added decoder scaling plot

2026-10-04 13:00:43 +08:00 · 2022-05-21 11:13:00 +02:00
1 parent af6b4ea765
commit cfc8daa4fa
@@ -2,8 +2,8 @@
 "cells": [
  {
   "cell_type": "code",
-   "execution_count": 55,
-   "id": "5a62c0ff",
+   "execution_count": 1,
+   "id": "d4aa17d8",
   "metadata": {},
   "outputs": [],
   "source": [
@@ -17,10 +17,18 @@
  },
  {
   "cell_type": "code",
-   "execution_count": 50,
-   "id": "c75f5208",
+   "execution_count": 2,
+   "id": "a45a0a88",
   "metadata": {},
-   "outputs": [],
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "WARNING:root:Pytorch pre-release version 1.12.0a0+git689df63 - assuming intent to test it\n"
+     ]
+    }
+   ],
   "source": [
    "from pytorch_lightning.utilities.model_summary import summarize\n",
    "from datasets import load_dataset\n",
@@ -32,7 +40,7 @@
  {
   "cell_type": "code",
   "execution_count": 8,
-   "id": "aa342b02",
+   "id": "6d78a5c6",
   "metadata": {},
   "outputs": [
    {
@@ -169,18 +177,18 @@
  {
   "cell_type": "code",
   "execution_count": 16,
-   "id": "5aefbb4a",
+   "id": "4a3efb88",
   "metadata": {},
   "outputs": [],
   "source": [
    "freq = \"1H\"\n",
-    "prediction_length = 24"
+    "if  = 24"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
-   "id": "047f9d32",
+   "id": "dde1b70c",
   "metadata": {},
   "outputs": [],
   "source": [
@@ -190,7 +198,7 @@
  {
   "cell_type": "code",
   "execution_count": 12,
-   "id": "513bbe1f",
+   "id": "06bfb217",
   "metadata": {},
   "outputs": [],
   "source": [
@@ -200,7 +208,7 @@
  {
   "cell_type": "code",
   "execution_count": 13,
-   "id": "2e484da6",
+   "id": "1fa6850b",
   "metadata": {},
   "outputs": [],
   "source": [
@@ -209,7 +217,7 @@
  },
  {
   "cell_type": "markdown",
-   "id": "b5630b79",
+   "id": "f3e9c192",
   "metadata": {},
   "source": [
    "## Sclaing Experiments\n",
@@ -228,7 +236,7 @@
  {
   "cell_type": "code",
   "execution_count": 51,
-   "id": "7945eff4",
+   "id": "5411bad5",
   "metadata": {},
   "outputs": [],
   "source": [
@@ -237,7 +245,7 @@
  },
  {
   "cell_type": "markdown",
-   "id": "b0b59bb2",
+   "id": "9c54e512",
   "metadata": {},
   "source": [
    "### Encoder Scaling"
@@ -246,7 +254,7 @@
  {
   "cell_type": "code",
   "execution_count": null,
-   "id": "681199ae",
+   "id": "9b68016d",
   "metadata": {},
   "outputs": [],
   "source": [
@@ -284,31 +292,34 @@
    "        predictor=predictor\n",
    "    )\n",
    "    forecasts = list(forecast_it)\n",
+    "    \n",
    "    if layer == layers[0]:\n",
    "        tss = list(ts_it)\n",
    "    \n",
    "    evaluator = Evaluator()\n",
    "    agg_metrics, _ = evaluator(iter(tss), iter(forecasts))\n",
    "    agg_metrics[\"trainable_parameters\"] = summarize(estimator.create_lightning_module()).trainable_parameters\n",
-    "    enc_metrics.append(agg_metrics.copy())"
+    "    enc_metrics.append(agg_metrics.copy())\n",
+    "    \n",
+    "with open(\"elec_enc_metrics.pkl\", \"wb\") as fp:\n",
+    "    pickle.dump(enc_metrics, fp)"
   ]
  },
  {
   "cell_type": "code",
-   "execution_count": 53,
-   "id": "c6cb9386",
+   "execution_count": 3,
+   "id": "2a0bc6a8",
   "metadata": {},
   "outputs": [],
   "source": [
-    "enc_metrics_out = open(\"elec_enc_metrics.pkl\", \"wb\")\n",
-    "pickle.dump(enc_metrics, enc_metrics_out)\n",
-    "enc_metrics_out.close()"
+    "with open(\"elec_enc_metrics.pkl\", \"rb\") as fp:\n",
+    "    enc_metrics = pickle.load(fp)"
   ]
  },
  {
   "cell_type": "code",
-   "execution_count": 66,
-   "id": "513cbbbf",
+   "execution_count": 4,
+   "id": "87875548",
   "metadata": {
    "scrolled": true
   },
@@ -319,7 +330,7 @@
       "Text(0.5, 1.0, 'Encoder Scaling')"
      ]
     },
-     "execution_count": 66,
+     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    },
@@ -348,8 +359,8 @@
  },
  {
   "cell_type": "code",
-   "execution_count": 68,
-   "id": "af5715b2",
+   "execution_count": 7,
+   "id": "bc9a04af",
   "metadata": {},
   "outputs": [
    {
@@ -358,13 +369,13 @@
       "Text(0.5, 1.0, 'Encoder Scaling')"
      ]
     },
-     "execution_count": 68,
+     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
-      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYoAAAEWCAYAAAB42tAoAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjUuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8qNh9FAAAACXBIWXMAAAsTAAALEwEAmpwYAABA/klEQVR4nO3dd3zc9X348ddbe50tyZJOxkNekmzZYGM77GHAGENCIAlJoc2gJaGkGU3yS7OaVUIz2+xVmhCSJiVNs0rAgBmWzTJgA97WyRPPu5Nt2drr3r8/vt8zh6yt++ru5Pfz8biH777zrbN07/tsUVWMMcaY/qQlOgBjjDHJzRKFMcaYAVmiMMYYMyBLFMYYYwZkicIYY8yALFEYY4wZkCUKY4ZBRGpF5P2JjmMgInK/iNzjPr9cROoSHZNJbZYoTMoQkX0i0iYizTGPHyY6rtEQkTtEZKeINIlIUERWiYgvXtdX1adVtTpe1zNnp4xEB2DMMN2oqk8kOojhEpEMVe3ute1K4KvASlV9RUSKgRsTEqAxA7AShRkXROR2EXlGRP5NRE6IyF4RuT5mf7GI/EJEDrv7/xyz7wMisktEjovIgyJyTsy+a91v/Cfd0ov0uu/ficgO95qPiUhFzD4VkQ+JSD1Q30fYbwKeV9VXAFT1uKr+UlWb3PNzReTfRWS/e/9nRCTX3fe/InLU3b5OROb3874sE5GDMa/3icgnRWSze+7/iEhOzP5PicgR9316v/szzBnif4MZpyxRmPHkQqAOKAG+CfxcRKIf7P8F5AHzgTLgOwAicjXwNeBdwGRgP/Bbd18J8Efg8+41dwOXRm8mIjcBnwPeDpQCTwMP9IrpZjeumj7ifQG4TkT+RUQuFZHsXvv/DVgCXAIUA58CIu6+R4BK92d5GfjNIO9NrHcBK4GZwHnA7e7PsxL4BLAcmAMsG8Y1zXimquPqAdwHhICtQzj2O8Cr7iMANCY6fnsM+P+1D2gGGmMeH3D33Q7sijk2D1CgHCcBRICiPq75c+CbMa8LgC5gBvBeYH3MPgEOAu93Xz8C3BGzPw1oBSrc1wpcPcjPdD3wF/dnaQa+DaS712oDFg7hfSl07zXRfX0/cI/7fBlwsNd7+O6Y198Efuo+vw/4Wsy+Oe515yT6/94eiX2MxxLF/Tjflgalqh9X1UWqugj4Ac63R5PcblbVwpjHf8bsOxp9oqqt7tMCYBpwXFVP9HG9c3BKEdHzmoFjwBR334GYfRr7GqgAvicijSLSCBzHSSZTYo6JPf4MqvqIqt6IU2K4CSfhvR+nBJODU4p5AxFJF5Gvi8huETmF8+GPe85QHI153orzHkGvn3ew2M3ZY9wlClVdh/MHe5qIzBaRR0Vko4g8LSJz+zj1Ns6sNjDjwwGgWEQK+9h3GOcDHwARyQcmAYeAIzhJJrpPYl+71/37XokrV1WfizlmSNMzq2pEVZ8EngIWAA1AOzC7j8P/GiepLAcm4pR+oFf7yQgcAabGvJ7W34Hm7DLuEkU/7gU+oqpLgE8CP47d6TZAzsT5IzXjjKoewakm+rGIFIlIpohc4e5+APhbEVnkthF8FXhBVfcBDwPzReTtIpIBfBSnKivqp8Bnow3JIjJRRN451LhE5CYRudWNSUTkAuBKnOquCE5V0LdF5By3FHGxG6MP6MAp+eS5McfD73Dei3kikgd8IU7XNSlu3CcKESnAaQz8XxF5FfgPnDrrWLcCv1fVnjEOzwzfX3qNo/jTEM97D07bw06cNqyPAajT1fYLwB9wvlHPxvl9QFUbgHcCX8f5UK4Eno1eUFX/BHwD+K1bBbQVp81hqE4AH8DpEXUK+DXwLVWNNkx/EtgCvIRTSv4Gzt/sr3Cqyw4B24H1w7hnv1T1EeD7wBpgV8x1O+JxfZO6xKl2HV9EZAbwkKouEJEJQJ2q9k4Osce/AnyoV5WBMWc1EZmHk/yytdcYEHN2GfclClU9BeyNVgm4RfyF0f1ue0UR8HyCQjQmaYjI20QkW0SKcEowf7EkYcZdohCRB3A+9KtF5KCI3AH8DXCHiGwCtuE0BEbdCvxWx2PRypjh+3ucqrndQA/wwcSGY5LBuKx6MsYYEz/jrkRhjDEmvsbVpIAlJSU6Y8aMRIdhjDEpY+PGjQ2qWjrQMeMqUcyYMYMNGzYkOgxjjEkZIrJ/sGOs6skYY8yALFEYY4wZkCUKY4wxA7JEYYwxZkCWKIwxxgzIEoUxxpgBWaIwxhgzIEsUxnO7w82sqQslOgxjzAhZojCe+/KD2/jQb14mErF5xYxJRZYojKdOtHTy3O5jtHb2cKixLdHhGGNGwBKF8dTj24P0uCWJuqNNCY7GGDMSliiMpx7ecgT/hGwAAiFLFMakIksUxjMnW7t4dlcDN58/hXMm5hCwEoUxKckShfHM6u1H6Y4oNyyYTFW5j7pgc6JDMsaMgCUK45lVW44wtSiX86ZOpNrvY3e4me6eSKLDMsYMkyUK44mTbV08s6uBG86djIhQ6ffR2R1h//HWRIdmjBkmzxKFiEwTkTUisl1EtonIP/ZxjIjI90Vkl4hsFpHFMfveJyL17uN9XsVpvPHE9iBdPcoN504GoNrvA7B2CmNSkJclim7g/6lqDXAR8CERqel1zPVApfu4E/gJgIgUA18CLgQuAL4kIkUexmribNWWI0wpzGXh1IkAzCkrQAQC1k5hTMrxLFGo6hFVfdl93gTsAKb0Ouwm4FfqWA8Uishk4DrgcVU9rqongMeBlV7FauLrVHsXT9c3cP2CckQEgNysdKYX5xEIWonCmFQzJm0UIjIDOB94odeuKcCBmNcH3W39be/r2neKyAYR2RAOh+MWsxm5J3cE6eyJcMN5k9+wvcrvo84ShTEpx/NEISIFwB+Aj6nqqXhfX1XvVdWlqrq0tLQ03pc3I/Dw5qNMnpjDoqmFb9he7fexr6GFju6exARmjBkRTxOFiGTiJInfqOof+zjkEDAt5vVUd1t/202Sa2rvYl19mOsXTCYtTd6wr9JfQHdE2dvQkqDojDEj4WWvJwF+DuxQ1W/3c9iDwHvd3k8XASdV9QjwGLBCRIrcRuwV7jaT5J7aGaKzO8Kbzys/Y191udPzyeZ8Mia1ZHh47UuB9wBbRORVd9vngOkAqvpTYBVwA7ALaAX+1t13XES+Arzknne3qh73MFYTJw9vPkL5hBzOn3ZmJ7VZJQVkpAn11vPJmJTiWaJQ1WcAGeQYBT7Uz777gPs8CM14pLmjm9pAmL++YPoZ1U4AWRlpzCjJtwZtY1KMjcw2cfN6tdPkfo+p9vusi6wxKcYShYmbVZuPUObLZsn0/sdGVvl9vHa8lbZO6/lkTKqwRGHioqWjmzV1Ia5fUN5ntVNUlb8AVdgVsnYKY1KFJQoTF0/tDNHRHTk9t1N/qqI9n6z6yZiUYYnCxMUjW49Q6stm6YziAY+rKM4jKyPN2imMSSGWKMyotXZ289TOECvnl5M+QLUTQEZ6GrNLCyxRGJNCLFGYUVuzM0x71+DVTlHV/gKbbtyYFGKJwozaqq1HKCnI4oKZA1c7RVWV+zh8sp1T7V0eR2aMiQdLFGZU2jp7eGpHiOuGUO0UVVXmNGjbCG1jUoMlCjMqtXUh2rp6ePMQq53g9TmfrJ3CmNRgicKMyqqtR5mUP/RqJ4AphbnkZaXb5IDGpAhLFGbE2rt6eHJHkBXzy8lIH/qvUlqaUFlWQH3IEoUxqcAShRmx2rowrZ3Dq3aKqvL7qDtqbRTGpAJLFGbEVm05QnF+FhfNGnq1U1R1uY+G5g6Ot3R6EJkxJp4sUZgRiVY7XTffP6xqp6hKvzVoG5MqLFGYEVkXCNPS2cP1C4Zf7QTOdONgicKYVGCJwozIqi1HKMzL5OLZk0Z0vn9CNhNyMqznkzEpwBKFGbaO7h6e2BHiuppyMkdQ7QQgIlSX+2zQnTEpwBKFGbanAw00d3Rz/bnlo7pOpd9HXbAJZ0VcY0yy8ixRiMh9IhISka397P8nEXnVfWwVkR4RKXb37RORLe6+DV7FaEZm1ZYjTMzN5NI5JaO6TrXfx8m2LkJNHXGKzBjjBS9LFPcDK/vbqarfUtVFqroI+CywVlWPxxxylbt/qYcxmmHq6O7h8R1BVtT4R1ztFFVlDdrGpATPEoWqrgOOD3qg4zbgAa9iMfHz7K4Gmtq7hzyl+ECq/AUA1qBtTJJLeBuFiOThlDz+ELNZgdUislFE7hzk/DtFZIOIbAiHw16GaoCHNx9lQk7GqKudACYVZFNSkGUlCmOSXMITBXAj8GyvaqfLVHUxcD3wIRG5or+TVfVeVV2qqktLS0u9jvWs1tkd4fHtR7m2ppysjPj86lT5fQSs55MxSS0ZEsWt9Kp2UtVD7r8h4E/ABQmIy/Ty7O4GTrV3c8MoezvFqvL7qA82EYlYzydjklVCE4WITASuBP4vZlu+iPiiz4EVQJ89p8zYWrX5CL7sDC6rHH21U1SV30dLZw+HGtvidk1jTHxleHVhEXkAWAaUiMhB4EtAJoCq/tQ97G3AalVtiTnVD/xJRKLx/beqPupVnGZounoirN4e5NoaP9kZ6XG7bnW506BdH2piWnFe3K5rjIkfzxKFqt42hGPux+lGG7ttD7DQm6jMSD23+xgn27ri0tsp1hx3WdS6o81cPdcf12sbY+IjGdooTApYtfkIBXGudgKYmJvJ5Ik51vPJmCRmicIMqqsnwmPbj7J8Xhk5mfGrdopyej5ZojAmWVmiMINav+cYja3xr3aKqvIXUB9qpsd6PhmTlCxRmEGt2nKE/Kx0rqjyZpxKld9HZ3eE/cdaBj/YGDPmLFGYAXX3RHhsW5Br5vk9qXYCZ1lUwAbeGZOkLFGYAb2w9zjHWzo9q3YCmFPmdJG1dgpjkpMlCjOgh7ccIS8rnWXV3k2PkpeVwfTiPOosURiTlCxRmH5190R4bOtRrp7rTW+nWNGpPIwxyccShenXi3uPc6yLine truncated
+      "image/png": "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 truncated
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
@@ -378,17 +389,17 @@
   "source": [
    "plt.plot(\n",
    "    [metrics[\"trainable_parameters\"] for metrics in enc_metrics],\n",
-    "    [metrics[\"MSE\"] for metrics in enc_metrics],   \n",
+    "    [metrics[\"MASE\"] for metrics in enc_metrics],   \n",
    ")\n",
    "plt.xlabel(\"trainable parameters\")\n",
-    "plt.ylabel(\"MSE\")\n",
+    "plt.ylabel(\"MASE\")\n",
    "plt.title(\"Encoder Scaling\")"
   ]
  },
  {
   "cell_type": "code",
-   "execution_count": 79,
-   "id": "1c64e02c",
+   "execution_count": 6,
+   "id": "0a6ab5ee",
   "metadata": {},
   "outputs": [
    {
@@ -397,7 +408,7 @@
       "Text(0.5, 1.0, 'Encoder Scaling')"
      ]
     },
-     "execution_count": 79,
+     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    },
@@ -426,7 +437,7 @@
  },
  {
   "cell_type": "markdown",
-   "id": "13356916",
+   "id": "2308a10f",
   "metadata": {},
   "source": [
    "### Decoder Scaling"
@@ -435,7 +446,7 @@
  {
   "cell_type": "code",
   "execution_count": null,
-   "id": "a175e127",
+   "id": "476beeac",
   "metadata": {},
   "outputs": [],
   "source": [
@@ -479,24 +490,143 @@
    "    evaluator = Evaluator()\n",
    "    agg_metrics, _ = evaluator(iter(tss), iter(forecasts))\n",
    "    agg_metrics[\"trainable_parameters\"] = summarize(estimator.create_lightning_module()).trainable_parameters\n",
-    "    dec_metrics.append(agg_metrics.copy())"
+    "    dec_metrics.append(agg_metrics.copy())\n",
+    "    \n",
+    "with open(\"elec_dec_metrics.pkl\", \"wb\") as fp:\n",
+    "    pickle.dump(dec_metrics, fp)"
   ]
  },
  {
   "cell_type": "code",
-   "execution_count": null,
-   "id": "c439510e",
+   "execution_count": 8,
+   "id": "2f00decc",
   "metadata": {},
   "outputs": [],
   "source": [
-    "dec_metrics_out = open(\"elec_dec_metrics.pkl\", \"wb\")\n",
-    "pickle.dump(dec_metrics, dec_metrics_out)\n",
-    "dec_metrics_out.close()"
+    "with open(\"elec_dec_metrics.pkl\", \"rb\") as fp:\n",
+    "    dec_metrics = pickle.load(fp)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 9,
+   "id": "fc3311e9",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "Text(0.5, 1.0, 'Decoder Scaling')"
+      ]
+     },
+     "execution_count": 9,
+     "metadata": {},
+     "output_type": "execute_result"
+    },
+    {
+     "data": {
+      "image/png": "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 truncated
+      "text/plain": [
+       "<Figure size 432x288 with 1 Axes>"
+      ]
+     },
+     "metadata": {
+      "needs_background": "light"
+     },
+     "output_type": "display_data"
+    }
+   ],
+   "source": [
+    "plt.plot(\n",
+    "    [metrics[\"trainable_parameters\"] for metrics in dec_metrics],\n",
+    "    [metrics[\"mean_wQuantileLoss\"] for metrics in dec_metrics],   \n",
+    ")\n",
+    "plt.xlabel(\"trainable parameters\")\n",
+    "plt.ylabel(\"CRPS\")\n",
+    "plt.title(\"Decoder Scaling\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 10,
+   "id": "be4e5502",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "Text(0.5, 1.0, 'Decoder Scaling')"
+      ]
+     },
+     "execution_count": 10,
+     "metadata": {},
+     "output_type": "execute_result"
+    },
+    {
+     "data": {
+      "image/png": "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 truncated
+      "text/plain": [
+       "<Figure size 432x288 with 1 Axes>"
+      ]
+     },
+     "metadata": {
+      "needs_background": "light"
+     },
+     "output_type": "display_data"
+    }
+   ],
+   "source": [
+    "plt.plot(\n",
+    "    [metrics[\"trainable_parameters\"] for metrics in dec_metrics],\n",
+    "    [metrics[\"MASE\"] for metrics in dec_metrics],   \n",
+    ")\n",
+    "plt.xlabel(\"trainable parameters\")\n",
+    "plt.ylabel(\"MASE\")\n",
+    "plt.title(\"Decoder Scaling\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 11,
+   "id": "5d6e2046",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "Text(0.5, 1.0, 'Decoder Scaling')"
+      ]
+     },
+     "execution_count": 11,
+     "metadata": {},
+     "output_type": "execute_result"
+    },
+    {
+     "data": {
+      "image/png": "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 truncated
+      "text/plain": [
+       "<Figure size 432x288 with 1 Axes>"
+      ]
+     },
+     "metadata": {
+      "needs_background": "light"
+     },
+     "output_type": "display_data"
+    }
+   ],
+   "source": [
+    "plt.plot(\n",
+    "    [metrics[\"trainable_parameters\"] for metrics in dec_metrics],\n",
+    "    [metrics[\"NRMSE\"] for metrics in dec_metrics],   \n",
+    ")\n",
+    "plt.xlabel(\"trainable parameters\")\n",
+    "plt.ylabel(\"NRMSE\")\n",
+    "plt.title(\"Decoder Scaling\")"
   ]
  },
  {
   "cell_type": "markdown",
-   "id": "78be6dd9",
+   "id": "4338c927",
   "metadata": {},
   "source": [
    "### Symmetric Scaling"
@@ -505,7 +635,7 @@
  {
   "cell_type": "code",
   "execution_count": null,
-   "id": "069b20d9",
+   "id": "88b5c69d",
   "metadata": {},
   "outputs": [],
   "source": [
@@ -549,19 +679,10 @@
    "    evaluator = Evaluator()\n",
    "    agg_metrics, _ = evaluator(iter(tss), iter(forecasts))\n",
    "    agg_metrics[\"trainable_parameters\"] = summarize(estimator.create_lightning_module()).trainable_parameters\n",
-    "    sym_metrics.append(agg_metrics.copy())"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "id": "9d48d029",
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "sym_metrics_out = open(\"elec_sym_metrics.pkl\", \"wb\")\n",
-    "pickle.dump(sym_metrics, sym_metrics_out)\n",
-    "sym_metrics_out.close()"
+    "    sym_metrics.append(agg_metrics.copy())\\\n",
+    "\n",
+    "with open(\"elec_sym_metrics.pkl\", \"wb\") as fp:\n",
+    "    pickle.dump(sym_metrics, fp)"
   ]
  }
 ],