demo/demo-7segment.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "6f349f5f-96a6-4a73-825f-8bc8190cd004",
   "metadata": {},
   "source": [
    "# 7-segment displays\n",
    "\n",
    "This notebook gives a cute example to show off memo's interoperability with the broader JAX ecosystem, which includes deep learning, graphics, physics simulation, and more.\n",
    "\n",
    "A [seven-segment display](https://en.wikipedia.org/wiki/Seven-segment_display) uses seven lights to display a variety of characters. But when designing a \"font\" for such a display, [you have to be careful to avoid ambiguities](https://harold.thimbleby.net/cv/files/seven-segment.pdf): for example, a capital \"B\" and the letter \"8\" naïvely map to the same pattern (all seven segments on), but are impossible to distinguish. Nonetheless, people have come up with all kinds of [creative variants](https://youtu.be/Th-u84OkpeQ?si=2K01bAX2LjjnXiMJ).\n",
    "\n",
    "Can we model intuitions about 7-segment displays as a kind of rational communication? To do so, let us start by training a simple neural network model of hand-written character recognition on the EMNIST dataset, using JAX's `flax` ecosystem for deep learning. We will then build a memo model that reasons _over_ this neural network!\n",
    "\n",
    "(This notebook has a lot of boilerplate/code for training a ResNet. For the most interesting parts, scroll to the end!)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "8e58e74e-e5b4-42a5-b651-e2035f63f017",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'cpu'"
      ]
     },
     "execution_count": 1,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import jax\n",
    "import jax.numpy as jnp\n",
    "import numpy as np\n",
    "import jax.scipy as jsp\n",
    "\n",
    "import torch\n",
    "import torchvision.datasets\n",
    "\n",
    "from matplotlib import pyplot as plt\n",
    "\n",
    "import os\n",
    "import os.path\n",
    "jax.default_backend()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f9f68f62-ad3b-408c-8956-72770beec4bc",
   "metadata": {},
   "source": [
    "Next, we use PyTorch's built-in tools to download and parse the EMNIST dataset."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "06ed0ff3-2413-4d39-b5bb-2b82653c434f",
   "metadata": {},
   "outputs": [],
   "source": [
    "ds_letter = torchvision.datasets.EMNIST(root='./7-aux/data', split='letters', download='True')\n",
    "ds_number = torchvision.datasets.EMNIST(root='./7-aux/data', split='digits', download='True')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "1b290fd8-71f6-4245-8ea9-d5a6f7c09412",
   "metadata": {},
   "outputs": [],
   "source": [
    "feats = []\n",
    "labls = []\n",
    "\n",
    "from collections import Counter\n",
    "cnt = Counter()\n",
    "for feat, labl in ds_number:\n",
    "    if cnt[labl] > 4800:\n",
    "        continue\n",
    "    cnt[labl] += 1\n",
    "    feats.append(np.array(feat).T)\n",
    "    labls.append(labl)\n",
    "for feat, labl in ds_letter:\n",
    "    if 1 <= labl <= 26:\n",
    "        feats.append(np.array(feat).T)\n",
    "        labls.append(labl + 10 - 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "bca460e4-034f-42bd-ba69-117de786524a",
   "metadata": {},
   "outputs": [],
   "source": [
    "feats_ = np.expand_dims(np.array(feats), -1)\n",
    "labls_ = np.expand_dims(np.array(labls), -1)\n",
    "ds = list(zip(feats_, labls_))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "6602602a",
   "metadata": {},
   "outputs": [],
   "source": [
    "import torchvision.transforms.v2 as transforms\n",
    "\n",
    "transforms = transforms.Compose([\n",
    "    transforms.RandomAffine(degrees=10, translate=(0.1, 0.1), scale=(0.9, 1.1)),\n",
    "    transforms.ToDtype(torch.float32, scale=True),\n",
    "])"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f9188578-12cb-45e6-a9e0-ea9221fefe0f",
   "metadata": {},
   "source": [
    "Next, we train a tiny neural network on this dataset. (Or, we load saved weights from a file if a checkpoint exists.)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "476f82b6-c801-43f6-895f-964b1ec106a6",
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "from functools import partial\n",
    "from typing import Any\n",
    "from collections.abc import Callable, Sequence\n",
    "\n",
    "from flax import linen as nn\n",
    "from flax.training import train_state\n",
    "import optax\n",
    "\n",
    "import orbax.checkpoint\n",
    "from flax.training import orbax_utils\n",
    "\n",
    "NUM_CLASS = 36\n",
    "\n",
    "ModuleDef = Any\n",
    "class ResNetBlock(nn.Module):\n",
    "  \"\"\"ResNet block.\"\"\"\n",
    "\n",
    "  filters: int\n",
    "  conv: ModuleDef\n",
    "  norm: ModuleDef\n",
    "  act: Callable\n",
    "  strides: tuple[int, int] = (1, 1)\n",
    "\n",
    "  @nn.compact\n",
    "  def __call__(\n",
    "      self,\n",
    "      x,\n",
    "  ):\n",
    "    residual = x\n",
    "    y = self.conv(self.filters, (3, 3), self.strides)(x)\n",
    "    y = self.norm()(y)\n",
    "    y = self.act(y)\n",
    "    y = self.conv(self.filters, (3, 3))(y)\n",
    "    y = self.norm(scale_init=nn.initializers.zeros_init())(y)\n",
    "\n",
    "    if residual.shape != y.shape:\n",
    "      residual = self.conv(\n",
    "          self.filters, (1, 1), self.strides, name='conv_proj'\n",
    "      )(residual)\n",
    "      residual = self.norm(name='norm_proj')(residual)\n",
    "\n",
    "    return self.act(residual + y)\n",
    "\n",
    "\n",
    "class BottleneckResNetBlock(nn.Module):\n",
    "  \"\"\"Bottleneck ResNet block.\"\"\"\n",
    "\n",
    "  filters: int\n",
    "  conv: ModuleDef\n",
    "  norm: ModuleDef\n",
    "  act: Callable\n",
    "  strides: tuple[int, int] = (1, 1)\n",
    "\n",
    "  @nn.compact\n",
    "  def __call__(self, x):\n",
    "    residual = x\n",
    "    y = self.conv(self.filters, (1, 1))(x)\n",
    "    y = self.norm()(y)\n",
    "    y = self.act(y)\n",
    "    y = self.conv(self.filters, (3, 3), self.strides)(y)\n",
    "    y = self.norm()(y)\n",
    "    y = self.act(y)\n",
    "    y = self.conv(self.filters * 4, (1, 1))(y)\n",
    "    y = self.norm(scale_init=nn.initializers.zeros_init())(y)\n",
    "\n",
    "    if residual.shape != y.shape:\n",
    "      residual = self.conv(\n",
    "          self.filters * 4, (1, 1), self.strides, name='conv_proj'\n",
    "      )(residual)\n",
    "      residual = self.norm(name='norm_proj')(residual)\n",
    "\n",
    "    return self.act(residual + y)\n",
    "\n",
    "\n",
    "class ResNet(nn.Module):\n",
    "  \"\"\"ResNetV1.5.\"\"\"\n",
    "\n",
    "  stage_sizes: Sequence[int]\n",
    "  block_cls: ModuleDef\n",
    "  num_classes: int\n",
    "  num_filters: int = 64\n",
    "  dtype: Any = jnp.float32\n",
    "  act: Callable = nn.relu\n",
    "  conv: ModuleDef = nn.Conv\n",
    "\n",
    "  @nn.compact\n",
    "  def __call__(self, x, train: bool = True):\n",
    "    conv = partial(self.conv, use_bias=False, dtype=self.dtype)\n",
    "    norm = partial(nn.LayerNorm, epsilon=1e-5, dtype=self.dtype)\n",
    "\n",
    "    x = conv(\n",
    "        self.num_filters,\n",
    "        (7, 7),\n",
    "        (2, 2),\n",
    "        padding=[(3, 3), (3, 3)],\n",
    "        name='conv_init',\n",
    "    )(x)\n",
    "    x = norm(name='bn_init')(x)\n",
    "    x = nn.relu(x)\n",
    "    x = nn.max_pool(x, (3, 3), strides=(2, 2), padding='SAME')\n",
    "    for i, block_size in enumerate(self.stage_sizes):\n",
    "      for j in range(block_size):\n",
    "        strides = (2, 2) if i > 0 and j == 0 else (1, 1)\n",
    "        x = self.block_cls(\n",
    "            self.num_filters * 2**i,\n",
    "            strides=strides,\n",
    "            conv=conv,\n",
    "            norm=norm,\n",
    "            act=self.act,\n",
    "        )(x)\n",
    "    x = jnp.mean(x, axis=(1, 2))\n",
    "    x = nn.Dense(self.num_classes, dtype=self.dtype)(x)\n",
    "    x = jnp.asarray(x, self.dtype)\n",
    "    return x\n",
    "\n",
    "\n",
    "ResNet18 = partial(ResNet, stage_sizes=[2, 2, 2, 2], block_cls=ResNetBlock)\n",
    "cnn = ResNet18(num_classes=NUM_CLASS)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "5628155a",
   "metadata": {},
   "outputs": [],
   "source": [
    "@jax.jit\n",
    "def apply_model(state, images, labels):\n",
    "    key = jax.random.fold_in(state.rng, state.step)\n",
    "\n",
    "    def loss_fn(params):\n",
    "        logits = state.apply_fn({'params': params}, images, train=True, rngs={\"dropout\": key})\n",
    "        one_hot = jax.nn.one_hot(labels, NUM_CLASS)[:, 0, :]\n",
    "        loss = jnp.mean(optax.softmax_cross_entropy(logits=logits, labels=one_hot))\n",
    "        return loss, logits\n",
    "\n",
    "    grad_fn = jax.value_and_grad(loss_fn, has_aux=True)\n",
    "    (loss, logits), grads = grad_fn(state.params)\n",
    "    accuracy = jnp.mean(jnp.argmax(logits, -1) == labels[:, 0])\n",
    "    return grads, loss, accuracy\n",
    "\n",
    "\n",
    "@jax.jit\n",
    "def update_model(state, grads):\n",
    "    state = state.apply_gradients(grads=grads)\n",
    "    return state\n",
    "\n",
    "class TrainState(train_state.TrainState):\n",
    "    rng: jax.random.PRNGKey\n",
    "\n",
    "rng = jax.random.key(0)\n",
    "rng, param_key, dropout_key = jax.random.split(rng, num=3)\n",
    "variables = cnn.init(param_key, jnp.empty([1, 28, 28, 1]), train=False)\n",
    "params = variables['params']\n",
    "tx = optax.adam(0.003)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "6bf84f59",
   "metadata": {},
   "outputs": [],
   "source": [
    "import re\n",
    "\n",
    "retrain = False\n",
    "checks = [f for f in os.listdir(\"7-aux\") if f.startswith(\"resnetweights\")]\n",
    "if len(checks) > 0:\n",
    "    restore = sorted(checks, key=lambda x: int(re.match(r'resnetweights(\\d+)', x).group(1)))[-1]\n",
    "\n",
    "if not retrain and len(checks) > 0:\n",
    "    target = jax.tree_util.tree_map(np.zeros_like, variables['params'])\n",
    "    path = os.path.abspath(os.path.join(\"7-aux\", restore))\n",
    "    params = orbax.checkpoint.StandardCheckpointer().restore(path, target)\n",
    "    state = TrainState.create(apply_fn=cnn.apply, params=params, tx=tx, rng=dropout_key)\n",
    "else:\n",
    "    state = TrainState.create(apply_fn=cnn.apply, params=params, tx=tx, rng=dropout_key)\n",
    "    losses = []\n",
    "    accs = []\n",
    "    for epoch in range(60):\n",
    "        print(epoch)\n",
    "        from torch.utils.data import DataLoader\n",
    "        dl = DataLoader(ds, batch_size=128, shuffle=True)\n",
    "        for t, (feat, labl) in enumerate((dl)):\n",
    "            feat = transforms(feat.permute(0, 3, 1, 2))\n",
    "            feat = jnp.array(feat).transpose(0, 2, 3, 1)\n",
    "            labl = jnp.array(labl)\n",
    "            grads, loss, accuracy = apply_model(state, feat, labl)\n",
    "            state = update_model(state, grads)\n",
    "            if t % 400 == 0:\n",
    "                losses.append(loss)\n",
    "                accs.append(accuracy)\n",
    "        print(epoch, losses[-1], accs[-1])\n",
    "\n",
    "        if epoch % 5 == 0:\n",
    "            ckpt = state.params\n",
    "            checkpointer = orbax.checkpoint.StandardCheckpointer()\n",
    "            save_args = orbax_utils.save_args_from_target(ckpt)\n",
    "            checkpointer.save(f'./7-aux/resnetweights{epoch}', ckpt, save_args=save_args, force=True)\n",
    "\n",
    "    plt.plot(accs)\n",
    "    plt.ylabel('Accuracy')\n",
    "    plt.xlabel('Time')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4d8ea964-4faf-42de-b0a8-6759a831aa54",
   "metadata": {},
   "source": [
    "Now, let's feed our trained neural network some 7-segment displays and see how it does. We can generate 7-segment displays in JAX by performing simple pixel manipulations on an array.\n",
    "\n",
    "Do the neural network's guesses match your intuitions?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "0ff20d05-a5f2-4555-937b-5d58e4818da3",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1000x600 with 128 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "@jax.jit\n",
    "def predict(state, img):\n",
    "    out = (state.apply_fn({\"params\": state.params}, jnp.expand_dims(img, (0, -1)), train=False)[0])\n",
    "    return out\n",
    "\n",
    "\n",
    "@jax.jit\n",
    "def make_7seg(i):\n",
    "    L = +9\n",
    "    R = -11\n",
    "    U = 4\n",
    "    M = 13\n",
    "    D = 22\n",
    "    canvas = jnp.zeros((28, 28))\n",
    "    segs = [(i & (1 << n)) != 0 for n in range(7)]\n",
    "    canvas = canvas.at[U, L:R].set(segs[0])\n",
    "    canvas = canvas.at[M, L:R].set(segs[1])\n",
    "    canvas = canvas.at[D, L:R].set(segs[2])\n",
    "    canvas = canvas.at[U : M, L].set(segs[3])\n",
    "    canvas = canvas.at[U : M, R].set(segs[4])\n",
    "    canvas = canvas.at[M : D, L].set(segs[5])\n",
    "    canvas = canvas.at[M : D, R].set(segs[6])\n",
    "    canvas = (\n",
    "        canvas\n",
    "        + jnp.roll(canvas, 1, axis=0)\n",
    "        + jnp.roll(canvas, 1, axis=1)\n",
    "        + jnp.roll(canvas, 2, axis=0)\n",
    "        + jnp.roll(canvas, 2, axis=1)\n",
    "        + jnp.roll(jnp.roll(canvas, 1, axis=0), 1, axis=1)\n",
    "        + jnp.roll(jnp.roll(canvas, 2, axis=0), 2, axis=1)\n",
    "    )\n",
    "    canvas = 255 * (canvas != 0)\n",
    "    canvas = jsp.signal.convolve2d(canvas, jnp.array([[1., 2., 1.], [2., 4., 2.], [1., 2., 1.]]) / 16)\n",
    "    canvas = jsp.signal.convolve2d(canvas, jnp.array([[1., 2., 1.], [2., 4., 2.], [1., 2., 1.]]) / 16)\n",
    "\n",
    "    return canvas\n",
    "\n",
    "\n",
    "plt.figure(figsize=(10, 6))\n",
    "for i in range(128):\n",
    "    plt.subplot(8, 16, i + 1)\n",
    "    img = np.array(make_7seg(i))\n",
    "    plt.imshow(img)\n",
    "    import string\n",
    "\n",
    "    chars = string.digits + string.ascii_uppercase\n",
    "    name = (\n",
    "        chars[predict(state, img).argsort()[-1]]\n",
    "    )\n",
    "    plt.title(name)\n",
    "    plt.axis(\"off\")\n",
    "plt.tight_layout()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "28c50283-394e-4225-a649-cfa0f7be6a4c",
   "metadata": {},
   "source": [
    "Finally, let's use memo to design a \"font\" by reasoning with respect to this computer vision model. We'll consider a subset of \"well formed\" characters to avoid getting too out of distribution from the training data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "5480a83d-9364-4ec9-b550-1e85228dcf11",
   "metadata": {},
   "outputs": [],
   "source": [
    "from memo import memo\n",
    "\n",
    "U = jnp.array([27, 40, 42, 43, 44, 45, 46, 47, 53, 55, 57, 59, 61, 62, 63, 79, 80, 81, 84, 87, 89, 90, 91, 94, 95, 102, 106, 109, 110, 111, 116, 117, 118, 119, 121, 122, 123, 124, 125, 127])\n",
    "R = jnp.arange(36)\n",
    "R_display = jnp.array([0, chars.index(\"O\"), 1, chars.index(\"I\"), 2, chars.index(\"Z\"), 9, chars.index(\"G\"), 6, chars.index(\"B\"), chars.index(\"K\"), chars.index(\"T\")])\n",
    "\n",
    "@jax.jit\n",
    "def pretrained_classifier(u, r):\n",
    "    return predict(state, make_7seg(u))[r]\n",
    "\n",
    "@memo\n",
    "def speaker[u: U, r: R](beta, t):\n",
    "    cast: [speaker, listener]\n",
    "    speaker: knows(r)\n",
    "    speaker: chooses(u in U, wpp=exp(beta * pretrained_classifier(u, r)) if t == 0 else exp(beta * listener[u, r](beta, t)))\n",
    "    return E[speaker.u == u]\n",
    "\n",
    "@memo\n",
    "def listener[u: U, r: R](beta, t):\n",
    "    cast: [listener, speaker]\n",
    "    listener: thinks[\n",
    "        speaker: given(r in R, wpp=1),\n",
    "        speaker: chooses(\n",
    "            u in U,\n",
    "            wpp=speaker[u, r](beta, t - 1),\n",
    "        )\n",
    "    ]\n",
    "    listener: observes [speaker.u] is u\n",
    "    listener: chooses(r in R, wpp=E[speaker.r == r])\n",
    "    return E[listener.r == r]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "44eef2d5-b809-4ed9-b20b-e1462f9df41b",
   "metadata": {},
   "source": [
    "Now, we can see how the emergent convention / \"font\" changes as we add more levels of recursive reasoning."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "be24d887-94f0-470e-a3dc-d067141ff041",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 1.23 s, sys: 90.1 ms, total: 1.32 s\n",
      "Wall time: 823 ms\n",
      "CPU times: user 956 ms, sys: 35.1 ms, total: 991 ms\n",
      "Wall time: 120 ms\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAK8AAAEKCAYAAABkNAWcAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjkuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8hTgPZAAAACXBIWXMAAA9hAAAPYQGoP6dpAAAt+0lEQVR4nO2de3gU5b3Hv7Mze8/uZjfJkg2QEBLCJchVoNBarNzCRY7lCIjQxypWULD1j2Nb2p4Hq8fTQ7F4OFopWEGPeEPEHqmXohYVrHDAJEYuEgKRS+633WQ3m92dmff8wZnpbm6AZW76fp5nH8js7M5nZ78z8877e2eWIYQQUCgGxKS1AIXyVaHhpRgWGl6KYaHhpRgWGl6KYaHhpRgWGl6KYaHhpRgWGl6KYflGhpdhGDz00ENaa2jKkCFDsGDBAq01/iEUCe+zzz4LhmFw9OhRJd5eVZ555hmMHDkSNpsNw4YNwxNPPKG1EuX/+Ubuea+UrVu34u6770ZxcTGeeOIJTJ06FT/+8Y+xYcMGrdUoADitBfRKNBrFL3/5S8yfPx+7d+8GAPzoRz+CKIp45JFHcM8998Dr9WpsqV8ikQicTqeiy9B0z1tTU4O77roLAwYMgNVqRXFxMbZv3y4/39DQAI7j8Otf/7rHa0+dOgWGYfDkk0/K04LBIB544AEMHjwYVqsVhYWF2LBhA0RRvGq3/fv3o6WlBffdd1/K9DVr1iASieDNN9/s9/UdHR144IEHMGTIEFitVvj9fsyaNQulpaXyPDfeeCNGjx6NTz/9FNOmTYPdbkd+fj7+8Ic/9Hi/WCyG9evXo7CwEFarFYMHD8ZPf/pTxGKxlPl27NiBm266CX6/H1arFaNGjcKWLVuu6DM/99xz4DgODz74oDzt8OHDKCkpgcfjgcPhwPTp0/Hxxx+nvO6hhx4CwzA4ceIEbr/9dni9XnznO9+5omX+I2i2521oaMC3vvUtMAyDtWvXIisrC2+//TZWrlyJ9vZ2PPDAAxgwYACmT5+OXbt2Yf369Smvf+WVV8CyLBYvXgwA6OzsxPTp01FTU4NVq1YhNzcXf/vb37Bu3TrU1dXhP//zP6/Kr6ysDABw/fXXp0yfOHEiTCYTysrKsGLFij5fv3r1auzevRtr167FqFGj0NLSgoMHD+LkyZOYMGGCPF9bWxvmzZuHJUuWYNmyZdi1axfuvfdeWCwW3HXXXQAAURSxcOFCHDx4EPfccw9GjhyJzz//HI8//jgqKyvxpz/9SX6/LVu2oLi4GAsXLgTHcdi7dy/uu+8+iKKINWvW9Om7bds2rF69Gr/4xS/wb//2bwCAv/71r5g7dy4mTpyI9evXw2QyyRvHgQMHMHny5JT3WLx4MYYNG4Z///d/hyojbYkC7NixgwAgR44c6XOelStXkkAgQJqbm1Om33bbbcTj8ZDOzk5CCCFbt24lAMjnn3+eMt+oUaPITTfdJP/9yCOPEKfTSSorK1Pm+/nPf05YliXnz5+XpwEg69ev7/czrFmzhrAs2+tzWVlZ5Lbbbuv39R6Ph6xZs6bfeaZPn04AkN/97nfytFgsRsaNG0f8fj+Jx+OEEEKef/55YjKZyIEDB1Je/4c//IEAIB9//LE8TVpvycyZM4cMHTo0ZVpeXh6ZP38+IYSQzZs3E4ZhyCOPPCI/L4oiGTZsGJkzZw4RRTHl/fPz88msWbPkaevXrycAyLJly/r9vNcaTZoNhBC89tpruPnmm0EIQXNzs/yYM2cOQqGQfHhdtGgROI7DK6+8Ir/+2LFjOHHiBJYuXSpPe/XVV3HDDTfA6/WmvN/MmTMhCAI++uijq3KMRqOwWCy9Pmez2RCNRvt9fXp6Og4fPoza2tp+5+M4DqtWrZL/tlgsWLVqFRobG/Hpp5/Kn23kyJEYMWJEyme76aabAFxq4kjY7Xb5/6FQCM3NzZg+fTrOnj2LUCjUY/m//e1v8ZOf/AQbNmzAr371K3l6eXk5Tp8+jdtvvx0tLS3yMiORCGbMmIGPPvqoR3Ns9erV/X7Wa44SW8Tl9rwNDQ0EQL+PPXv2yPPPmTOHFBUVyX//6le/IhzHkaamJnma3W7v9/02bdokzwsV9ryvvPIKsdlsxGQykUmTJpH169eTM2fOpMwzffp0kpub2+O177//PgFAXnrpJUIIISNHjuz3s/34xz+WX3vw4EEyY8YM4nA4esx37tw5eb68vDySlpZGAJCf/exnvfpf7jtqbW0lhPx9z5t8dFMDTdq80ha7YsUK3HHHHb3OM2bMGPn/t912G+68806Ul5dj3Lhx2LVrF2bMmIHMzMyU95w1axZ++tOf9vp+RUVFV+UYCAQgCAIaGxvh9/vl6fF4HC0tLcjJyen39UuWLMENN9yA119/Hfv27cPGjRuxYcMG7NmzB3Pnzr0qF1EUcd1112HTpk29Pj948GAAwJkzZzBjxgyMGDECmzZtwuDBg2GxWPDWW2/h8ccf77GnLC4uRjAYxPPPP49Vq1YhPz8/ZZkAsHHjRowbN67X5aalpaX8nbzXVwNNwpuVlQWXywVBEDBz5szLzn/LLbdg1apVctOhsrIS69atS5mnoKAA4XD4it7vSpC+sKNHj2LevHny9KNHj0IUxT6/0GQCgQDuu+8+3HfffWhsbMSECRPw6KOPpoS3tra2R7dSZWUlgEtVMODSZ/vss88wY8YMMAzT5/L27t2LWCyGN954A7m5ufL05GZFMpmZmdi9eze+853vYMaMGTh48KC8URYUFAAA3G73NVun1xpN2rwsy+Kf//mf8dprr+HYsWM9nm9qakr5Oz09HXPmzMGuXbvw8ssvw2Kx4JZbbkmZZ8mSJfjkk0/wl7/8pcf7BYNB8Dx/VY433XQTfD5fj26mLVu2wOFwYP78+X2+VhCEHu1Lv9+PnJycHl1bPM9j69at8t/xeBxbt25FVlYWJk6cKH+2mpoaPP300z2WFY1GEYlEAFxarwBSzvRDoRB27NjRp+ugQYPw3nvvIRqNYtasWWhpaQFwqVeloKAAjz32GMLhcI/Xdf+OtEDRPe/27dvxzjvv9Jj+k5/8BP/xH/+B/fv3Y8qUKfjRj36EUaNGobW1FaWlpXjvvffQ2tqa8pqlS5dixYoVeOqppzBnzhykp6enPP/ggw/ijTfewIIFC/DDH/4QEydORCQSweeff47du3fjyy+/TGlmXA673Y5HHnkEa9asweLFizFnzhwcOHAAO3fuxKOPPgqfz9fnazs6OjBo0CDceuutGDt2LNLS0vDee+/hyJEj+N3vfpcyb05ODjZs2IAvv/wSRUVFeOWVV1BeXo5t27bBbDYDAH7wgx9g165dWL16Nfbv349vf/vbEAQBX3zxBXbt2oW//OUvuP766zF79mxYLBbcfPPNWLVqFcLhMJ5++mn4/X7U1dX16VtYWIh9+/bhxhtvxJw5c/DXv/4Vbrcbf/zjHzF37lwUFxfjzjvvxMCBA1FTU4P9+/fD7XZj7969V7w+FUGJhrR0wtbX48KFC4SQSydua9asIYMHDyZms5lkZ2eTGTNmkG3btvV4z/b2dvmkbOfOnb0ut6Ojg6xbt44UFhYSi8VCMjMzybRp08hjjz0mdzsRcmUnbBLbtm0jw4cPJxaLhRQUFJDHH388peuoN2KxGHnwwQfJ2LFjicvlIk6nk4wdO5Y89dRTKfNNnz6dFBcXk6NHj5KpU6cSm81G8vLyyJNPPtnjPePxONmwYQMpLi4mVquVeL1eMnHiRPLrX/+ahEIheb433niDjBkzhthsNjJkyBCyYcMGsn37dgKAVFdXy/Mld5VJHD58mLhcLvLd735X7nIrKysjixYtIhkZGcRqtZK8vDyyZMkS8v7778uvk07Ykk+g1YAhhN63QStuvPFGNDc399p0olweOjCHYlhoeCmGhYaXYlhom5diWOiel2JYaHgphoWGl2JYaHgphoWGl2JYaHgphoWGl2JYaHgphoWGl2JYaHgphkW18MZiMfzsZz9DTk4O7HY7pkyZgnfffVetxffL8ePHsWLFCgwcOBBWqxU5OTlYvnw5jh8/rqlXOBzG+vXrUVJSAp/PB4Zh8Oyzz2rq1Bta3ZtOtfD+8Ic/xKZNm7B8+XJs3rwZLMti3rx5OHjwoFoKvbJnzx5MmDAB77//Pu6880489dRTWLlyJfbv348JEybg9ddf18ytubkZDz/8ME6ePImxY8dq5qFb1BjxfvjwYQKAbNy4UZ4WjUZJQUEBmTp1qhoKvVJVVUUcDgcZMWIEaWxsTHmuqamJjBgxgjidzh6XrKtFV1cXqaurI4QQcuTIEQKA7NixQxOX/riSm8wogSp73t27d4NlWdxzzz3yNJvNhpUrV+KTTz7BhQsX1NDowcaNG9HZ2Ylt27YhKysr5bnMzExs3boVkUgEv/3tbzXxs1qtyM7O1mTZRkCV8JaVlaGoqAhutztlunSvq/LycjU0erB3714MGTIEN9xwQ6/Pf/e738WQIUMue1M9ijaoEt66ujoEAoEe06Vpl7slkhKEQiHU1tZeti05ZswYXLx4ER0dHSqZUa4UVcIbjUZhtVp7TLfZbPLzaiOF0eVy9Tuf9Hx7e7viTpSrQ5Xw2u32HjfbAICuri75ebWRQnm5PeqVhpyiPqqENxAI9HrTC2na5e77pQQejweBQAAVFRX9zldRUYGBAwf2aK9TtEeV8I4bNw6VlZU9Dr2HDx+Wn9eCBQsWoLq6us++5gMHDuDLL780/K/mfF1RJby33norBEHAtm3b5GmxWAw7duzAlClT5Lscqs2DDz4Iu92OVatWyffokmhtbcXq1avhcDhSbnNP0Q+q3CVyypQpWLx4MdatW4fGxkYUFhbiueeew5dffolnnnlGDYVeGTZsGJ577jksX74c1113HVauXIn8/HzZq7m5GS+99JJ8x0QtePLJJxEMBuUemb179+LixYsAgPvvvx8ej0czN81RqxoSjUbJv/zLv5Ds7GxitVrJpEmTyDvvvKPW4vuloqKCLFu2jAQCAfmeacuWLevxUwJakJeX1+c935LvPaYlWlXY6H0bKIaFDomkGBYaXophoeGlGBYaXophoeGlGBYaXophoeGlGBYaXophoeGlGBYaXophueYDc651tbm/nyu9WvTq1t0r+e+vsgwl15me3BQdVUYIkR9XgvTBTCblDwh6cyOEgOd5JBIJ8DwPURRhMplgsVjAsmyfy1VjnenVTbHw8jwPnuchCAIEQbhsSBiGgclkAsuyMJvNYFn2mm6lenfjeR7Nzc2or69HS0sLOjs7kZaWhpycHKSnp8NqtfYIgVrrTK9uioRX2lK7uroQi8WQSCQgimK/r5E+qMVigcPhgMlkUuSL0KtbPB5HXV0dysvLUVlZiebmZmRnZ2PSpEnIy8uD2+0Gx6V+XWqtM726KRZeQRAQj8flkIii2Kc8IQQMw8g/FN3blcZfdzee59Ha2oqqqiqUlZXh4sWLGDJkCAKBADwejxwEtb307KZYs0EURfA8j3g8jlgsBkKIvPVJQZHanFJ4CCHgOO6KDuVfNzdRFBEOh1FfX4/q6mqcO3cODMOgra0NXV1dSCQSYFlWk3WmVzfF9rzJHyR5S0xu4EuH8Hg8Ls97uUP419WNEIJEIpFyNJBuDWA2m2Gz2WCz2TRbZ3p0U/waNmlvZjabkZaWBqvVKjfepZUQDoeRSCQUO0EzqpvFYoHL5UJ6ejq8Xi/sdrvmXnpyU+UCTJPJBI7jYLVa4XA4wHEcGIaRDyddXV3geV4NFUO5sSwrezmdTtjtdl146cVNlfACl7ZWlmXBsqwcEGma9H8t9iJ6dUt24DhOfmjtpSc31crDyR+ot4eW6NVNWrbevPTipkqFTWq4Jz+kQ7MgCCnTrqbq9XV1S3bs/u/VVgaVQg9uihcpIpEIotEoYrEYGIZBLBaT+/8EQUAkEkEwGEQsFgPLsgAu3dROqRWgZ7fungDkjSd5Q5J6SZLnUxO9uCkSXkEQ0NbWhurqaly4cAEtLS1gGAYul6vHGX00GkU4HIYoinA6nQgEAmBZFmlpaYqUFPXs1p3kjSwcDstHA+mESBpjoAV6cFMkvIlEAk1NTThx4gQqKipw9uxZxGIx2Gw2cByX0peaSCQQi8XAcRz8fj9GjRoFn8+HQCCQUrX5JrglIy2/s7MTHR0dCIVCckElkUjIZW0t2r56cVMkvDzPIxQK4fz58zh16hROnz4tV2G6b43SYYdlWbS2tsJms2H8+PHgeV6Rw46e3bqTXMaORqPyMgVBkLuipOaM2ujBTZHwiqKIzs5OtLS0oL6+Hk1NTQAgd61IpVhRFJFIJCAIAoBLZ64NDQ2IRCKKVWb07NYd6YQyuU0pTdMaPbgpOiQyHo8jkUjI09xuN7KysmC328EwDOLxOOrr6xEMBuUV0dXVpfjYBj27JSMNeJHKr72VYLVCD26qFCmkE6KSkhKMHj0aGRkZYFkW4XAYpaWl2LdvX4/746qFXt2kEmxaWhrS09Ph8/lgs9l0VR7W2k2V8LIsC7/fjzFjxmDatGkIBAIwm81oa2uDIAioqKhAJBJRQ8VwbsklWJvNptvysBZuqpWHrVYr3G43MjIykJWVJY/i8ng8vY7EVxM9uiWXYLuXrvVUHtbSTbVmgzQAxmKxwGKxwGw2w2w2g+M4VfpMjeom/UvLwz1RJbzJZUOpM9tkMqWcpWp18qF3t+7/0vLw31Gt2RCPxxEOhxEKheB0OsFxHILBIMLhMOLxOERR1KzPUo9uvZVgu3dLJc/3TXRTJbyCIKChoQFlZWXgeV4+o+/o6EBZWRkaGxsRj8flQ7aah0Q9uUlNmOQ2diwWk0uwUtk1uctOqYsujeCmWrOho6MDH374IU6dOgWn05nSl9rU1ARCCJxOp3wptVoB1pObxWKRTxyj0ShcLhcEQZCPClIJVrp0P/mkSen1pUc3xcIrXa8vjdISRRGNjY1ob29PuVpB2lK9Xi9ycnKQk5MDt9ut6GFaj27SgJ+BAwdi2LBhcLvdyM7Oli+vkcIh+XYfDK7kXk6vboqE12QywW63yx3YLS0tcplVFEXE43F5XrPZDI/Hg9zcXBQWFiI/Px8+n0+xQ7Re3TiOg9frxdChQ8GyLNra2uByuZCRkSEPApJKr1I4zGaz3JxRcmPXq5si4ZU+bEFBASKRCJxOZ68d/QzDgOM4+Hw+5OXloaCgACNHjkR2djYsFosyW6tO3cxmM/x+PxiGQSAQQGdnJziOQ3p6OlwuV0oIugfEYrHIRwwl0KvbNf8dNmlYXFNTE2pra9HU1IS2traUcQQpAgwDm80Gj8eDzMxM+P1+ZGZmwmazXfM+Q726Sd1LsVhMHnMhHQ2kO890X15ym7J7f/S1Xmd6dVMkvIQQxONx+ZFIJPrtNuleJEjesynxRejN7av0lfZWJOj+3NfdTZHwSv9eTSe/9CGTG/fX+lCjV7dr+RUotc6uBdfajf58K8Ww0DujUwwLDS/FsNDwUgwLDS/FsNDwUgwLDS/FsNDwUgwLDS/FsNDwUgwLDS/FsNDwUgwLDS/FsNDwUgwLDS/FsNDwUgwLDS/FsNDwUgwLDS/FsNDwUgwLDS/FsNDwUgwLDS/FsKgS3iNHjmDt2rUoLi6G0+lEbm4ulixZgsrKSjUW3y993dk7+fHQQw9p5vfpp5+ipKQEbrcbLpcLs2fPRnl5uWY+yVRXV2Pt2rUoKiqCw+GAw+HAqFGjsGbNGlRUVCi+fFXu23Drrbfi448/xuLFizFmzBjU19fjySefRDgcxqFDhzB69GilFfpk586dfT730EMP4cyZM3jrrbcwd+5cFa0uUVpaim9/+9sYPHgwVq1aBVEU8dRTT6G1tRX/+7//i+HDh6vuJPHnP/8ZS5cuBcdxWL58OcaOHQuTyYQvvvgCe/bswblz51BdXY28vDzlJIgKfPzxxyQWi6VMq6ysJFarlSxfvlwNhavm6aefJgDI/fffr5nDvHnziNfrJc3NzfK02tpakpaWRhYtWqSZV1VVFXE6nWTkyJGktra2x/OJRIJs3ryZnD9/XlEPVcLbFxMmTCATJkzQUqFXjh07Rux2Oxk/fjzp6urSzMPlcpHFixf3mD5//nxisVhIR0eHBlaE3HPPPQQAOXTokCbLl9DshI0QgoaGBmRmZmql0CudnZ1YsmQJWJbFyy+/DKvVqplLLBaD3W7vMd3hcCAej+PYsWMaWF1qMhQWFmLKlCmaLF9Cs/C+8MILqKmpwdKlS7VS6JX7778fJ06cwO9//3sUFRVp6jJ8+HAcOnRIvqUocOnHXw4fPgwAqKmpUd2pvb0dtbW1vZ6nBINBNDc3y49oNKqsjBa7+5MnTxK3202mTp1KeJ7XQqFXXnjhBQKA/OAHP9BahRBCyJYtWwgAcscdd5Djx4+Tzz//nCxdupSYzWYCgDz//POqO124cIEAICtWrOjx3NixYwkA+bFx40ZFXVQPb11dHRk6dCgZPHgwqampUXvxfVJZWUlcLhcpKirSrC3ZG7/4xS/ksAIg119/PfnlL39JAJDXX39ddZ9gMEgAkFtuuaXHc4cOHSLvvvsu2blz59cvvMFgkIwbN474fD5y/PhxNRfdL11dXWT8+PHEarWS0tJSrXV60NraSg4cOEAqKioIIYSsW7eOANBsHQYCAVJYWNjn89XV1V+v8EajUXLDDTcQh8NB/va3v6m12Cti7dq1BADZvHmz1ipXxKRJk8igQYOIIAiaLP/uu+8mAMjhw4d7ff5rFV6e58nChQsJx3HkzTffVGORV8yePXsIALJw4UKtVa6Il19+mQAgjz32mGYOlZWVxOFwkOLiYlJfX9/j+bNnz6oSXlUqbA888AA2b96Mm2++GUuWLOnx/IoVK5RW6JW6ujoUFxejvb0dmzZtgs/n63W+goICTJ06VWU74KOPPsLDDz+M2bNnIyMjA4cOHcKOHTswa9Ys7N27Fxyn2q/v9uB//ud/sGzZMlgsFrnCRghBdXU1XnzxRdTU1OCFF17AbbfdppyEopvG/zN9+vSUs9DuD63Yv39/v17S44477tDEr6qqisyePZtkZmYSq9VKRowYQX7zm9/0qFZqRVVVFbn33ntJYWEhsdlsxG63kxEjRpDVq1eT8vJyxZdPf5OCYljokEiKYaHhpRgWGl6KYaHhpRgWGl6KYaHhpRgWGl6KYaHhpRgWGl6KYaHhpRiWaz6yo7dq89VWoBmG6fX//yh6devu8FUq9mqtMz25KTYsiRACnufB8zwSiQR4nu/3gzMMA5PJBI7jYLVawXHcNf2geneTnCSfyzmp5aVnN8XCG4vFEAwG0dTUhIaGBrS1tSGRSPQtwnFwOBzIyMjA4MGD4ff7YTabFfky9OjG8zyam5tRX1+PpqYmhEKhfp3U8tKzmyLhJYSgq6sLdXV1+OKLL1BRUYEzZ84gGo32ucXabDZkZWVh2LBhMJlMSE9PV2S8ql7d4vE46urqUF5ejpMnT+LcuXPo7Ozs9zVqrTO9uikW3ng8jmAwiAsXLuDYsWOoqKhAIpHoMyBmsxnZ2dngeR5FRUX9zvt1dON5Hq2traiqqkJ5eTlOnz6NWCzW72vUWmd6dVOs2SCKImKxGDo6OtDa2orm5uZeG/+JRAKiKAK4dDj3+Xxob29PuVfBN8FNFEWEw2HU19fj3LlzOH/+PADI7UaTyaSJl57dFNvzMgwDq9UKr9eLQYMGIRqNIpFIgGVZmEwmeQ9YX1+PUCgEURTB8zw6Ozuv6ITg6+YmffFdXV0pezW3243MzEzY7XaYTCbN1pke3RTb85rNZmRmZmLEiBFIS0vD5MmTYbPZYLVawbIsRFFER0cHysrKsG/fPrS0tCilYig3CYZh4HK5MHfuXIwePRo+nw8cx2nupSc3xcLLsizS0tIQCATgcrlgMpng8XjgcDjAcRwEQUBbWxsEQcBnn32GcDislIqh3JIds7KycN1112HatGkIBALgOE5zLz25KRZehmHAcZx8SJEO0w6HA2azGYIggOM4eDwe2Gw2sCyrlIqh3JKxWq3weDzIyMhAVlYWzGazLrz04qbotdMmk0lu1FssFlitVlitVjkgFosFZrMZLMsqVpAwohtwaQNjWVbu5JfctPbSk5uiYxvIpZuaQBTFPh/SPGpfxKxnt2S/3hy19NKTm2K9DTzPo6urC5FIBNFoFLFYDAzDIBaLyXu3YDCIcDiMeDyu2hehZ7fuxONxhMNhBINBuUmjBy+9uCkSXumEp7q6GhcuXEBLS4t8htrbGX1jY6McHCXHNOjdrbtnY2MjysvLIQgCMjIywLIsOjo6UFpaitraWnR1dSEtLU0VHz26KRLeRCKBpqYmnDhxAhUVFTh79ixisRhsNpvcqS0VCurq6tDU1ASe52GxWOByuRQb06B3N6kdybIsCCHo6OjAgQMHUFVVBYfDAZPJhFgshgsXLqC1tRU8zyviYRQ3RcLL8zxCoRDOnz+PU6dO4fTp0ylFAAlBEBCNRhGPx+F0OpGZmYmsrCw4nc6U+b4JbizLwm63w+VyweFwyA51dXUIhULyXl8QBEQiEcTj8WvuYDQ3RcIriiI6OzvR0tIij0QC/r71mkwmee/FMAycTicCgQByc3MxcOBAeDwexQ7RenVjWRYulwsDBgxAdnY2ampq5BDE43H5/1LlSk306qZYVxnP84jH4ylD59xuN7KysmC32+XDDyEEDocDgUAAQ4cORWFhodxvqNThWY9uUtWvsLAQHR0dMJlMqK+vl7ulGIZJKVsHg0F5DIHS6NVNlXtkSidEJSUlGD16NDIyMmCxWOS2ZTweh81mQ0ZGBgYNGgSfz6doePXoxrIsfD4fRowYAafTiaFDhyISicBut6ecSIbDYZSWlqpaHtarmyrhZVkWfr8fY8aMkcuJNpsNoigiEomgvb1dPinyeDywWCyqndXrxS15sBDDMMjIyJDHwdrtdrkLTyrBVlRUIBKJXHMPI7mpdndiq9UKt9stlxNtNhsEQZBLibFYDCaTSbU9rh7dGIaB2WxOKVv7fL6UsjXLsvB4PLBarYqd1BrFTbVmQ3IpVirHCoKARCIBi8UCQRBSfqxaLfTm1puPVHYVBEHub9aiPKw3N1XCm1wyFASh14coiprU6o3gJv1fa5LLwMnrJvn/ySVipVGt2SCVE0OhEJxOJ7q6uuR+wfb2diQSCXmLVvuL0qubNHZAEASYTKaUkKg5tkEqqcdiMUQiEXlQujQAXRpGmlweFkVR8ZFlqoRXEAQ0NDSgrKwMPM+nnNF3dXWB53k4HA5kZmaCYRh4vV41tHTtJh0Jurq6AFw6sZQ2qFgsJgdXaouqUVI/e/YsLl68iObm5suW1OPxuNykUMpNtWZDR0cHPvzwQ5w6dQpOpzOlL9XlciE/Px/Dhw+H1+tVfe+mNzdpj5tIJNDZ2Yl4PA6TyQSe59HR0QGe5+V+aYvFgoyMDFitVkVL6o2NjTh+/Dg+++wzuaQu3Y9BKqlL/bxNTU0ghMDpdCI9PV0xN0WvpDCbzTCbzQAutZcaGxvR3t4uf2DperKsrCw4HA4MGjRIlfalXt2kE0KpyicFgud5ufwq9Tvn5ubKly4VFBQgPT1dscO0IAgIh8NoaGhAdXU1Tp48CZ7n5THRyfNJTS6v14ucnBzk5OTA7XYr4qZIeE0mE+x2O9LT0+Hz+dDS0iJfPSp9IRLSlyJd3aD0YGa9uiUHVypVSxtL8kllcsEgNzcXVqsV2dnZilclk9vfyY/umM1meDwe5ObmorCwEPn5+YoVdhQJL8dx8Hq9KCgoQCQSgdPp7LXTWgqF3+/HkCFD4Pf7kZaWpmw7ScduLMvK7W2pHZkMIQRmsxm5ubnw+XwghMij3fx+v2IFFJZl4Xa7MWjQIIwYMQJms7nfdebz+ZCXl4eCggKMHDkS2dnZirgpEl6r1Yrc3FxwHIdhw4b1ezslhmFgs9ng8/ng9/sV/RL07JZcBJH6lrsj7X0zMzMhiqK8l5YuwVFqMNNXWWcejweZmZnw+/3IzMxUZKO/5j8iKK1gabSRNACmv8V0v5Ys+YNe6zse6tEtuS/3ai6jSS6cKHmXyH9knUkPJb5PRcIr/Xs1/ZHdT1akad8Et2v5FXxT1hmgQHgpFLWgd0anGBYaXophoeGlGBYaXophoeGlGBYaXophoeGlGBYaXophoeGlGBYaXophoeGlGBYaXophoeGlGBYaXophoeGlGBYaXophoeGlGBYaXophoeGlGBYaXophoeGlGBYaXophUTW8paWlWLhwoXw7+NGjR+O//uu/1FTowbPPPtvj5h1+vx/f+9738Pbbb+vC7ejRoynTQ6EQJk+eDJvNhnfeeUdVp+7rqq/HBx98oLiLajeX3rdvH26++WaMHz8e//qv/4q0tDScOXMGFy9eVEuhXx5++GHk5+eDEIKGhgY8++yzmDdvHvbu3YsFCxZorSfT3t6O2bNno6KiAq+//jpKSkpUXf7zzz+f8vd///d/49133+0xfeTIkcrLEBUIhUJkwIAB5Pvf/z4RBEGNRV4xO3bsIADIkSNHUqa3trYSs9lMbr/9do3Merq1t7eTb33rW8RisZA///nPmnkls2bNGqJSjHqgSrPhxRdfRENDAx599FGYTCZEIhHVfgDvqyL9TBPHqXZw6pdwOIySkhKUlpbitddew/z587VW0hxVwvvee+/B7XajpqYGw4cPR1paGtxuN+699175tvVaEwqF0NzcjKamJhw/fhz33nsvwuEwVqxYobUaIpEI5s6diyNHjuDVV1/VVTNGS1TZrZw+fRo8z+Of/umfsHLlSvzmN7/BBx98gCeeeALBYBAvvfSSGhr9MnPmzJS/rVYrtm/fjlmzZmlk9HfuuOMO1NbW4tVXX8XChQu11tENqoQ3HA6js7MTq1evlnsXFi1ahHg8jq1bt+Lhhx/GsGHD1FDpk9///vcoKioCADQ0NGDnzp24++674XK5sGjRIk3dGhoaYLPZMHjwYE099IYqzQa73Q4AWLZsWcr022+/HQDwySefqKHRL5MnT8bMmTMxc+ZMLF++HG+++SZGjRqFtWvXptzqXwu2bt0Ki8WCkpISnDp1SlMXPaFKeHNycgAAAwYMSJnu9/sBAG1tbWpoXBUmkwnf+973UFdXh9OnT2vqMmrUKLz11luIRqOYNWsWLly4oKmPXlAlvBMnTgQA1NTUpEyvra0FAGRlZamhcdXwPA/gUrNHayZPnow//elPaGxsxKxZs9DU1KS1kuaoEt4lS5YAAJ555pmU6X/84x/BcRxuvPFGNTSuikQigX379sFisajT4X4FzJgxAy+99BKqqqpQUlKC9vZ2rZU0RZUTtvHjx+Ouu+7C9u3bwfM8pk+fjg8++ACvvvoq1q1bJzcrtOTtt9/GF198AQBobGzEiy++iNOnT+PnP/853G63xnZ/5/vf/z6efvpp3HXXXVi4cCHeeecd2Gw2rbW0Qa1qSDweJw899BDJy8sjZrOZFBYWkscff1ytxfeJVMVKfthsNjJu3DiyZcsWIoqi5m7dq3+EEPLYY48RAGTBggUkkUhoYHcJLSts9DcpKIaFDomkGBYaXophoeGlGBYaXophoeGlGBYaXophoeGlGBYaXophoeGlGBYaXophueYDc7pXm5P/Zhjmqt/vq7ymL/Tqplev7i7d/9baTZFRZYQQ8DyPRCIBnuchiiJMJhMsFgtYloXJ1PsOX/pgfT3/dXbTq5ee3RQJL8/zaG5uRn19PVpaWtDZ2Ym0tDTk5OQgPT0dVqu1xwdiGAYmkwksy8JsNoNl2Wu6lerdTa9eenZTJLzxeBx1dXUoLy9HZWUlmpubkZ2djUmTJiEvLw9ut7vH/RCkD2qxWOBwOGAymRT5IvTqplcvPbsptudtbW1FVVUVysrKcPHiRQwZMgSBQAAej0f+UBKEEDAMA7PZDODSZedKoVc3vXrp2U2R8IqiiHA4jPr6elRXV+PcuXNgGAZtbW3o6upCIpEAy7IghIAQAlEUwTAMCCHgOA6CIPQ4Ufi6u+nVS89uip2wJRIJdHV1IRaLQRRF+c44ZrMZNpsNNptNPhGIx+Pyh1b6NlB6ddOrl57dVLmGjWEYWCwWuFwupKenw+v1wm63yyshHA4jkUgo0l4zqptevfTkptpd5FiWhdVqhcPhgNPphN1ulw8nXV1d8mXmWqBXN7166cVNtT0vwzBgWRYcx8kPaVryTYnVRq9uevXSk5tq5WHpg/R1J20t0aubXr304qbKnlc6C01uxIuiCEEQIAiC/Ld0hqrmBc16dbsSL60u/NaLm2pt3ng8jnA4jGAwCIfDgWg0CkEQEIlEEAwGEYvFwLIsAMDlcqn6xejVrT+vUCiERCIBs9kMs9msepD14KZKeAVBQGNjI8rLyyEIAjIyMmCxWCCKIqLRKMLhMERRhNPpRCAQAMuySEtLU6zcaQS3/rykvlWHw4GMjAwMGjQIXq9XMRe9uqnWbOjo6MCBAwdQVVUFh8MBjuPk/sNYLAaO4+D3+zFq1Cj4fD4EAoGUqs03za0/L2ljys3NRVFRETweDwRBUNRHj26qNRsEQUBdXR1CoZB8ZgpAbiuxLIvW1lbYbDaMHz8ePM+rdijUq1tfXgDg9XphMpng9/vlooCa6MFN1V8LEQQB0WhUHqQhiiISiYS8ZTIMg4aGBk1+cEWvbt29JKxWK2KxmDxEUQu0dlM1vG63G1lZWbDb7WAYBvF4HPX19QgGg/JerqurS9E6vdHckr2Sxw94PB5kZ2fD7XbDYrFo0nWmtZtqRQqXy4WSkhKMHj0aGRkZYFkW4XAYpaWl2LdvH1paWtRQMYxbb17SSVEsFpNPirKysuSRXd80N1XCy7Is/H4/xowZg2nTpiEQCMBsNqOtrQ2CIKCiogKRSEQNFcO49eZls9kgiiIikQja29vB8zwsFgvS0tIUvZJCr26qNRusVivcbjcyMjKQlZUlj673eDy9jsRXE726dfey2WwQBAE2mw0syyIWi8FkMsFsNqvebNCDm2rNBpPJBI7jYLFYYLFY5A5sjuNU6c81mltvXlarFYIgIJFIwGKxQBAETcrFenFTvTwslRBNJpNcVlS7JGwEt968uj+ksvU31U318nAoFILT6QTHcQgGgwiHw4jH4xBFUdWTDiO4dfeSejukdmUikZD3flqVh7V0U6083NDQgLKyMvA8L5/Rd3R0oKysDI2NjYjH4/IhW829iV7devNKLsHyPA+Hw4HMzEwwDKN6eVgPbqqWhz/88EOcOnUKTqczpS+1qakJhBA4nU75Umq1QqJXt968kvtSXS4X8vPzMXz4cHi9XtVH4unBTbHwStfrS1eQiqKIxsZGtLe3y+XE5I5/r9eLnJwc5OTkwO12K3qY1qvb5bxMJpN8ZW5WVhYcDgcGDRqkSvtSj26KhNdkMsFutyM9PR0+nw8tLS1ymVUUxZTf8jWbzfB4PMjNzUVhYSHy8/Ph8/kUO0Tr1e1qvKSNi2EYcByn6A1H9OymSHg5joPX60VBQQEikQicTmevHf3SB/T5fMjLy0NBQQFGjhyJ7OxsxcqKenW7Wi+/348hQ4bA7/cjLS1N2f5UnbopEl6r1Yrc3FxwHIdhw4ahra0NiUSi13kZhoHNZoPH40FmZib8fj8yMzMV+8B6dfsqXj6fD36/H36/X9ExBHp1u+Y/Iig12uPxuPxIJBL9Ntq7d3gnf9hr+aH16vaPelmt1pQNSk/rTEk3RcIr/Xs1nfxSJSZ5eN213lr16qZXL7270Z9vpRgWemd0imGh4aUYFhpeimGh4aUYFhpeimGh4aUYFhpeimGh4aUYFhpeimGh4aUYFhpeimGh4aUYlv8DKkhds1wxFVYAAAAASUVORK5CYII=",
      "text/plain": [
       "<Figure size 200x300 with 12 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 200x300 with 12 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "grid_full = make_7seg(127)\n",
    "grid_full = grid_full / (255 * 5)\n",
    "\n",
    "for t in range(2):\n",
    "    plt.figure(figsize=(2, 3))\n",
    "    plt.gcf().subplots_adjust(top=0.9, hspace=0, wspace=0)\n",
    "    %time S = speaker(2.0, t)\n",
    "    plt.suptitle(f'Level {t} speaker', y=1.0)\n",
    "    for i in range(len(R_display)):\n",
    "        plt.subplot(3, 4, i+1)\n",
    "        grid = make_7seg(U[jnp.argmax(S[:, R_display[i]])])\n",
    "        grid = grid / 255\n",
    "        grid = grid + grid_full\n",
    "        grid = np.stack([grid, grid, grid], axis=-1)\n",
    "        grid = 1 - grid\n",
    "        plt.imshow(np.clip(grid, 0., 1.))\n",
    "        plt.title(chars[R_display[i]])\n",
    "        plt.axis(\"off\")\n",
    "    plt.savefig(f\"7seg-{t}.pdf\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "4e1ac7b1",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": ".cpu_env",
   "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": "3.12.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}