xautodl/notebooks/TOT/synthetic-env.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "filled-multiple",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The root path: /Users/xuanyidong/Desktop/AutoDL-Projects\n",
      "The library path: /Users/xuanyidong/Desktop/AutoDL-Projects/lib\n"
     ]
    }
   ],
   "source": [
    "import os, sys\n",
    "import torch\n",
    "from pathlib import Path\n",
    "import numpy as np\n",
    "import matplotlib\n",
    "from matplotlib import cm\n",
    "# matplotlib.use(\"agg\")\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib.ticker as ticker\n",
    "\n",
    "\n",
    "__file__ = os.path.dirname(os.path.realpath(\"__file__\"))\n",
    "root_dir = (Path(__file__).parent / \"..\").resolve()\n",
    "lib_dir = (root_dir / \"lib\").resolve()\n",
    "print(\"The root path: {:}\".format(root_dir))\n",
    "print(\"The library path: {:}\".format(lib_dir))\n",
    "assert lib_dir.exists(), \"{:} does not exist\".format(lib_dir)\n",
    "if str(lib_dir) not in sys.path:\n",
    "    sys.path.insert(0, str(lib_dir))\n",
    "\n",
    "from datasets import ConstantGenerator, SinGenerator, SyntheticDEnv\n",
    "from datasets import DynamicQuadraticFunc\n",
    "from datasets.synthetic_example import create_example_v1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "detected-second",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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
      "text/plain": [
       "<Figure size 5760x2880 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "def visualize_env():\n",
    "    \n",
    "    dpi, width, height = 10, 800, 400\n",
    "    figsize = width / float(dpi), height / float(dpi)\n",
    "    LabelSize, LegendFontsize, font_gap = 40, 40, 5\n",
    "\n",
    "    fig = plt.figure(figsize=figsize)\n",
    "\n",
    "    dynamic_env, function = create_example_v1(num_per_task=250)\n",
    "    \n",
    "    timeaxis, xaxis, yaxis = [], [], []\n",
    "    for timestamp, dataset in dynamic_env:\n",
    "        num = dataset.shape[0]\n",
    "        timeaxis.append(torch.zeros(num) + timestamp)\n",
    "        xaxis.append(dataset[:,0])\n",
    "        # compute the ground truth\n",
    "        function.set_timestamp(timestamp)\n",
    "        yaxis.append(function(dataset[:,0]))\n",
    "    timeaxis = torch.cat(timeaxis).numpy()\n",
    "    xaxis = torch.cat(xaxis).numpy()\n",
    "    yaxis = torch.cat(yaxis).numpy()\n",
    "\n",
    "    cur_ax = fig.add_subplot(2, 1, 1)\n",
    "    cur_ax.scatter(timeaxis, xaxis, color=\"k\", linestyle=\"-\", alpha=0.9, label=None)\n",
    "    cur_ax.set_xlabel(\"Time\", fontsize=LabelSize)\n",
    "    cur_ax.set_ylabel(\"X\", rotation=0, fontsize=LabelSize)\n",
    "    for tick in cur_ax.xaxis.get_major_ticks():\n",
    "        tick.label.set_fontsize(LabelSize - font_gap)\n",
    "        tick.label.set_rotation(10)\n",
    "    for tick in cur_ax.yaxis.get_major_ticks():\n",
    "        tick.label.set_fontsize(LabelSize - font_gap)\n",
    "    \n",
    "    cur_ax = fig.add_subplot(2, 1, 2)\n",
    "    cur_ax.scatter(timeaxis, yaxis, color=\"k\", linestyle=\"-\", alpha=0.9, label=None)\n",
    "    cur_ax.set_xlabel(\"Time\", fontsize=LabelSize)\n",
    "    cur_ax.set_ylabel(\"Y\", rotation=0, fontsize=LabelSize)\n",
    "    for tick in cur_ax.xaxis.get_major_ticks():\n",
    "        tick.label.set_fontsize(LabelSize - font_gap)\n",
    "        tick.label.set_rotation(10)\n",
    "    for tick in cur_ax.yaxis.get_major_ticks():\n",
    "        tick.label.set_fontsize(LabelSize - font_gap)\n",
    "    plt.show()\n",
    "\n",
    "visualize_env()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "supreme-basis",
   "metadata": {},
   "outputs": [],
   "source": [
    "# def optimize_fn(xs, ys, test_sets):\n",
    "#     xs = torch.FloatTensor(xs).view(-1, 1)\n",
    "#     ys = torch.FloatTensor(ys).view(-1, 1)\n",
    "    \n",
    "#     model = SuperSequential(\n",
    "#         SuperMLPv1(1, 10, 20, torch.nn.ReLU),\n",
    "#         SuperMLPv1(20, 10, 1, torch.nn.ReLU)\n",
    "#     )\n",
    "#     optimizer = torch.optim.Adam(\n",
    "#         model.parameters(),\n",
    "#         lr=0.01, weight_decay=1e-4, amsgrad=True\n",
    "#     )\n",
    "#     for _iter in range(100):\n",
    "#         preds = model(ys)\n",
    "\n",
    "#         optimizer.zero_grad()\n",
    "#         loss = torch.nn.functional.mse_loss(preds, ys)\n",
    "#         loss.backward()\n",
    "#         optimizer.step()\n",
    "        \n",
    "#     with torch.no_grad():\n",
    "#         answers = []\n",
    "#         for test_set in test_sets:\n",
    "#             test_set = torch.FloatTensor(test_set).view(-1, 1)\n",
    "#             preds = model(test_set).view(-1).numpy()\n",
    "#             answers.append(preds.tolist())\n",
    "#     return answers\n",
    "\n",
    "# def f(x):\n",
    "#     return np.cos( 0.5 * x + x * x)\n",
    "\n",
    "# def get_data(mode):\n",
    "#     dataset = SynAdaptiveEnv(mode=mode)\n",
    "#     times, xs, ys = [], [], []\n",
    "#     for i, (_, t, x) in enumerate(dataset):\n",
    "#         times.append(t)\n",
    "#         xs.append(x)\n",
    "#     dataset.set_transform(f)\n",
    "#     for i, (_, _, y) in enumerate(dataset):\n",
    "#         ys.append(y)\n",
    "#     return times, xs, ys\n",
    "\n",
    "# def visualize_syn(save_path):\n",
    "#     save_dir = (save_path / '..').resolve()\n",
    "#     save_dir.mkdir(parents=True, exist_ok=True)\n",
    "    \n",
    "#     dpi, width, height = 40, 2000, 900\n",
    "#     figsize = width / float(dpi), height / float(dpi)\n",
    "#     LabelSize, LegendFontsize, font_gap = 40, 40, 5\n",
    "    \n",
    "#     fig = plt.figure(figsize=figsize)\n",
    "    \n",
    "#     times, xs, ys = get_data(None)\n",
    "    \n",
    "#     def draw_ax(cur_ax, xaxis, yaxis, xlabel, ylabel,\n",
    "#                 alpha=0.1, color='k', linestyle='-', legend=None, plot_only=False):\n",
    "#         if legend is not None:\n",
    "#             cur_ax.plot(xaxis[:1], yaxis[:1], color=color, label=legend)\n",
    "#         cur_ax.plot(xaxis, yaxis, color=color, linestyle=linestyle, alpha=alpha, label=None)\n",
    "#         if not plot_only:\n",
    "#             cur_ax.set_xlabel(xlabel, fontsize=LabelSize)\n",
    "#             cur_ax.set_ylabel(ylabel, rotation=0, fontsize=LabelSize)\n",
    "#             for tick in cur_ax.xaxis.get_major_ticks():\n",
    "#                 tick.label.set_fontsize(LabelSize - font_gap)\n",
    "#                 tick.label.set_rotation(10)\n",
    "#             for tick in cur_ax.yaxis.get_major_ticks():\n",
    "#                 tick.label.set_fontsize(LabelSize - font_gap)\n",
    "    \n",
    "#     cur_ax = fig.add_subplot(2, 1, 1)\n",
    "#     draw_ax(cur_ax, times, xs, \"time\", \"x\", alpha=1.0, legend=None)\n",
    "\n",
    "#     cur_ax = fig.add_subplot(2, 1, 2)\n",
    "#     draw_ax(cur_ax, times, ys, \"time\", \"y\", alpha=0.1, legend=\"ground truth\")\n",
    "    \n",
    "#     train_times, train_xs, train_ys = get_data(\"train\")\n",
    "#     draw_ax(cur_ax, train_times, train_ys, None, None, alpha=1.0, color='r', legend=None, plot_only=True)\n",
    "    \n",
    "#     valid_times, valid_xs, valid_ys = get_data(\"valid\")\n",
    "#     draw_ax(cur_ax, valid_times, valid_ys, None, None, alpha=1.0, color='g', legend=None, plot_only=True)\n",
    "    \n",
    "#     test_times, test_xs, test_ys = get_data(\"test\")\n",
    "#     draw_ax(cur_ax, test_times, test_ys, None, None, alpha=1.0, color='b', legend=None, plot_only=True)\n",
    "    \n",
    "#     # optimize MLP models\n",
    "# #     [train_preds, valid_preds, test_preds] = optimize_fn(train_xs, train_ys, [train_xs, valid_xs, test_xs])\n",
    "# #     draw_ax(cur_ax, train_times, train_preds, None, None,\n",
    "# #             alpha=1.0, linestyle='--', color='r', legend=\"MLP\", plot_only=True)\n",
    "# #     import pdb; pdb.set_trace()\n",
    "# #     draw_ax(cur_ax, valid_times, valid_preds, None, None,\n",
    "# #             alpha=1.0, linestyle='--', color='g', legend=None, plot_only=True)\n",
    "# #     draw_ax(cur_ax, test_times, test_preds, None, None,\n",
    "# #             alpha=1.0, linestyle='--', color='b', legend=None, plot_only=True)\n",
    "\n",
    "#     plt.legend(loc=1, fontsize=LegendFontsize)\n",
    "\n",
    "#     fig.savefig(save_path, dpi=dpi, bbox_inches=\"tight\", format=\"pdf\")\n",
    "#     plt.close(\"all\")\n",
    "#     # plt.show()"
   ]
  }
 ],
 "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": "3.8.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}