFromGenerator(simulation↑)</h4>
        <h4>.map(gameToFeaturesAndLabel)</h4>
        <h4>.batch(
          <input type="number" id="generator-batch" value="100">)</h4>
        <h4>.take(
          <input type="number" id="generator-take" value="5">)</h4>
        <h4>.toArray()
          <button id="dataset-to-array">dataset-to-array</button>
        </h4>
        <p></p>
        <div class="container" id="to-array-container">
        </div>
      </div>
      <!-- Right hand pane -->
      <div class="divTableCell" style='width:50%'>
        <p class='section-head'>Train & Evaluate Model</p>
        <div id="train-model-message"></div>
        <h4>batchesPerEpoch
          <input type="number" id="batches-per-epoch" value="50">
          <h4>Epochs to train
            <input type="number" id="epochs-to-train" value="50">
          </h4>
          Expected simulations = batchSize * batchesPerEpoch * epochs =
          <div id='expected-simulations' style='display:inline'></div>
          <button id="train-model-using-fit-dataset">
            train-model-using-fit-dataset
          </button>
          <button id="stop-training" disabled="true">stop-training</button>

          <section>
            <p class='section-head'>Training Progress</p>
            <p id="training-status"></p>
            <p id="training-message"></p>
            <div id="training-stats">
              <div class="canvases">
                <label id="training-loss-label"></label>
                <div id="training-loss-canvas"></div>
              </div>
              <div class="canvases">
                <label id="training-accuracy-label"></label>
                <div id="training-accuracy-canvas"></div>
              </div>
            </div>
            Note that since each player has an equal chance of winning, we
            expect that a completely naive estimator will have an accuracy of
            0.5. An estimator with perfect accuracy is not possible, since the
            estimator does not have access to the opponent player's hand.
          </section>

          <section>
            <p class='section-head'>Use Trained Model</p>
            <div id='prediction-input'></div>
            <button id="predict" disabled="true">predict</button>
            Output of model:
            <div class="container" id="prediction" style='display:inline'> </div>
            <p></p>
            Note that this prediction is larger for hands that the model
            considers more likely to win, but are not calibrated probabilities.
          </section>

      </div>
    </div>
  </div>
  <script type="module" src="./index.js"></script>
</body>

</html>