metaphase_testing.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 158,
   "metadata": {},
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "from tensorflow import keras\n",
    "from keras import layers, models, optimizers\n",
    "import numpy as np\n",
    "import datetime\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 193,
   "metadata": {},
   "outputs": [],
   "source": [
    "def datagen(tasks, sampl):\n",
    "    x = tf.random.uniform([tasks, sampl, 1], -1.0, 1.0)\n",
    "    y = tf.sin(x * tf.random.uniform([tasks, 1, 1], 0.0, 8.0) + tf.random.uniform([tasks, 1, 1], 0.0, 3.0)) * tf.random.uniform([tasks, 1, 1], 0.1, 5.0)\n",
    "    l = y[:, -1:, :]\n",
    "    y = tf.concat([y[:, :-1, :], tf.zeros([tasks, 1, 1])], axis=1)\n",
    "    label_indicator = tf.concat([tf.ones([tasks, sampl-1, 1]), tf.zeros([tasks, 1, 1])], axis=1)\n",
    "    return tf.concat([x, label_indicator, y], axis=2), l\n",
    "\n",
    "def datagen_constant(tasks, sampl):\n",
    "    x = tf.random.normal([tasks, sampl, 1])\n",
    "    y = x * tf.random.uniform([tasks, 1, 1], 0.1, .5)\n",
    "    l = y[:, -1:, :]\n",
    "    y = tf.concat([y[:, :-1, :], tf.zeros([tasks, 1, 1])], axis=1)\n",
    "    label_indicator = tf.concat([tf.ones([tasks, sampl-1, 1]), tf.zeros([tasks, 1, 1])], axis=1)\n",
    "    return tf.concat([x, label_indicator, y], axis=2), l\n"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "By default, all Transformers have a key, value, and query size of 32, 8 heads, and 4 layers, and model size of NM = 256. The model size defines the dimensionality of each token, and the MLP between layers scales this size up to a hidden representation of 4 × NM where NM corresponds to the model size."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 160,
   "metadata": {},
   "outputs": [],
   "source": [
    "class BaseAttention(layers.Layer):\n",
    "  def __init__(self, **kwargs):\n",
    "    super().__init__()\n",
    "    self.mha = layers.MultiHeadAttention(**kwargs)\n",
    "    self.layernorm = layers.LayerNormalization()\n",
    "    self.add = layers.Add()\n",
    "\n",
    "class GSA(BaseAttention):\n",
    "  def call(self, x):\n",
    "    attn_output = self.mha(\n",
    "        query=x,\n",
    "        value=x,\n",
    "        key=x)\n",
    "    x = self.add([x, attn_output])\n",
    "    x = self.layernorm(x)\n",
    "    return x\n",
    "\n",
    "class FeedForward(layers.Layer):\n",
    "  def __init__(self, d_model, dff):\n",
    "    super().__init__()\n",
    "    self.seq = models.Sequential([\n",
    "      layers.Dense(dff, activation='relu'),\n",
    "      layers.Dense(d_model)\n",
    "    ])\n",
    "    self.add = layers.Add()\n",
    "\n",
    "  def call(self, x):\n",
    "    x = self.add([x, self.seq(x)])\n",
    "    return x\n",
    "\n",
    "class TransformLayer(layers.Layer):\n",
    "  def __init__(self,*, d_model, num_heads, dff):\n",
    "    super().__init__()\n",
    "\n",
    "    self.self_attention = GSA(\n",
    "        num_heads=num_heads,\n",
    "        key_dim=d_model)\n",
    "\n",
    "    self.ffn = FeedForward(d_model, dff)\n",
    "\n",
    "  def call(self, x):\n",
    "    x = self.self_attention(x)\n",
    "    x = self.ffn(x)\n",
    "    return x\n",
    "\n",
    "class Transformer(models.Model):\n",
    "  def __init__(self, *, num_layers, d_model, num_heads, dff):\n",
    "    super().__init__()\n",
    "    self.embedder = layers.Dense(d_model)\n",
    "    self.layerstack = [TransformLayer(d_model=d_model, num_heads=num_heads, dff=dff) for _ in range(num_layers)]\n",
    "\n",
    "  def call(self, x):\n",
    "    x = self.embedder(x)\n",
    "    for i in range(3):\n",
    "      for layer in self.layerstack:\n",
    "        x = layer(x)\n",
    "    return x[:, -1:, -1]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 161,
   "metadata": {},
   "outputs": [],
   "source": [
    "sampl = 100\n",
    "\n",
    "mdim = 16\n",
    "head = 2\n",
    "mlay = 8\n",
    "assert mdim % head == 0\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 178,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(400, 100, 3) (400, 1, 1)\n",
      "(100, 100, 3) (100, 1, 1)\n",
      "82\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "train_s, train_l = datagen(400, sampl)\n",
    "valid_s, valid_l = datagen_constant(100, sampl)\n",
    "print(train_s.shape, train_l.shape)\n",
    "print(valid_s.shape, valid_l.shape)\n",
    "\n",
    "random_sample_number = np.random.randint(0, 99) \n",
    "print(random_sample_number)\n",
    "\n",
    "# plot data points, no lines\n",
    "plt.plot(train_s[random_sample_number, :-1, 0], train_s[random_sample_number, :-1, -1], 'o')\n",
    "# plot the last 1 points as red points\n",
    "\n",
    "plt.plot(train_s[random_sample_number, -1:, 0], train_l[random_sample_number], 'ro')\n",
    "plt.show()\n",
    "\n",
    "plt.plot(valid_s[random_sample_number, :-1, 0], valid_s[random_sample_number, :-1, -1], 'o')\n",
    "# plot the last 1 points as red points\n",
    "plt.plot(valid_s[random_sample_number, -1:, 0], valid_l[random_sample_number], 'ro')\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 116,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The tensorboard extension is already loaded. To reload it, use:\n",
      "  %reload_ext tensorboard\n"
     ]
    }
   ],
   "source": [
    "\n",
    "%load_ext tensorboard\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 117,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "gpicl = Transformer(num_layers=mlay, d_model=mdim, num_heads=head, dff=mdim*4)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 183,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/500\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2023-03-31 11:50:09.454720: I tensorflow/core/grappler/optimizers/custom_graph_optimizer_registry.cc:113] Plugin optimizer for device_type GPU is enabled.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2/2 [==============================] - ETA: 0s - loss: 1.8621 - mae: 0.9646"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2023-03-31 11:50:14.423947: I tensorflow/core/grappler/optimizers/custom_graph_optimizer_registry.cc:113] Plugin optimizer for device_type GPU is enabled.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2/2 [==============================] - 8s 2s/step - loss: 1.8621 - mae: 0.9646 - val_loss: 0.3385 - val_mae: 0.5073\n",
      "Epoch 2/500\n",
      "2/2 [==============================] - 2s 1s/step - loss: 2.4482 - mae: 1.1363 - val_loss: 0.3989 - val_mae: 0.5903\n",
      "Epoch 3/500\n",
      "2/2 [==============================] - 2s 1s/step - loss: 2.9371 - mae: 1.2463 - val_loss: 0.2445 - val_mae: 0.4257\n",
      "Epoch 4/500\n",
      "2/2 [==============================] - 2s 931ms/step - loss: 2.2745 - mae: 0.9804 - val_loss: 0.7135 - val_mae: 0.6477\n",
      "Epoch 5/500\n",
      "2/2 [==============================] - 2s 929ms/step - loss: 2.8303 - mae: 1.2484 - val_loss: 0.2883 - val_mae: 0.4459\n",
      "Epoch 6/500\n",
      "2/2 [==============================] - 2s 912ms/step - loss: 1.9818 - mae: 0.9997 - val_loss: 0.3598 - val_mae: 0.4778\n",
      "Epoch 7/500\n",
      "2/2 [==============================] - 2s 917ms/step - loss: 1.8789 - mae: 0.9697 - val_loss: 0.3805 - val_mae: 0.5367\n",
      "Epoch 8/500\n",
      "2/2 [==============================] - 2s 913ms/step - loss: 1.6971 - mae: 0.9126 - val_loss: 0.5146 - val_mae: 0.6626\n",
      "Epoch 9/500\n",
      "2/2 [==============================] - 2s 962ms/step - loss: 1.2863 - mae: 0.7578 - val_loss: 0.4742 - val_mae: 0.6239\n",
      "Epoch 10/500\n",
      "2/2 [==============================] - 2s 936ms/step - loss: 1.5000 - mae: 0.8248 - val_loss: 0.3958 - val_mae: 0.5432\n",
      "Epoch 11/500\n",
      "2/2 [==============================] - 2s 901ms/step - loss: 3.0734 - mae: 1.1643 - val_loss: 0.8599 - val_mae: 0.7346\n",
      "Epoch 12/500\n",
      "2/2 [==============================] - 2s 921ms/step - loss: 2.8663 - mae: 1.2824 - val_loss: 1.0889 - val_mae: 0.9534\n",
      "Epoch 13/500\n",
      "2/2 [==============================] - 2s 917ms/step - loss: 2.0545 - mae: 1.0190 - val_loss: 0.9766 - val_mae: 0.8909\n",
      "Epoch 14/500\n",
      "2/2 [==============================] - 2s 911ms/step - loss: 1.6530 - mae: 0.8731 - val_loss: 0.5514 - val_mae: 0.6486\n",
      "Epoch 15/500\n",
      "2/2 [==============================] - 2s 910ms/step - loss: 2.4091 - mae: 1.0888 - val_loss: 1.0331 - val_mae: 0.7908\n",
      "Epoch 16/500\n",
      "2/2 [==============================] - 2s 935ms/step - loss: 2.3330 - mae: 1.1574 - val_loss: 0.4319 - val_mae: 0.5910\n",
      "Epoch 17/500\n",
      "2/2 [==============================] - 2s 901ms/step - loss: 1.8987 - mae: 0.9793 - val_loss: 0.4822 - val_mae: 0.5997\n",
      "Epoch 18/500\n",
      "2/2 [==============================] - 2s 890ms/step - loss: 1.7954 - mae: 0.9003 - val_loss: 0.3451 - val_mae: 0.4949\n",
      "Epoch 19/500\n",
      "2/2 [==============================] - 2s 876ms/step - loss: 1.3326 - mae: 0.7708 - val_loss: 0.3500 - val_mae: 0.5132\n",
      "Epoch 20/500\n",
      "2/2 [==============================] - 2s 950ms/step - loss: 1.3431 - mae: 0.7822 - val_loss: 0.4142 - val_mae: 0.5734\n",
      "Epoch 21/500\n",
      "2/2 [==============================] - 2s 925ms/step - loss: 1.2654 - mae: 0.7663 - val_loss: 0.4107 - val_mae: 0.5631\n",
      "Epoch 22/500\n",
      "2/2 [==============================] - 2s 941ms/step - loss: 1.3993 - mae: 0.8284 - val_loss: 0.4068 - val_mae: 0.5522\n",
      "Epoch 23/500\n",
      "2/2 [==============================] - 2s 929ms/step - loss: 2.8165 - mae: 1.1920 - val_loss: 0.3585 - val_mae: 0.4867\n",
      "Epoch 24/500\n",
      "2/2 [==============================] - 2s 922ms/step - loss: 2.2738 - mae: 1.0368 - val_loss: 0.3075 - val_mae: 0.4865\n",
      "Epoch 25/500\n",
      "2/2 [==============================] - 2s 954ms/step - loss: 1.5330 - mae: 0.8060 - val_loss: 0.3292 - val_mae: 0.4958\n",
      "Epoch 26/500\n",
      "2/2 [==============================] - 2s 914ms/step - loss: 1.2114 - mae: 0.7079 - val_loss: 0.3710 - val_mae: 0.5273\n",
      "Epoch 27/500\n",
      "2/2 [==============================] - 2s 903ms/step - loss: 1.1627 - mae: 0.7023 - val_loss: 0.3583 - val_mae: 0.5132\n",
      "Epoch 28/500\n",
      "2/2 [==============================] - 2s 877ms/step - loss: 1.6062 - mae: 0.8198 - val_loss: 0.4960 - val_mae: 0.5985\n",
      "Epoch 29/500\n",
      "2/2 [==============================] - 2s 885ms/step - loss: 1.4323 - mae: 0.8635 - val_loss: 0.6645 - val_mae: 0.7152\n",
      "Epoch 30/500\n",
      "2/2 [==============================] - 2s 916ms/step - loss: 2.7502 - mae: 1.2122 - val_loss: 1.9016 - val_mae: 1.0811\n",
      "Epoch 31/500\n",
      "2/2 [==============================] - 2s 941ms/step - loss: 2.9243 - mae: 1.3275 - val_loss: 0.8553 - val_mae: 0.8289\n",
      "Epoch 32/500\n",
      "2/2 [==============================] - 2s 965ms/step - loss: 1.9705 - mae: 1.0177 - val_loss: 0.9023 - val_mae: 0.8261\n",
      "Epoch 33/500\n",
      "2/2 [==============================] - 2s 931ms/step - loss: 1.9223 - mae: 1.0203 - val_loss: 0.3032 - val_mae: 0.4454\n",
      "Epoch 34/500\n",
      "2/2 [==============================] - 2s 898ms/step - loss: 2.9791 - mae: 1.1783 - val_loss: 0.2117 - val_mae: 0.3627\n",
      "Epoch 35/500\n",
      "2/2 [==============================] - 2s 866ms/step - loss: 1.8200 - mae: 0.9471 - val_loss: 0.3184 - val_mae: 0.4581\n",
      "Epoch 36/500\n",
      "2/2 [==============================] - 2s 913ms/step - loss: 1.6114 - mae: 0.8377 - val_loss: 0.4813 - val_mae: 0.5399\n",
      "Epoch 37/500\n",
      "2/2 [==============================] - 2s 927ms/step - loss: 1.7237 - mae: 0.8674 - val_loss: 0.4115 - val_mae: 0.5361\n",
      "Epoch 38/500\n",
      "2/2 [==============================] - 2s 901ms/step - loss: 1.3909 - mae: 0.8200 - val_loss: 0.3956 - val_mae: 0.5120\n",
      "Epoch 39/500\n",
      "2/2 [==============================] - 2s 891ms/step - loss: 1.3502 - mae: 0.7818 - val_loss: 0.3831 - val_mae: 0.5299\n",
      "Epoch 40/500\n",
      "2/2 [==============================] - 2s 889ms/step - loss: 1.0480 - mae: 0.6714 - val_loss: 0.3814 - val_mae: 0.5248\n",
      "Epoch 41/500\n",
      "2/2 [==============================] - 2s 907ms/step - loss: 1.1287 - mae: 0.7154 - val_loss: 0.3556 - val_mae: 0.5106\n",
      "Epoch 42/500\n",
      "2/2 [==============================] - 2s 931ms/step - loss: 1.4073 - mae: 0.8212 - val_loss: 0.3940 - val_mae: 0.5277\n",
      "Epoch 43/500\n",
      "2/2 [==============================] - 2s 906ms/step - loss: 1.6584 - mae: 0.8708 - val_loss: 0.4060 - val_mae: 0.5500\n",
      "Epoch 44/500\n",
      "2/2 [==============================] - 2s 936ms/step - loss: 1.2024 - mae: 0.7274 - val_loss: 0.6338 - val_mae: 0.7044\n",
      "Epoch 45/500\n",
      "2/2 [==============================] - 2s 916ms/step - loss: 1.5437 - mae: 0.8761 - val_loss: 0.3230 - val_mae: 0.4909\n",
      "Epoch 46/500\n",
      "2/2 [==============================] - 2s 935ms/step - loss: 1.1289 - mae: 0.7112 - val_loss: 0.3651 - val_mae: 0.5359\n",
      "Epoch 47/500\n",
      "2/2 [==============================] - 2s 950ms/step - loss: 1.3405 - mae: 0.7375 - val_loss: 0.3630 - val_mae: 0.4937\n",
      "Epoch 48/500\n",
      "2/2 [==============================] - 2s 921ms/step - loss: 2.7037 - mae: 1.1612 - val_loss: 0.2922 - val_mae: 0.3797\n",
      "Epoch 49/500\n",
      "2/2 [==============================] - 2s 897ms/step - loss: 1.8107 - mae: 0.9432 - val_loss: 0.3398 - val_mae: 0.4843\n",
      "Epoch 50/500\n",
      "2/2 [==============================] - 2s 912ms/step - loss: 1.1759 - mae: 0.7122 - val_loss: 0.4189 - val_mae: 0.5479\n",
      "Epoch 51/500\n",
      "2/2 [==============================] - 2s 1s/step - loss: 1.0952 - mae: 0.6565 - val_loss: 0.4791 - val_mae: 0.5868\n",
      "Epoch 52/500\n",
      "2/2 [==============================] - 2s 911ms/step - loss: 1.1451 - mae: 0.6770 - val_loss: 0.5453 - val_mae: 0.6185\n",
      "Epoch 53/500\n",
      "2/2 [==============================] - 2s 992ms/step - loss: 1.7874 - mae: 0.9414 - val_loss: 0.3547 - val_mae: 0.4982\n",
      "Epoch 54/500\n",
      "2/2 [==============================] - 2s 920ms/step - loss: 1.3853 - mae: 0.8198 - val_loss: 0.4450 - val_mae: 0.5842\n",
      "Epoch 55/500\n",
      "2/2 [==============================] - 2s 956ms/step - loss: 0.9322 - mae: 0.6055 - val_loss: 0.5006 - val_mae: 0.6078\n",
      "Epoch 56/500\n",
      "2/2 [==============================] - 2s 975ms/step - loss: 0.9548 - mae: 0.6269 - val_loss: 0.4868 - val_mae: 0.6108\n",
      "Epoch 57/500\n",
      "2/2 [==============================] - 2s 928ms/step - loss: 0.9719 - mae: 0.6348 - val_loss: 0.4556 - val_mae: 0.6000\n",
      "Epoch 58/500\n",
      "2/2 [==============================] - 2s 872ms/step - loss: 0.9032 - mae: 0.6275 - val_loss: 0.6478 - val_mae: 0.7017\n",
      "Epoch 59/500\n",
      "2/2 [==============================] - 2s 908ms/step - loss: 1.8994 - mae: 0.9767 - val_loss: 0.2065 - val_mae: 0.3938\n",
      "Epoch 60/500\n",
      "2/2 [==============================] - 2s 954ms/step - loss: 2.3846 - mae: 1.0749 - val_loss: 0.1775 - val_mae: 0.2805\n",
      "Epoch 61/500\n",
      "2/2 [==============================] - 2s 888ms/step - loss: 1.5210 - mae: 0.8248 - val_loss: 0.5116 - val_mae: 0.5930\n",
      "Epoch 62/500\n",
      "2/2 [==============================] - 2s 950ms/step - loss: 1.2263 - mae: 0.7319 - val_loss: 0.4446 - val_mae: 0.5290\n",
      "Epoch 63/500\n",
      "2/2 [==============================] - 2s 986ms/step - loss: 0.9194 - mae: 0.6402 - val_loss: 0.3876 - val_mae: 0.4988\n",
      "Epoch 64/500\n",
      "2/2 [==============================] - 2s 944ms/step - loss: 0.8480 - mae: 0.5985 - val_loss: 0.4605 - val_mae: 0.5516\n",
      "Epoch 65/500\n",
      "2/2 [==============================] - 2s 911ms/step - loss: 0.8441 - mae: 0.5651 - val_loss: 0.5731 - val_mae: 0.6632\n",
      "Epoch 66/500\n",
      "2/2 [==============================] - 2s 915ms/step - loss: 1.0456 - mae: 0.6585 - val_loss: 0.3610 - val_mae: 0.4626\n",
      "Epoch 67/500\n",
      "2/2 [==============================] - 2s 927ms/step - loss: 1.6985 - mae: 0.8405 - val_loss: 0.4312 - val_mae: 0.5590\n",
      "Epoch 68/500\n",
      "2/2 [==============================] - 2s 923ms/step - loss: 1.3761 - mae: 0.8100 - val_loss: 0.4890 - val_mae: 0.5750\n",
      "Epoch 69/500\n",
      "2/2 [==============================] - 2s 902ms/step - loss: 1.2598 - mae: 0.7061 - val_loss: 0.5219 - val_mae: 0.6086\n",
      "Epoch 70/500\n",
      "2/2 [==============================] - 2s 914ms/step - loss: 1.1568 - mae: 0.6867 - val_loss: 0.4452 - val_mae: 0.5634\n",
      "Epoch 71/500\n",
      "2/2 [==============================] - 2s 970ms/step - loss: 0.7929 - mae: 0.5756 - val_loss: 0.4725 - val_mae: 0.5977\n",
      "Epoch 72/500\n",
      "2/2 [==============================] - 2s 910ms/step - loss: 1.0640 - mae: 0.6618 - val_loss: 0.2944 - val_mae: 0.4738\n",
      "Epoch 73/500\n",
      "2/2 [==============================] - 2s 905ms/step - loss: 1.0166 - mae: 0.6594 - val_loss: 0.3307 - val_mae: 0.4981\n",
      "Epoch 74/500\n",
      "2/2 [==============================] - 2s 932ms/step - loss: 0.9580 - mae: 0.6404 - val_loss: 0.6024 - val_mae: 0.6803\n",
      "Epoch 75/500\n",
      "2/2 [==============================] - 2s 928ms/step - loss: 0.9089 - mae: 0.6070 - val_loss: 0.4308 - val_mae: 0.5326\n",
      "Epoch 76/500\n",
      "2/2 [==============================] - 2s 950ms/step - loss: 0.7323 - mae: 0.5221 - val_loss: 0.5116 - val_mae: 0.5839\n",
      "Epoch 77/500\n",
      "2/2 [==============================] - 2s 907ms/step - loss: 0.6806 - mae: 0.5027 - val_loss: 0.5198 - val_mae: 0.6037\n",
      "Epoch 78/500\n",
      "2/2 [==============================] - 2s 932ms/step - loss: 0.9001 - mae: 0.6132 - val_loss: 0.4777 - val_mae: 0.5931\n",
      "Epoch 79/500\n",
      "2/2 [==============================] - 2s 929ms/step - loss: 0.9352 - mae: 0.6417 - val_loss: 0.6313 - val_mae: 0.6573\n",
      "Epoch 80/500\n",
      "2/2 [==============================] - 2s 938ms/step - loss: 2.9368 - mae: 1.1690 - val_loss: 0.1029 - val_mae: 0.2493\n",
      "Epoch 81/500\n",
      "2/2 [==============================] - 2s 917ms/step - loss: 3.3800 - mae: 1.3979 - val_loss: 1.6623 - val_mae: 0.9470\n",
      "Epoch 82/500\n",
      "2/2 [==============================] - 2s 896ms/step - loss: 2.3326 - mae: 1.1560 - val_loss: 0.1960 - val_mae: 0.3471\n",
      "Epoch 83/500\n",
      "2/2 [==============================] - 2s 958ms/step - loss: 1.8218 - mae: 0.9058 - val_loss: 0.1791 - val_mae: 0.2929\n",
      "Epoch 84/500\n",
      "2/2 [==============================] - 2s 967ms/step - loss: 1.7851 - mae: 0.8880 - val_loss: 0.1644 - val_mae: 0.2756\n",
      "Epoch 85/500\n",
      "2/2 [==============================] - 2s 925ms/step - loss: 2.3551 - mae: 1.0918 - val_loss: 0.3827 - val_mae: 0.4727\n",
      "Epoch 86/500\n",
      "2/2 [==============================] - 2s 959ms/step - loss: 1.3610 - mae: 0.8117 - val_loss: 0.5763 - val_mae: 0.6130\n",
      "Epoch 87/500\n",
      "2/2 [==============================] - 2s 914ms/step - loss: 1.2169 - mae: 0.7508 - val_loss: 1.0338 - val_mae: 0.8993\n",
      "Epoch 88/500\n",
      "2/2 [==============================] - 2s 900ms/step - loss: 0.9634 - mae: 0.6236 - val_loss: 0.6037 - val_mae: 0.6368\n",
      "Epoch 89/500\n",
      "2/2 [==============================] - 2s 969ms/step - loss: 0.8082 - mae: 0.5518 - val_loss: 0.4921 - val_mae: 0.5585\n",
      "Epoch 90/500\n",
      "2/2 [==============================] - 2s 950ms/step - loss: 1.4370 - mae: 0.7305 - val_loss: 0.4414 - val_mae: 0.5485\n",
      "Epoch 91/500\n",
      "2/2 [==============================] - 2s 919ms/step - loss: 1.2568 - mae: 0.7005 - val_loss: 0.5020 - val_mae: 0.5767\n",
      "Epoch 92/500\n",
      "2/2 [==============================] - 2s 899ms/step - loss: 0.9813 - mae: 0.6779 - val_loss: 0.6176 - val_mae: 0.6738\n",
      "Epoch 93/500\n",
      "2/2 [==============================] - 2s 882ms/step - loss: 0.8384 - mae: 0.5905 - val_loss: 0.4610 - val_mae: 0.5703\n",
      "Epoch 94/500\n",
      "2/2 [==============================] - 2s 922ms/step - loss: 0.7785 - mae: 0.5502 - val_loss: 0.4661 - val_mae: 0.5621\n",
      "Epoch 95/500\n",
      "2/2 [==============================] - 2s 960ms/step - loss: 0.7843 - mae: 0.5584 - val_loss: 0.6275 - val_mae: 0.6698\n",
      "Epoch 96/500\n",
      "2/2 [==============================] - 2s 954ms/step - loss: 0.8085 - mae: 0.5728 - val_loss: 0.4396 - val_mae: 0.5473\n",
      "Epoch 97/500\n",
      "2/2 [==============================] - 2s 910ms/step - loss: 0.6331 - mae: 0.4979 - val_loss: 0.4528 - val_mae: 0.5593\n",
      "Epoch 98/500\n",
      "2/2 [==============================] - 2s 880ms/step - loss: 0.8804 - mae: 0.5857 - val_loss: 0.5332 - val_mae: 0.6599\n",
      "Epoch 99/500\n",
      "2/2 [==============================] - 2s 926ms/step - loss: 1.3275 - mae: 0.7802 - val_loss: 0.4793 - val_mae: 0.5956\n",
      "Epoch 100/500\n",
      "2/2 [==============================] - 2s 1s/step - loss: 1.5838 - mae: 0.8479 - val_loss: 0.5445 - val_mae: 0.6172\n",
      "Epoch 101/500\n",
      "2/2 [==============================] - 2s 942ms/step - loss: 1.6308 - mae: 0.9038 - val_loss: 0.4304 - val_mae: 0.5444\n",
      "Epoch 102/500\n",
      "2/2 [==============================] - 2s 926ms/step - loss: 0.7503 - mae: 0.5423 - val_loss: 0.5566 - val_mae: 0.6505\n",
      "Epoch 103/500\n",
      "2/2 [==============================] - 2s 950ms/step - loss: 0.8588 - mae: 0.5641 - val_loss: 0.4412 - val_mae: 0.5695\n",
      "Epoch 104/500\n",
      "2/2 [==============================] - 2s 931ms/step - loss: 0.6802 - mae: 0.5001 - val_loss: 0.5638 - val_mae: 0.6381\n",
      "Epoch 105/500\n",
      "2/2 [==============================] - 2s 909ms/step - loss: 0.7919 - mae: 0.5688 - val_loss: 0.4983 - val_mae: 0.6108\n",
      "Epoch 106/500\n",
      "2/2 [==============================] - 2s 944ms/step - loss: 0.7885 - mae: 0.5740 - val_loss: 0.5884 - val_mae: 0.6676\n",
      "Epoch 107/500\n",
      "2/2 [==============================] - 2s 945ms/step - loss: 0.9029 - mae: 0.5812 - val_loss: 0.5390 - val_mae: 0.6154\n",
      "Epoch 108/500\n",
      "2/2 [==============================] - 2s 937ms/step - loss: 0.7882 - mae: 0.5638 - val_loss: 0.5452 - val_mae: 0.6188\n",
      "Epoch 109/500\n",
      "2/2 [==============================] - 2s 940ms/step - loss: 0.6572 - mae: 0.5243 - val_loss: 0.5735 - val_mae: 0.6484\n",
      "Epoch 110/500\n",
      "2/2 [==============================] - 2s 913ms/step - loss: 0.5897 - mae: 0.4767 - val_loss: 0.5612 - val_mae: 0.6188\n",
      "Epoch 111/500\n",
      "2/2 [==============================] - 2s 917ms/step - loss: 0.9375 - mae: 0.6242 - val_loss: 0.3735 - val_mae: 0.5328\n",
      "Epoch 112/500\n",
      "2/2 [==============================] - 2s 939ms/step - loss: 0.6950 - mae: 0.5139 - val_loss: 0.4363 - val_mae: 0.5527\n",
      "Epoch 113/500\n",
      "2/2 [==============================] - 2s 948ms/step - loss: 0.6026 - mae: 0.4666 - val_loss: 0.4265 - val_mae: 0.5149\n",
      "Epoch 114/500\n",
      "2/2 [==============================] - 2s 937ms/step - loss: 1.0248 - mae: 0.6474 - val_loss: 0.5237 - val_mae: 0.6200\n",
      "Epoch 115/500\n",
      "2/2 [==============================] - 2s 900ms/step - loss: 1.3032 - mae: 0.7820 - val_loss: 0.4859 - val_mae: 0.5047\n",
      "Epoch 116/500\n",
      "2/2 [==============================] - 2s 924ms/step - loss: 1.3288 - mae: 0.8015 - val_loss: 0.8832 - val_mae: 0.8005\n",
      "Epoch 117/500\n",
      "2/2 [==============================] - 2s 930ms/step - loss: 2.0989 - mae: 0.9548 - val_loss: 0.3071 - val_mae: 0.4115\n",
      "Epoch 118/500\n",
      "2/2 [==============================] - 2s 932ms/step - loss: 1.3042 - mae: 0.7582 - val_loss: 0.4207 - val_mae: 0.5240\n",
      "Epoch 119/500\n",
      "2/2 [==============================] - 2s 921ms/step - loss: 0.8315 - mae: 0.5587 - val_loss: 0.4669 - val_mae: 0.5350\n",
      "Epoch 120/500\n",
      "2/2 [==============================] - 2s 969ms/step - loss: 0.8807 - mae: 0.6090 - val_loss: 0.4559 - val_mae: 0.5600\n",
      "Epoch 121/500\n",
      "2/2 [==============================] - 2s 950ms/step - loss: 0.7260 - mae: 0.5168 - val_loss: 0.6751 - val_mae: 0.6873\n",
      "Epoch 122/500\n",
      "2/2 [==============================] - 2s 968ms/step - loss: 0.6152 - mae: 0.4680 - val_loss: 0.4790 - val_mae: 0.5242\n",
      "Epoch 123/500\n",
      "2/2 [==============================] - 2s 914ms/step - loss: 0.6760 - mae: 0.5048 - val_loss: 0.5455 - val_mae: 0.6251\n",
      "Epoch 124/500\n",
      "2/2 [==============================] - 2s 943ms/step - loss: 0.5726 - mae: 0.4477 - val_loss: 0.4398 - val_mae: 0.5236\n",
      "Epoch 125/500\n",
      "2/2 [==============================] - 2s 959ms/step - loss: 0.5181 - mae: 0.4349 - val_loss: 0.5015 - val_mae: 0.5888\n",
      "Epoch 126/500\n",
      "2/2 [==============================] - 2s 947ms/step - loss: 0.4709 - mae: 0.4055 - val_loss: 0.5386 - val_mae: 0.5849\n",
      "Epoch 127/500\n",
      "2/2 [==============================] - 2s 927ms/step - loss: 0.6591 - mae: 0.4963 - val_loss: 0.4281 - val_mae: 0.5290\n",
      "Epoch 128/500\n",
      "2/2 [==============================] - 2s 909ms/step - loss: 0.9404 - mae: 0.5988 - val_loss: 0.4897 - val_mae: 0.5884\n",
      "Epoch 129/500\n",
      "2/2 [==============================] - 2s 925ms/step - loss: 0.7743 - mae: 0.5625 - val_loss: 0.4720 - val_mae: 0.5150\n",
      "Epoch 130/500\n",
      "2/2 [==============================] - 2s 919ms/step - loss: 0.7875 - mae: 0.5546 - val_loss: 0.5370 - val_mae: 0.5890\n",
      "Epoch 131/500\n",
      "2/2 [==============================] - 2s 907ms/step - loss: 1.5086 - mae: 0.8191 - val_loss: 0.6153 - val_mae: 0.7062\n",
      "Epoch 132/500\n",
      "2/2 [==============================] - 2s 887ms/step - loss: 1.0527 - mae: 0.6625 - val_loss: 0.5425 - val_mae: 0.5484\n",
      "Epoch 133/500\n",
      "2/2 [==============================] - 2s 912ms/step - loss: 1.4030 - mae: 0.7979 - val_loss: 0.5249 - val_mae: 0.6035\n",
      "Epoch 134/500\n",
      "2/2 [==============================] - 2s 931ms/step - loss: 1.4509 - mae: 0.8129 - val_loss: 0.4488 - val_mae: 0.5845\n",
      "Epoch 135/500\n",
      "2/2 [==============================] - 2s 890ms/step - loss: 0.9538 - mae: 0.6616 - val_loss: 0.5661 - val_mae: 0.6225\n",
      "Epoch 136/500\n",
      "2/2 [==============================] - 2s 911ms/step - loss: 0.7417 - mae: 0.5404 - val_loss: 0.4574 - val_mae: 0.5662\n",
      "Epoch 137/500\n",
      "2/2 [==============================] - 2s 910ms/step - loss: 0.7695 - mae: 0.5432 - val_loss: 0.7753 - val_mae: 0.7252\n",
      "Epoch 138/500\n",
      "2/2 [==============================] - 2s 921ms/step - loss: 0.7959 - mae: 0.5493 - val_loss: 0.5195 - val_mae: 0.5704\n",
      "Epoch 139/500\n",
      "2/2 [==============================] - 2s 917ms/step - loss: 0.5595 - mae: 0.4414 - val_loss: 0.6141 - val_mae: 0.5947\n",
      "Epoch 140/500\n",
      "2/2 [==============================] - 2s 907ms/step - loss: 0.8773 - mae: 0.5810 - val_loss: 1.0168 - val_mae: 0.8256\n",
      "Epoch 141/500\n",
      "2/2 [==============================] - 2s 950ms/step - loss: 0.6520 - mae: 0.5336 - val_loss: 0.6232 - val_mae: 0.6496\n",
      "Epoch 142/500\n",
      "2/2 [==============================] - 2s 904ms/step - loss: 0.5268 - mae: 0.3966 - val_loss: 0.5915 - val_mae: 0.5825\n",
      "Epoch 143/500\n",
      "2/2 [==============================] - 2s 945ms/step - loss: 0.7056 - mae: 0.5142 - val_loss: 1.0684 - val_mae: 0.9404\n",
      "Epoch 144/500\n",
      "2/2 [==============================] - 2s 941ms/step - loss: 1.0141 - mae: 0.6502 - val_loss: 0.5848 - val_mae: 0.5397\n",
      "Epoch 145/500\n",
      "2/2 [==============================] - 2s 930ms/step - loss: 1.3563 - mae: 0.7731 - val_loss: 0.4644 - val_mae: 0.5043\n",
      "Epoch 146/500\n",
      "2/2 [==============================] - 2s 914ms/step - loss: 1.3922 - mae: 0.7521 - val_loss: 0.4527 - val_mae: 0.5246\n",
      "Epoch 147/500\n",
      "2/2 [==============================] - 2s 952ms/step - loss: 0.6418 - mae: 0.4698 - val_loss: 0.4762 - val_mae: 0.5161\n",
      "Epoch 148/500\n",
      "2/2 [==============================] - 2s 929ms/step - loss: 0.5131 - mae: 0.4357 - val_loss: 0.4456 - val_mae: 0.5136\n",
      "Epoch 149/500\n",
      "2/2 [==============================] - 2s 899ms/step - loss: 0.5420 - mae: 0.4390 - val_loss: 0.6628 - val_mae: 0.6631\n",
      "Epoch 150/500\n",
      "2/2 [==============================] - 2s 944ms/step - loss: 0.7043 - mae: 0.4775 - val_loss: 0.6657 - val_mae: 0.7063\n",
      "Epoch 151/500\n",
      "2/2 [==============================] - 2s 930ms/step - loss: 0.7494 - mae: 0.5728 - val_loss: 0.4745 - val_mae: 0.5486\n",
      "Epoch 152/500\n",
      "2/2 [==============================] - 2s 927ms/step - loss: 0.7267 - mae: 0.5525 - val_loss: 0.4903 - val_mae: 0.5499\n",
      "Epoch 153/500\n",
      "2/2 [==============================] - 2s 917ms/step - loss: 0.5433 - mae: 0.4660 - val_loss: 0.4277 - val_mae: 0.4921\n",
      "Epoch 154/500\n",
      "2/2 [==============================] - 2s 947ms/step - loss: 0.6386 - mae: 0.4554 - val_loss: 0.4969 - val_mae: 0.5000\n",
      "Epoch 155/500\n",
      "2/2 [==============================] - 2s 928ms/step - loss: 0.7231 - mae: 0.5391 - val_loss: 0.9465 - val_mae: 0.7984\n",
      "Epoch 156/500\n",
      "2/2 [==============================] - 2s 922ms/step - loss: 0.8569 - mae: 0.6323 - val_loss: 0.5640 - val_mae: 0.5244\n",
      "Epoch 157/500\n",
      "2/2 [==============================] - 2s 932ms/step - loss: 0.7514 - mae: 0.5932 - val_loss: 0.5856 - val_mae: 0.5922\n",
      "Epoch 158/500\n",
      "2/2 [==============================] - 2s 930ms/step - loss: 0.6189 - mae: 0.5066 - val_loss: 0.5224 - val_mae: 0.5452\n",
      "Epoch 159/500\n",
      "2/2 [==============================] - 2s 906ms/step - loss: 0.6528 - mae: 0.5202 - val_loss: 0.4582 - val_mae: 0.5161\n",
      "Epoch 160/500\n",
      "2/2 [==============================] - 2s 918ms/step - loss: 0.7358 - mae: 0.5395 - val_loss: 0.4729 - val_mae: 0.5773\n",
      "Epoch 161/500\n",
      "2/2 [==============================] - 2s 928ms/step - loss: 0.9603 - mae: 0.5985 - val_loss: 0.5110 - val_mae: 0.5561\n",
      "Epoch 162/500\n",
      "2/2 [==============================] - 2s 934ms/step - loss: 0.5515 - mae: 0.4606 - val_loss: 0.5322 - val_mae: 0.5663\n",
      "Epoch 163/500\n",
      "2/2 [==============================] - 2s 941ms/step - loss: 0.4555 - mae: 0.3910 - val_loss: 0.6926 - val_mae: 0.5956\n",
      "Epoch 164/500\n",
      "2/2 [==============================] - 2s 978ms/step - loss: 0.4278 - mae: 0.3751 - val_loss: 0.6858 - val_mae: 0.6156\n",
      "Epoch 165/500\n",
      "2/2 [==============================] - 2s 975ms/step - loss: 0.7635 - mae: 0.5230 - val_loss: 0.5617 - val_mae: 0.6214\n",
      "Epoch 166/500\n",
      "2/2 [==============================] - 2s 947ms/step - loss: 0.4686 - mae: 0.4087 - val_loss: 0.8142 - val_mae: 0.7110\n",
      "Epoch 167/500\n",
      "2/2 [==============================] - 2s 1s/step - loss: 0.7006 - mae: 0.4751 - val_loss: 0.7798 - val_mae: 0.7403\n",
      "Epoch 168/500\n",
      "2/2 [==============================] - 2s 928ms/step - loss: 0.6724 - mae: 0.5074 - val_loss: 0.6554 - val_mae: 0.6761\n",
      "Epoch 169/500\n",
      "2/2 [==============================] - 2s 913ms/step - loss: 0.4596 - mae: 0.4248 - val_loss: 0.5715 - val_mae: 0.5594\n",
      "Epoch 170/500\n",
      "2/2 [==============================] - 2s 943ms/step - loss: 0.4830 - mae: 0.4121 - val_loss: 0.5926 - val_mae: 0.6033\n",
      "Epoch 171/500\n",
      "2/2 [==============================] - 2s 966ms/step - loss: 0.5069 - mae: 0.3871 - val_loss: 0.5674 - val_mae: 0.5902\n",
      "Epoch 172/500\n",
      "2/2 [==============================] - 2s 933ms/step - loss: 0.6875 - mae: 0.5015 - val_loss: 0.5150 - val_mae: 0.5344\n",
      "Epoch 173/500\n",
      "2/2 [==============================] - 2s 886ms/step - loss: 0.4637 - mae: 0.4152 - val_loss: 0.6660 - val_mae: 0.6478\n",
      "Epoch 174/500\n",
      "2/2 [==============================] - 2s 929ms/step - loss: 0.4740 - mae: 0.4017 - val_loss: 0.5717 - val_mae: 0.5577\n",
      "Epoch 175/500\n",
      "2/2 [==============================] - 2s 924ms/step - loss: 0.4022 - mae: 0.3900 - val_loss: 0.6109 - val_mae: 0.5743\n",
      "Epoch 176/500\n",
      "2/2 [==============================] - 2s 966ms/step - loss: 0.3313 - mae: 0.3440 - val_loss: 0.7670 - val_mae: 0.6945\n",
      "Epoch 177/500\n",
      "2/2 [==============================] - 2s 941ms/step - loss: 0.6649 - mae: 0.4625 - val_loss: 0.7067 - val_mae: 0.6779\n",
      "Epoch 178/500\n",
      "2/2 [==============================] - 2s 922ms/step - loss: 1.1050 - mae: 0.7295 - val_loss: 0.4767 - val_mae: 0.4231\n",
      "Epoch 179/500\n",
      "2/2 [==============================] - 2s 920ms/step - loss: 2.1192 - mae: 1.0158 - val_loss: 1.2681 - val_mae: 1.0065\n",
      "Epoch 180/500\n",
      "2/2 [==============================] - 2s 935ms/step - loss: 1.5123 - mae: 0.8919 - val_loss: 0.8056 - val_mae: 0.6881\n",
      "Epoch 181/500\n",
      "2/2 [==============================] - 2s 890ms/step - loss: 0.7051 - mae: 0.5610 - val_loss: 0.6906 - val_mae: 0.6739\n",
      "Epoch 182/500\n",
      "2/2 [==============================] - 2s 980ms/step - loss: 0.6410 - mae: 0.4936 - val_loss: 0.6530 - val_mae: 0.6529\n",
      "Epoch 183/500\n",
      "2/2 [==============================] - 2s 923ms/step - loss: 0.6049 - mae: 0.4695 - val_loss: 0.6258 - val_mae: 0.6211\n",
      "Epoch 184/500\n",
      "2/2 [==============================] - 2s 927ms/step - loss: 0.4245 - mae: 0.4052 - val_loss: 0.6278 - val_mae: 0.5753\n",
      "Epoch 185/500\n",
      "2/2 [==============================] - 2s 920ms/step - loss: 0.4681 - mae: 0.4290 - val_loss: 0.6523 - val_mae: 0.6309\n",
      "Epoch 186/500\n",
      "2/2 [==============================] - 2s 908ms/step - loss: 0.4092 - mae: 0.3934 - val_loss: 0.6722 - val_mae: 0.6198\n",
      "Epoch 187/500\n",
      "2/2 [==============================] - 2s 900ms/step - loss: 0.6440 - mae: 0.5103 - val_loss: 0.5556 - val_mae: 0.5784\n",
      "Epoch 188/500\n",
      "2/2 [==============================] - 2s 940ms/step - loss: 0.3709 - mae: 0.3838 - val_loss: 0.8180 - val_mae: 0.6878\n",
      "Epoch 189/500\n",
      "2/2 [==============================] - 2s 945ms/step - loss: 0.5862 - mae: 0.4624 - val_loss: 0.6023 - val_mae: 0.6128\n",
      "Epoch 190/500\n",
      "2/2 [==============================] - 2s 953ms/step - loss: 0.5781 - mae: 0.4177 - val_loss: 0.4209 - val_mae: 0.4075\n",
      "Epoch 191/500\n",
      "2/2 [==============================] - 2s 898ms/step - loss: 0.5399 - mae: 0.4738 - val_loss: 0.3816 - val_mae: 0.4168\n",
      "Epoch 192/500\n",
      "2/2 [==============================] - 2s 924ms/step - loss: 0.5235 - mae: 0.4600 - val_loss: 0.5225 - val_mae: 0.5005\n",
      "Epoch 193/500\n",
      "2/2 [==============================] - 2s 908ms/step - loss: 0.3333 - mae: 0.3445 - val_loss: 0.7127 - val_mae: 0.7006\n",
      "Epoch 194/500\n",
      "2/2 [==============================] - 2s 880ms/step - loss: 0.3260 - mae: 0.3644 - val_loss: 0.6113 - val_mae: 0.5878\n",
      "Epoch 195/500\n",
      "2/2 [==============================] - 2s 917ms/step - loss: 0.3268 - mae: 0.3764 - val_loss: 0.4267 - val_mae: 0.4621\n",
      "Epoch 196/500\n",
      "2/2 [==============================] - 2s 929ms/step - loss: 0.4730 - mae: 0.4447 - val_loss: 0.5611 - val_mae: 0.5415\n",
      "Epoch 197/500\n",
      "2/2 [==============================] - 2s 958ms/step - loss: 0.7760 - mae: 0.5220 - val_loss: 0.5340 - val_mae: 0.5267\n",
      "Epoch 198/500\n",
      "2/2 [==============================] - 2s 979ms/step - loss: 0.9640 - mae: 0.5995 - val_loss: 0.4599 - val_mae: 0.4862\n",
      "Epoch 199/500\n",
      "2/2 [==============================] - 2s 923ms/step - loss: 1.0018 - mae: 0.6724 - val_loss: 0.7291 - val_mae: 0.6435\n",
      "Epoch 200/500\n",
      "2/2 [==============================] - 2s 898ms/step - loss: 0.8082 - mae: 0.5659 - val_loss: 0.9742 - val_mae: 0.8332\n",
      "Epoch 201/500\n",
      "2/2 [==============================] - 2s 929ms/step - loss: 1.2454 - mae: 0.7431 - val_loss: 2.8912 - val_mae: 1.3365\n",
      "Epoch 202/500\n",
      "2/2 [==============================] - 2s 942ms/step - loss: 2.2878 - mae: 1.1603 - val_loss: 0.1851 - val_mae: 0.3184\n",
      "Epoch 203/500\n",
      "2/2 [==============================] - 2s 895ms/step - loss: 1.4376 - mae: 0.8798 - val_loss: 0.2628 - val_mae: 0.3304\n",
      "Epoch 204/500\n",
      "2/2 [==============================] - 2s 947ms/step - loss: 1.3300 - mae: 0.7467 - val_loss: 0.3010 - val_mae: 0.4094\n",
      "Epoch 205/500\n",
      "2/2 [==============================] - 2s 954ms/step - loss: 0.9239 - mae: 0.5983 - val_loss: 0.4410 - val_mae: 0.5119\n",
      "Epoch 206/500\n",
      "2/2 [==============================] - 2s 941ms/step - loss: 0.5894 - mae: 0.5072 - val_loss: 0.5448 - val_mae: 0.5883\n",
      "Epoch 207/500\n",
      "2/2 [==============================] - 2s 921ms/step - loss: 0.6311 - mae: 0.5218 - val_loss: 0.5244 - val_mae: 0.5769\n",
      "Epoch 208/500\n",
      "2/2 [==============================] - 2s 943ms/step - loss: 0.5384 - mae: 0.4605 - val_loss: 0.5568 - val_mae: 0.6155\n",
      "Epoch 209/500\n",
      "2/2 [==============================] - 2s 942ms/step - loss: 0.5221 - mae: 0.3896 - val_loss: 0.5571 - val_mae: 0.6081\n",
      "Epoch 210/500\n",
      "2/2 [==============================] - 2s 1s/step - loss: 0.4622 - mae: 0.3759 - val_loss: 0.4470 - val_mae: 0.5108\n",
      "Epoch 211/500\n",
      "2/2 [==============================] - 2s 936ms/step - loss: 0.3906 - mae: 0.3673 - val_loss: 0.5718 - val_mae: 0.5964\n",
      "Epoch 212/500\n",
      "2/2 [==============================] - 2s 912ms/step - loss: 0.3239 - mae: 0.3331 - val_loss: 0.6100 - val_mae: 0.6290\n",
      "Epoch 213/500\n",
      "2/2 [==============================] - 2s 943ms/step - loss: 0.6051 - mae: 0.4460 - val_loss: 0.9060 - val_mae: 0.8358\n",
      "Epoch 214/500\n",
      "2/2 [==============================] - 2s 906ms/step - loss: 1.6175 - mae: 0.8835 - val_loss: 0.5518 - val_mae: 0.5580\n",
      "Epoch 215/500\n",
      "2/2 [==============================] - 2s 882ms/step - loss: 1.3921 - mae: 0.7887 - val_loss: 0.6107 - val_mae: 0.5987\n",
      "Epoch 216/500\n",
      "2/2 [==============================] - 2s 901ms/step - loss: 0.7564 - mae: 0.5699 - val_loss: 0.6507 - val_mae: 0.6083\n",
      "Epoch 217/500\n",
      "2/2 [==============================] - 2s 914ms/step - loss: 0.5914 - mae: 0.4424 - val_loss: 0.4944 - val_mae: 0.5311\n",
      "Epoch 218/500\n",
      "2/2 [==============================] - 2s 924ms/step - loss: 0.4140 - mae: 0.3594 - val_loss: 0.4087 - val_mae: 0.4438\n",
      "Epoch 219/500\n",
      "2/2 [==============================] - 2s 920ms/step - loss: 0.7999 - mae: 0.5070 - val_loss: 0.2762 - val_mae: 0.4040\n",
      "Epoch 220/500\n",
      "2/2 [==============================] - 2s 941ms/step - loss: 0.7652 - mae: 0.5556 - val_loss: 0.4426 - val_mae: 0.5253\n",
      "Epoch 221/500\n",
      "2/2 [==============================] - 2s 951ms/step - loss: 0.4100 - mae: 0.3961 - val_loss: 0.5212 - val_mae: 0.5816\n",
      "Epoch 222/500\n",
      "2/2 [==============================] - 2s 931ms/step - loss: 0.3690 - mae: 0.3624 - val_loss: 0.4888 - val_mae: 0.5251\n",
      "Epoch 223/500\n",
      "2/2 [==============================] - 2s 936ms/step - loss: 0.5101 - mae: 0.4060 - val_loss: 0.5534 - val_mae: 0.5595\n",
      "Epoch 224/500\n",
      "2/2 [==============================] - 2s 946ms/step - loss: 0.6565 - mae: 0.4616 - val_loss: 0.5183 - val_mae: 0.5913\n",
      "Epoch 225/500\n",
      "2/2 [==============================] - 2s 948ms/step - loss: 0.5948 - mae: 0.4531 - val_loss: 0.4070 - val_mae: 0.4392\n",
      "Epoch 226/500\n",
      "2/2 [==============================] - 2s 897ms/step - loss: 0.3941 - mae: 0.4070 - val_loss: 0.5005 - val_mae: 0.5057\n",
      "Epoch 227/500\n",
      "2/2 [==============================] - 2s 923ms/step - loss: 0.5136 - mae: 0.4436 - val_loss: 0.3340 - val_mae: 0.3855\n",
      "Epoch 228/500\n",
      "2/2 [==============================] - 2s 936ms/step - loss: 0.8775 - mae: 0.6170 - val_loss: 0.5828 - val_mae: 0.5669\n",
      "Epoch 229/500\n",
      "2/2 [==============================] - 2s 918ms/step - loss: 1.0079 - mae: 0.6417 - val_loss: 0.5300 - val_mae: 0.4448\n",
      "Epoch 230/500\n",
      "2/2 [==============================] - 2s 918ms/step - loss: 0.8431 - mae: 0.5885 - val_loss: 0.5702 - val_mae: 0.5705\n",
      "Epoch 231/500\n",
      "2/2 [==============================] - 2s 936ms/step - loss: 0.6011 - mae: 0.4996 - val_loss: 0.4537 - val_mae: 0.5008\n",
      "Epoch 232/500\n",
      "2/2 [==============================] - 2s 925ms/step - loss: 0.8106 - mae: 0.5638 - val_loss: 0.4793 - val_mae: 0.4433\n",
      "Epoch 233/500\n",
      "2/2 [==============================] - 2s 923ms/step - loss: 0.4368 - mae: 0.4034 - val_loss: 0.5340 - val_mae: 0.5228\n",
      "Epoch 234/500\n",
      "2/2 [==============================] - 2s 927ms/step - loss: 0.3147 - mae: 0.3352 - val_loss: 0.5309 - val_mae: 0.5485\n",
      "Epoch 235/500\n",
      "2/2 [==============================] - 2s 910ms/step - loss: 0.3054 - mae: 0.3230 - val_loss: 0.7541 - val_mae: 0.6473\n",
      "Epoch 236/500\n",
      "2/2 [==============================] - 2s 896ms/step - loss: 0.4485 - mae: 0.3944 - val_loss: 0.6123 - val_mae: 0.6013\n",
      "Epoch 237/500\n",
      "2/2 [==============================] - 2s 926ms/step - loss: 0.5575 - mae: 0.4734 - val_loss: 0.6323 - val_mae: 0.5832\n",
      "Epoch 238/500\n",
      "2/2 [==============================] - 2s 909ms/step - loss: 0.6083 - mae: 0.4804 - val_loss: 0.7333 - val_mae: 0.6161\n",
      "Epoch 239/500\n",
      "2/2 [==============================] - 2s 929ms/step - loss: 0.5000 - mae: 0.4415 - val_loss: 0.6947 - val_mae: 0.6241\n",
      "Epoch 240/500\n",
      "2/2 [==============================] - 2s 920ms/step - loss: 0.6811 - mae: 0.5337 - val_loss: 0.3971 - val_mae: 0.4547\n",
      "Epoch 241/500\n",
      "2/2 [==============================] - 2s 934ms/step - loss: 0.4144 - mae: 0.4006 - val_loss: 0.5330 - val_mae: 0.5234\n",
      "Epoch 242/500\n",
      "2/2 [==============================] - 2s 926ms/step - loss: 0.5059 - mae: 0.4413 - val_loss: 0.4202 - val_mae: 0.4554\n",
      "Epoch 243/500\n",
      "2/2 [==============================] - 2s 906ms/step - loss: 0.4454 - mae: 0.4443 - val_loss: 0.6485 - val_mae: 0.6075\n",
      "Epoch 244/500\n",
      "2/2 [==============================] - 2s 925ms/step - loss: 0.2794 - mae: 0.3397 - val_loss: 0.6796 - val_mae: 0.6082\n",
      "Epoch 245/500\n",
      "2/2 [==============================] - 2s 960ms/step - loss: 0.2850 - mae: 0.3143 - val_loss: 0.5437 - val_mae: 0.5366\n",
      "Epoch 246/500\n",
      "2/2 [==============================] - 2s 965ms/step - loss: 0.4004 - mae: 0.3521 - val_loss: 0.5106 - val_mae: 0.5280\n",
      "Epoch 247/500\n",
      "2/2 [==============================] - 2s 947ms/step - loss: 0.3030 - mae: 0.3103 - val_loss: 0.6521 - val_mae: 0.5706\n",
      "Epoch 248/500\n",
      "2/2 [==============================] - 2s 903ms/step - loss: 0.2163 - mae: 0.2933 - val_loss: 0.6914 - val_mae: 0.6504\n",
      "Epoch 249/500\n",
      "2/2 [==============================] - 2s 902ms/step - loss: 0.2880 - mae: 0.3209 - val_loss: 0.6378 - val_mae: 0.5833\n",
      "Epoch 250/500\n",
      "2/2 [==============================] - 2s 931ms/step - loss: 0.3782 - mae: 0.3758 - val_loss: 0.6690 - val_mae: 0.6340\n",
      "Epoch 251/500\n",
      "2/2 [==============================] - 2s 899ms/step - loss: 0.3611 - mae: 0.4172 - val_loss: 0.6031 - val_mae: 0.5728\n",
      "Epoch 252/500\n",
      "2/2 [==============================] - 2s 893ms/step - loss: 0.3314 - mae: 0.3867 - val_loss: 0.7708 - val_mae: 0.6586\n",
      "Epoch 253/500\n",
      "2/2 [==============================] - 2s 918ms/step - loss: 1.3574 - mae: 0.7536 - val_loss: 0.3441 - val_mae: 0.4203\n",
      "Epoch 254/500\n",
      "2/2 [==============================] - 2s 936ms/step - loss: 1.3177 - mae: 0.7527 - val_loss: 0.7588 - val_mae: 0.6909\n",
      "Epoch 255/500\n",
      "2/2 [==============================] - 2s 899ms/step - loss: 0.4792 - mae: 0.4525 - val_loss: 0.5665 - val_mae: 0.5421\n",
      "Epoch 256/500\n",
      "2/2 [==============================] - 2s 919ms/step - loss: 0.2832 - mae: 0.3350 - val_loss: 0.5551 - val_mae: 0.5495\n",
      "Epoch 257/500\n",
      "2/2 [==============================] - 2s 1s/step - loss: 0.2039 - mae: 0.2922 - val_loss: 0.6543 - val_mae: 0.6015\n",
      "Epoch 258/500\n",
      "2/2 [==============================] - 2s 959ms/step - loss: 0.3254 - mae: 0.3641 - val_loss: 0.8587 - val_mae: 0.6706\n",
      "Epoch 259/500\n",
      "2/2 [==============================] - 2s 918ms/step - loss: 0.7471 - mae: 0.5749 - val_loss: 0.8552 - val_mae: 0.6862\n",
      "Epoch 260/500\n",
      "2/2 [==============================] - 2s 983ms/step - loss: 0.3655 - mae: 0.4147 - val_loss: 0.7198 - val_mae: 0.6364\n",
      "Epoch 261/500\n",
      "2/2 [==============================] - 2s 926ms/step - loss: 0.3432 - mae: 0.3691 - val_loss: 0.6954 - val_mae: 0.6382\n",
      "Epoch 262/500\n",
      "2/2 [==============================] - 2s 943ms/step - loss: 0.5407 - mae: 0.4482 - val_loss: 0.5984 - val_mae: 0.5647\n",
      "Epoch 263/500\n",
      "2/2 [==============================] - 2s 932ms/step - loss: 0.4856 - mae: 0.4353 - val_loss: 0.6653 - val_mae: 0.6346\n",
      "Epoch 264/500\n",
      "2/2 [==============================] - 2s 947ms/step - loss: 0.5757 - mae: 0.4613 - val_loss: 0.8310 - val_mae: 0.6443\n",
      "Epoch 265/500\n",
      "2/2 [==============================] - 2s 916ms/step - loss: 0.9536 - mae: 0.5735 - val_loss: 0.5397 - val_mae: 0.5796\n",
      "Epoch 266/500\n",
      "2/2 [==============================] - 2s 964ms/step - loss: 0.4110 - mae: 0.3993 - val_loss: 0.5664 - val_mae: 0.5457\n",
      "Epoch 267/500\n",
      "2/2 [==============================] - 2s 974ms/step - loss: 0.5079 - mae: 0.4053 - val_loss: 0.5090 - val_mae: 0.4938\n",
      "Epoch 268/500\n",
      "2/2 [==============================] - 2s 948ms/step - loss: 0.3236 - mae: 0.3554 - val_loss: 0.5247 - val_mae: 0.5107\n",
      "Epoch 269/500\n",
      "2/2 [==============================] - 2s 902ms/step - loss: 0.2146 - mae: 0.2963 - val_loss: 0.5489 - val_mae: 0.5054\n",
      "Epoch 270/500\n",
      "2/2 [==============================] - 2s 951ms/step - loss: 0.1704 - mae: 0.2503 - val_loss: 0.6121 - val_mae: 0.5622\n",
      "Epoch 271/500\n",
      "2/2 [==============================] - 2s 926ms/step - loss: 0.1803 - mae: 0.2711 - val_loss: 0.4757 - val_mae: 0.4748\n",
      "Epoch 272/500\n",
      "2/2 [==============================] - 2s 889ms/step - loss: 0.1612 - mae: 0.2402 - val_loss: 0.4880 - val_mae: 0.4749\n",
      "Epoch 273/500\n",
      "2/2 [==============================] - 2s 913ms/step - loss: 0.1518 - mae: 0.2523 - val_loss: 0.4552 - val_mae: 0.4650\n",
      "Epoch 274/500\n",
      "2/2 [==============================] - 2s 905ms/step - loss: 0.1644 - mae: 0.2760 - val_loss: 0.4726 - val_mae: 0.4822\n",
      "Epoch 275/500\n",
      "2/2 [==============================] - 2s 943ms/step - loss: 0.3145 - mae: 0.3680 - val_loss: 0.4725 - val_mae: 0.5258\n",
      "Epoch 276/500\n",
      "2/2 [==============================] - 2s 957ms/step - loss: 0.5680 - mae: 0.4213 - val_loss: 0.4496 - val_mae: 0.4778\n",
      "Epoch 277/500\n",
      "2/2 [==============================] - 2s 993ms/step - loss: 0.7243 - mae: 0.5116 - val_loss: 0.5207 - val_mae: 0.4904\n",
      "Epoch 278/500\n",
      "2/2 [==============================] - 2s 944ms/step - loss: 0.9000 - mae: 0.5561 - val_loss: 0.6893 - val_mae: 0.5938\n",
      "Epoch 279/500\n",
      "2/2 [==============================] - 2s 902ms/step - loss: 1.0216 - mae: 0.6714 - val_loss: 0.4006 - val_mae: 0.4290\n",
      "Epoch 280/500\n",
      "2/2 [==============================] - 2s 914ms/step - loss: 0.7805 - mae: 0.5336 - val_loss: 0.5945 - val_mae: 0.6018\n",
      "Epoch 281/500\n",
      "2/2 [==============================] - 2s 918ms/step - loss: 0.5411 - mae: 0.4631 - val_loss: 0.7095 - val_mae: 0.6599\n",
      "Epoch 282/500\n",
      "2/2 [==============================] - 2s 915ms/step - loss: 0.6596 - mae: 0.5355 - val_loss: 0.7159 - val_mae: 0.6492\n",
      "Epoch 283/500\n",
      "2/2 [==============================] - 2s 900ms/step - loss: 0.7578 - mae: 0.5751 - val_loss: 0.5175 - val_mae: 0.5469\n",
      "Epoch 284/500\n",
      "2/2 [==============================] - 2s 901ms/step - loss: 0.3161 - mae: 0.3558 - val_loss: 0.6967 - val_mae: 0.6149\n",
      "Epoch 285/500\n",
      "2/2 [==============================] - 2s 925ms/step - loss: 0.4044 - mae: 0.3738 - val_loss: 0.6478 - val_mae: 0.6443\n",
      "Epoch 286/500\n",
      "2/2 [==============================] - 2s 906ms/step - loss: 0.8318 - mae: 0.5541 - val_loss: 0.5857 - val_mae: 0.4986\n",
      "Epoch 287/500\n",
      "2/2 [==============================] - 2s 968ms/step - loss: 0.5212 - mae: 0.4634 - val_loss: 0.6374 - val_mae: 0.5586\n",
      "Epoch 288/500\n",
      "2/2 [==============================] - 2s 920ms/step - loss: 0.1990 - mae: 0.2910 - val_loss: 0.6636 - val_mae: 0.5912\n",
      "Epoch 289/500\n",
      "2/2 [==============================] - 2s 911ms/step - loss: 0.1588 - mae: 0.2479 - val_loss: 0.7089 - val_mae: 0.6235\n",
      "Epoch 290/500\n",
      "2/2 [==============================] - 2s 969ms/step - loss: 0.2387 - mae: 0.2980 - val_loss: 0.6427 - val_mae: 0.5823\n",
      "Epoch 291/500\n",
      "2/2 [==============================] - 2s 981ms/step - loss: 0.1815 - mae: 0.2903 - val_loss: 0.4681 - val_mae: 0.4758\n",
      "Epoch 292/500\n",
      "2/2 [==============================] - 2s 944ms/step - loss: 0.2059 - mae: 0.3117 - val_loss: 0.4716 - val_mae: 0.4944\n",
      "Epoch 293/500\n",
      "2/2 [==============================] - 2s 999ms/step - loss: 0.2524 - mae: 0.3625 - val_loss: 0.5226 - val_mae: 0.4982\n",
      "Epoch 294/500\n",
      "2/2 [==============================] - 2s 944ms/step - loss: 0.2102 - mae: 0.2976 - val_loss: 0.5115 - val_mae: 0.5229\n",
      "Epoch 295/500\n",
      "2/2 [==============================] - 2s 908ms/step - loss: 0.5387 - mae: 0.4220 - val_loss: 0.6213 - val_mae: 0.6740\n",
      "Epoch 296/500\n",
      "2/2 [==============================] - 2s 932ms/step - loss: 1.1030 - mae: 0.7087 - val_loss: 0.3402 - val_mae: 0.3100\n",
      "Epoch 297/500\n",
      "2/2 [==============================] - 2s 964ms/step - loss: 1.2280 - mae: 0.7437 - val_loss: 1.0166 - val_mae: 0.8626\n",
      "Epoch 298/500\n",
      "2/2 [==============================] - 2s 900ms/step - loss: 0.5916 - mae: 0.5012 - val_loss: 0.6642 - val_mae: 0.5947\n",
      "Epoch 299/500\n",
      "2/2 [==============================] - 2s 904ms/step - loss: 0.2643 - mae: 0.3077 - val_loss: 0.7074 - val_mae: 0.5986\n",
      "Epoch 300/500\n",
      "2/2 [==============================] - 2s 905ms/step - loss: 0.2360 - mae: 0.3224 - val_loss: 0.6599 - val_mae: 0.5663\n",
      "Epoch 301/500\n",
      "2/2 [==============================] - 2s 935ms/step - loss: 0.2063 - mae: 0.2929 - val_loss: 0.5456 - val_mae: 0.5012\n",
      "Epoch 302/500\n",
      "2/2 [==============================] - 2s 907ms/step - loss: 0.2684 - mae: 0.3334 - val_loss: 0.4210 - val_mae: 0.4387\n",
      "Epoch 303/500\n",
      "2/2 [==============================] - 2s 939ms/step - loss: 0.3807 - mae: 0.4074 - val_loss: 0.4015 - val_mae: 0.4694\n",
      "Epoch 304/500\n",
      "2/2 [==============================] - 2s 1s/step - loss: 0.6945 - mae: 0.5649 - val_loss: 0.6086 - val_mae: 0.5107\n",
      "Epoch 305/500\n",
      "2/2 [==============================] - 2s 951ms/step - loss: 0.3442 - mae: 0.4168 - val_loss: 0.5051 - val_mae: 0.5149\n",
      "Epoch 306/500\n",
      "2/2 [==============================] - 2s 948ms/step - loss: 0.3010 - mae: 0.3567 - val_loss: 0.5373 - val_mae: 0.5147\n",
      "Epoch 307/500\n",
      "2/2 [==============================] - 2s 970ms/step - loss: 0.2239 - mae: 0.3237 - val_loss: 0.5464 - val_mae: 0.5462\n",
      "Epoch 308/500\n",
      "2/2 [==============================] - 2s 945ms/step - loss: 0.1621 - mae: 0.2584 - val_loss: 0.5521 - val_mae: 0.5317\n",
      "Epoch 309/500\n",
      "2/2 [==============================] - 2s 932ms/step - loss: 0.1469 - mae: 0.2454 - val_loss: 0.5377 - val_mae: 0.4933\n",
      "Epoch 310/500\n",
      "2/2 [==============================] - 2s 909ms/step - loss: 0.2668 - mae: 0.3066 - val_loss: 0.5459 - val_mae: 0.5650\n",
      "Epoch 311/500\n",
      "2/2 [==============================] - 2s 959ms/step - loss: 0.4894 - mae: 0.4283 - val_loss: 0.9144 - val_mae: 0.6971\n",
      "Epoch 312/500\n",
      "2/2 [==============================] - 2s 903ms/step - loss: 0.7344 - mae: 0.5744 - val_loss: 0.5555 - val_mae: 0.5669\n",
      "Epoch 313/500\n",
      "2/2 [==============================] - 2s 892ms/step - loss: 1.3197 - mae: 0.6952 - val_loss: 0.4315 - val_mae: 0.4840\n",
      "Epoch 314/500\n",
      "2/2 [==============================] - 2s 964ms/step - loss: 2.4818 - mae: 1.0824 - val_loss: 0.8654 - val_mae: 0.6988\n",
      "Epoch 315/500\n",
      "2/2 [==============================] - 2s 964ms/step - loss: 2.0673 - mae: 1.0869 - val_loss: 0.2429 - val_mae: 0.2970\n",
      "Epoch 316/500\n",
      "2/2 [==============================] - 2s 908ms/step - loss: 1.8633 - mae: 0.9750 - val_loss: 0.4831 - val_mae: 0.4986\n",
      "Epoch 317/500\n",
      "2/2 [==============================] - 2s 922ms/step - loss: 0.6631 - mae: 0.4983 - val_loss: 0.4938 - val_mae: 0.5283\n",
      "Epoch 318/500\n",
      "2/2 [==============================] - 2s 948ms/step - loss: 0.4810 - mae: 0.3932 - val_loss: 0.4280 - val_mae: 0.4510\n",
      "Epoch 319/500\n",
      "2/2 [==============================] - 2s 940ms/step - loss: 0.3311 - mae: 0.3827 - val_loss: 0.4603 - val_mae: 0.4773\n",
      "Epoch 320/500\n",
      "2/2 [==============================] - 2s 928ms/step - loss: 0.2117 - mae: 0.3095 - val_loss: 0.5455 - val_mae: 0.5425\n",
      "Epoch 321/500\n",
      "2/2 [==============================] - 2s 916ms/step - loss: 0.1731 - mae: 0.2767 - val_loss: 0.4866 - val_mae: 0.4895\n",
      "Epoch 322/500\n",
      "2/2 [==============================] - 2s 915ms/step - loss: 0.2078 - mae: 0.3002 - val_loss: 0.4118 - val_mae: 0.4707\n",
      "Epoch 323/500\n",
      "2/2 [==============================] - 2s 916ms/step - loss: 0.2565 - mae: 0.2945 - val_loss: 0.3944 - val_mae: 0.4315\n",
      "Epoch 324/500\n",
      "2/2 [==============================] - 2s 932ms/step - loss: 0.2937 - mae: 0.3288 - val_loss: 0.5243 - val_mae: 0.5577\n",
      "Epoch 325/500\n",
      "2/2 [==============================] - 2s 914ms/step - loss: 0.1736 - mae: 0.2795 - val_loss: 0.4475 - val_mae: 0.4820\n",
      "Epoch 326/500\n",
      "2/2 [==============================] - 2s 898ms/step - loss: 0.2578 - mae: 0.2784 - val_loss: 0.4020 - val_mae: 0.4485\n",
      "Epoch 327/500\n",
      "2/2 [==============================] - 2s 923ms/step - loss: 0.2283 - mae: 0.2857 - val_loss: 0.4747 - val_mae: 0.5133\n",
      "Epoch 328/500\n",
      "2/2 [==============================] - 2s 954ms/step - loss: 0.3036 - mae: 0.2972 - val_loss: 0.4176 - val_mae: 0.4599\n",
      "Epoch 329/500\n",
      "2/2 [==============================] - 2s 928ms/step - loss: 0.2146 - mae: 0.2992 - val_loss: 0.4266 - val_mae: 0.4854\n",
      "Epoch 330/500\n",
      "2/2 [==============================] - 2s 916ms/step - loss: 0.2685 - mae: 0.2795 - val_loss: 0.4829 - val_mae: 0.5105\n",
      "Epoch 331/500\n",
      "2/2 [==============================] - 2s 943ms/step - loss: 0.1566 - mae: 0.2454 - val_loss: 0.5666 - val_mae: 0.5336\n",
      "Epoch 332/500\n",
      "2/2 [==============================] - 2s 894ms/step - loss: 0.2986 - mae: 0.3430 - val_loss: 0.4538 - val_mae: 0.4818\n",
      "Epoch 333/500\n",
      "2/2 [==============================] - 2s 910ms/step - loss: 0.2486 - mae: 0.3287 - val_loss: 0.5177 - val_mae: 0.5258\n",
      "Epoch 334/500\n",
      "2/2 [==============================] - 2s 910ms/step - loss: 0.2447 - mae: 0.3326 - val_loss: 0.4717 - val_mae: 0.5315\n",
      "Epoch 335/500\n",
      "2/2 [==============================] - 2s 920ms/step - loss: 0.5224 - mae: 0.4122 - val_loss: 0.4163 - val_mae: 0.4834\n",
      "Epoch 336/500\n",
      "2/2 [==============================] - 2s 969ms/step - loss: 0.3416 - mae: 0.3549 - val_loss: 0.3687 - val_mae: 0.4307\n",
      "Epoch 337/500\n",
      "2/2 [==============================] - 2s 962ms/step - loss: 0.3586 - mae: 0.3474 - val_loss: 0.4293 - val_mae: 0.4719\n",
      "Epoch 338/500\n",
      "2/2 [==============================] - 2s 908ms/step - loss: 0.2307 - mae: 0.2797 - val_loss: 0.4870 - val_mae: 0.5246\n",
      "Epoch 339/500\n",
      "2/2 [==============================] - 2s 949ms/step - loss: 0.5201 - mae: 0.3729 - val_loss: 0.5432 - val_mae: 0.5122\n",
      "Epoch 340/500\n",
      "2/2 [==============================] - 2s 929ms/step - loss: 0.3188 - mae: 0.3942 - val_loss: 0.4149 - val_mae: 0.4397\n",
      "Epoch 341/500\n",
      "2/2 [==============================] - 2s 916ms/step - loss: 0.2921 - mae: 0.3454 - val_loss: 0.4533 - val_mae: 0.5352\n",
      "Epoch 342/500\n",
      "2/2 [==============================] - 2s 905ms/step - loss: 0.5512 - mae: 0.4965 - val_loss: 0.5704 - val_mae: 0.5774\n",
      "Epoch 343/500\n",
      "2/2 [==============================] - 2s 944ms/step - loss: 0.1964 - mae: 0.3002 - val_loss: 0.5572 - val_mae: 0.5768\n",
      "Epoch 344/500\n",
      "2/2 [==============================] - 2s 950ms/step - loss: 0.1211 - mae: 0.2470 - val_loss: 0.5194 - val_mae: 0.5321\n",
      "Epoch 345/500\n",
      "2/2 [==============================] - 2s 919ms/step - loss: 0.1070 - mae: 0.2072 - val_loss: 0.5181 - val_mae: 0.5169\n",
      "Epoch 346/500\n",
      "2/2 [==============================] - 2s 945ms/step - loss: 0.1502 - mae: 0.2315 - val_loss: 0.5593 - val_mae: 0.5516\n",
      "Epoch 347/500\n",
      "2/2 [==============================] - 2s 925ms/step - loss: 0.1925 - mae: 0.2660 - val_loss: 0.5113 - val_mae: 0.4994\n",
      "Epoch 348/500\n",
      "2/2 [==============================] - 2s 903ms/step - loss: 0.1506 - mae: 0.2308 - val_loss: 0.4744 - val_mae: 0.5030\n",
      "Epoch 349/500\n",
      "2/2 [==============================] - 2s 917ms/step - loss: 0.0829 - mae: 0.1850 - val_loss: 0.4722 - val_mae: 0.5007\n",
      "Epoch 350/500\n",
      "2/2 [==============================] - 2s 949ms/step - loss: 0.0726 - mae: 0.1760 - val_loss: 0.4936 - val_mae: 0.5073\n",
      "Epoch 351/500\n",
      "2/2 [==============================] - 2s 953ms/step - loss: 0.0757 - mae: 0.1784 - val_loss: 0.5070 - val_mae: 0.5139\n",
      "Epoch 352/500\n",
      "2/2 [==============================] - 2s 937ms/step - loss: 0.1125 - mae: 0.2189 - val_loss: 0.4326 - val_mae: 0.4640\n",
      "Epoch 353/500\n",
      "2/2 [==============================] - 2s 958ms/step - loss: 0.0861 - mae: 0.1979 - val_loss: 0.4560 - val_mae: 0.4716\n",
      "Epoch 354/500\n",
      "2/2 [==============================] - 2s 957ms/step - loss: 0.1639 - mae: 0.2392 - val_loss: 0.4213 - val_mae: 0.4447\n",
      "Epoch 355/500\n",
      "2/2 [==============================] - 2s 912ms/step - loss: 0.5219 - mae: 0.3811 - val_loss: 0.2663 - val_mae: 0.3180\n",
      "Epoch 356/500\n",
      "2/2 [==============================] - 2s 911ms/step - loss: 0.3903 - mae: 0.3971 - val_loss: 0.4329 - val_mae: 0.4523\n",
      "Epoch 357/500\n",
      "2/2 [==============================] - 2s 942ms/step - loss: 0.1230 - mae: 0.2378 - val_loss: 0.4597 - val_mae: 0.4897\n",
      "Epoch 358/500\n",
      "2/2 [==============================] - 2s 909ms/step - loss: 0.1886 - mae: 0.2446 - val_loss: 0.5169 - val_mae: 0.5451\n",
      "Epoch 359/500\n",
      "2/2 [==============================] - 2s 961ms/step - loss: 0.1450 - mae: 0.2740 - val_loss: 0.5990 - val_mae: 0.5859\n",
      "Epoch 360/500\n",
      "2/2 [==============================] - 2s 989ms/step - loss: 0.1110 - mae: 0.2486 - val_loss: 0.5887 - val_mae: 0.5852\n",
      "Epoch 361/500\n",
      "2/2 [==============================] - 2s 967ms/step - loss: 0.0813 - mae: 0.2071 - val_loss: 0.4928 - val_mae: 0.5077\n",
      "Epoch 362/500\n",
      "2/2 [==============================] - 2s 890ms/step - loss: 0.0759 - mae: 0.1769 - val_loss: 0.5027 - val_mae: 0.5141\n",
      "Epoch 363/500\n",
      "2/2 [==============================] - 2s 935ms/step - loss: 0.2234 - mae: 0.2540 - val_loss: 0.4825 - val_mae: 0.4853\n",
      "Epoch 364/500\n",
      "2/2 [==============================] - 2s 936ms/step - loss: 0.1966 - mae: 0.2628 - val_loss: 0.5470 - val_mae: 0.5587\n",
      "Epoch 365/500\n",
      "2/2 [==============================] - 2s 1s/step - loss: 0.1830 - mae: 0.2742 - val_loss: 0.5138 - val_mae: 0.5023\n",
      "Epoch 366/500\n",
      "2/2 [==============================] - 2s 1s/step - loss: 0.5222 - mae: 0.4451 - val_loss: 0.4504 - val_mae: 0.4934\n",
      "Epoch 367/500\n",
      "2/2 [==============================] - 2s 940ms/step - loss: 0.4670 - mae: 0.3942 - val_loss: 0.3800 - val_mae: 0.5029\n",
      "Epoch 368/500\n",
      "2/2 [==============================] - 2s 946ms/step - loss: 0.5332 - mae: 0.4764 - val_loss: 0.5496 - val_mae: 0.5516\n",
      "Epoch 369/500\n",
      "2/2 [==============================] - 2s 890ms/step - loss: 0.2484 - mae: 0.3282 - val_loss: 0.4739 - val_mae: 0.4831\n",
      "Epoch 370/500\n",
      "2/2 [==============================] - 2s 911ms/step - loss: 0.1504 - mae: 0.2681 - val_loss: 0.5230 - val_mae: 0.5265\n",
      "Epoch 371/500\n",
      "2/2 [==============================] - 2s 982ms/step - loss: 0.1572 - mae: 0.2544 - val_loss: 0.4546 - val_mae: 0.4539\n",
      "Epoch 372/500\n",
      "2/2 [==============================] - 2s 907ms/step - loss: 0.2683 - mae: 0.3082 - val_loss: 0.3699 - val_mae: 0.4221\n",
      "Epoch 373/500\n",
      "2/2 [==============================] - 2s 963ms/step - loss: 0.4204 - mae: 0.4004 - val_loss: 0.5179 - val_mae: 0.5050\n",
      "Epoch 374/500\n",
      "2/2 [==============================] - 2s 948ms/step - loss: 0.2519 - mae: 0.2931 - val_loss: 0.4715 - val_mae: 0.4870\n",
      "Epoch 375/500\n",
      "2/2 [==============================] - 2s 939ms/step - loss: 0.1492 - mae: 0.2405 - val_loss: 0.5092 - val_mae: 0.4809\n",
      "Epoch 376/500\n",
      "2/2 [==============================] - 2s 995ms/step - loss: 0.1030 - mae: 0.2020 - val_loss: 0.5487 - val_mae: 0.5500\n",
      "Epoch 377/500\n",
      "2/2 [==============================] - 2s 949ms/step - loss: 0.0718 - mae: 0.1863 - val_loss: 0.5402 - val_mae: 0.5296\n",
      "Epoch 378/500\n",
      "2/2 [==============================] - 2s 984ms/step - loss: 0.0758 - mae: 0.1845 - val_loss: 0.5566 - val_mae: 0.5432\n",
      "Epoch 379/500\n",
      "2/2 [==============================] - 2s 915ms/step - loss: 0.0822 - mae: 0.1928 - val_loss: 0.5163 - val_mae: 0.5145\n",
      "Epoch 380/500\n",
      "2/2 [==============================] - 2s 951ms/step - loss: 0.0900 - mae: 0.2072 - val_loss: 0.4677 - val_mae: 0.4650\n",
      "Epoch 381/500\n",
      "2/2 [==============================] - 2s 935ms/step - loss: 0.1264 - mae: 0.2114 - val_loss: 0.5320 - val_mae: 0.5254\n",
      "Epoch 382/500\n",
      "2/2 [==============================] - 2s 924ms/step - loss: 0.1194 - mae: 0.2237 - val_loss: 0.4869 - val_mae: 0.4930\n",
      "Epoch 383/500\n",
      "2/2 [==============================] - 2s 943ms/step - loss: 0.0713 - mae: 0.1850 - val_loss: 0.4785 - val_mae: 0.4807\n",
      "Epoch 384/500\n",
      "2/2 [==============================] - 2s 926ms/step - loss: 0.0801 - mae: 0.1915 - val_loss: 0.4412 - val_mae: 0.4357\n",
      "Epoch 385/500\n",
      "2/2 [==============================] - 2s 956ms/step - loss: 0.1449 - mae: 0.2260 - val_loss: 0.4976 - val_mae: 0.4924\n",
      "Epoch 386/500\n",
      "2/2 [==============================] - 2s 921ms/step - loss: 0.0804 - mae: 0.2057 - val_loss: 0.4322 - val_mae: 0.4219\n",
      "Epoch 387/500\n",
      "2/2 [==============================] - 2s 922ms/step - loss: 0.0957 - mae: 0.2250 - val_loss: 0.4519 - val_mae: 0.4552\n",
      "Epoch 388/500\n",
      "2/2 [==============================] - 2s 946ms/step - loss: 0.1505 - mae: 0.2226 - val_loss: 0.4567 - val_mae: 0.4545\n",
      "Epoch 389/500\n",
      "2/2 [==============================] - 2s 945ms/step - loss: 0.1138 - mae: 0.2122 - val_loss: 0.4739 - val_mae: 0.4739\n",
      "Epoch 390/500\n",
      "2/2 [==============================] - 2s 949ms/step - loss: 0.0893 - mae: 0.2137 - val_loss: 0.5120 - val_mae: 0.4912\n",
      "Epoch 391/500\n",
      "2/2 [==============================] - 2s 944ms/step - loss: 0.1660 - mae: 0.2576 - val_loss: 0.6069 - val_mae: 0.5543\n",
      "Epoch 392/500\n",
      "2/2 [==============================] - 2s 917ms/step - loss: 0.7556 - mae: 0.5231 - val_loss: 0.4317 - val_mae: 0.4616\n",
      "Epoch 393/500\n",
      "2/2 [==============================] - 2s 960ms/step - loss: 0.3494 - mae: 0.3854 - val_loss: 0.3703 - val_mae: 0.3688\n",
      "Epoch 394/500\n",
      "2/2 [==============================] - 2s 936ms/step - loss: 0.4073 - mae: 0.4225 - val_loss: 1.2967 - val_mae: 0.8514\n",
      "Epoch 395/500\n",
      "2/2 [==============================] - 2s 903ms/step - loss: 0.8497 - mae: 0.6499 - val_loss: 0.6999 - val_mae: 0.6331\n",
      "Epoch 396/500\n",
      "2/2 [==============================] - 2s 930ms/step - loss: 0.3292 - mae: 0.3822 - val_loss: 0.6538 - val_mae: 0.6449\n",
      "Epoch 397/500\n",
      "2/2 [==============================] - 2s 943ms/step - loss: 0.4624 - mae: 0.3971 - val_loss: 0.4801 - val_mae: 0.4819\n",
      "Epoch 398/500\n",
      "2/2 [==============================] - 2s 984ms/step - loss: 0.9990 - mae: 0.6160 - val_loss: 1.1549 - val_mae: 0.9235\n",
      "Epoch 399/500\n",
      "2/2 [==============================] - 2s 901ms/step - loss: 1.1752 - mae: 0.7086 - val_loss: 0.7011 - val_mae: 0.6429\n",
      "Epoch 400/500\n",
      "2/2 [==============================] - 2s 906ms/step - loss: 1.1877 - mae: 0.6907 - val_loss: 1.6799 - val_mae: 1.0432\n",
      "Epoch 401/500\n",
      "2/2 [==============================] - 2s 942ms/step - loss: 2.2523 - mae: 1.0215 - val_loss: 0.4796 - val_mae: 0.5720\n",
      "Epoch 402/500\n",
      "2/2 [==============================] - 2s 925ms/step - loss: 0.9476 - mae: 0.6770 - val_loss: 0.7590 - val_mae: 0.6797\n",
      "Epoch 403/500\n",
      "2/2 [==============================] - 2s 908ms/step - loss: 0.3533 - mae: 0.3577 - val_loss: 0.6988 - val_mae: 0.6292\n",
      "Epoch 404/500\n",
      "2/2 [==============================] - 2s 904ms/step - loss: 0.1736 - mae: 0.2832 - val_loss: 0.6990 - val_mae: 0.6220\n",
      "Epoch 405/500\n",
      "2/2 [==============================] - 2s 938ms/step - loss: 0.1922 - mae: 0.2682 - val_loss: 0.6222 - val_mae: 0.5866\n",
      "Epoch 406/500\n",
      "2/2 [==============================] - 2s 927ms/step - loss: 0.1542 - mae: 0.2673 - val_loss: 0.7632 - val_mae: 0.6591\n",
      "Epoch 407/500\n",
      "2/2 [==============================] - 2s 932ms/step - loss: 0.1226 - mae: 0.2189 - val_loss: 0.6434 - val_mae: 0.5991\n",
      "Epoch 408/500\n",
      "2/2 [==============================] - 2s 969ms/step - loss: 0.4729 - mae: 0.3342 - val_loss: 0.4950 - val_mae: 0.4739\n",
      "Epoch 409/500\n",
      "2/2 [==============================] - 2s 930ms/step - loss: 0.4644 - mae: 0.4151 - val_loss: 0.4084 - val_mae: 0.4679\n",
      "Epoch 410/500\n",
      "2/2 [==============================] - 2s 905ms/step - loss: 0.6088 - mae: 0.4537 - val_loss: 0.6100 - val_mae: 0.5800\n",
      "Epoch 411/500\n",
      "2/2 [==============================] - 2s 953ms/step - loss: 0.4882 - mae: 0.4137 - val_loss: 0.6667 - val_mae: 0.6121\n",
      "Epoch 412/500\n",
      "2/2 [==============================] - 2s 912ms/step - loss: 0.5175 - mae: 0.4627 - val_loss: 0.7676 - val_mae: 0.7205\n",
      "Epoch 413/500\n",
      "2/2 [==============================] - 2s 930ms/step - loss: 0.5867 - mae: 0.5176 - val_loss: 0.7726 - val_mae: 0.7011\n",
      "Epoch 414/500\n",
      "2/2 [==============================] - 2s 932ms/step - loss: 0.2806 - mae: 0.3385 - val_loss: 0.8011 - val_mae: 0.7000\n",
      "Epoch 415/500\n",
      "2/2 [==============================] - 2s 1s/step - loss: 0.1595 - mae: 0.2511 - val_loss: 0.6639 - val_mae: 0.6063\n",
      "Epoch 416/500\n",
      "2/2 [==============================] - 2s 946ms/step - loss: 0.1219 - mae: 0.2164 - val_loss: 0.6822 - val_mae: 0.6154\n",
      "Epoch 417/500\n",
      "2/2 [==============================] - 2s 928ms/step - loss: 0.1194 - mae: 0.2124 - val_loss: 0.6764 - val_mae: 0.6185\n",
      "Epoch 418/500\n",
      "2/2 [==============================] - 2s 897ms/step - loss: 0.1179 - mae: 0.2101 - val_loss: 0.5621 - val_mae: 0.5404\n",
      "Epoch 419/500\n",
      "2/2 [==============================] - 2s 889ms/step - loss: 0.1244 - mae: 0.2045 - val_loss: 0.5731 - val_mae: 0.5587\n",
      "Epoch 420/500\n",
      "2/2 [==============================] - 2s 912ms/step - loss: 0.0707 - mae: 0.1776 - val_loss: 0.5802 - val_mae: 0.5484\n",
      "Epoch 421/500\n",
      "2/2 [==============================] - 2s 912ms/step - loss: 0.0616 - mae: 0.1702 - val_loss: 0.6363 - val_mae: 0.5861\n",
      "Epoch 422/500\n",
      "2/2 [==============================] - 2s 899ms/step - loss: 0.0866 - mae: 0.1801 - val_loss: 0.5616 - val_mae: 0.5216\n",
      "Epoch 423/500\n",
      "2/2 [==============================] - 2s 936ms/step - loss: 0.0824 - mae: 0.1752 - val_loss: 0.5994 - val_mae: 0.5645\n",
      "Epoch 424/500\n",
      "2/2 [==============================] - 2s 939ms/step - loss: 0.0594 - mae: 0.1578 - val_loss: 0.5924 - val_mae: 0.5551\n",
      "Epoch 425/500\n",
      "2/2 [==============================] - 2s 895ms/step - loss: 0.0746 - mae: 0.1800 - val_loss: 0.6253 - val_mae: 0.5821\n",
      "Epoch 426/500\n",
      "2/2 [==============================] - 2s 931ms/step - loss: 0.0925 - mae: 0.2020 - val_loss: 0.5884 - val_mae: 0.5577\n",
      "Epoch 427/500\n",
      "2/2 [==============================] - 2s 938ms/step - loss: 0.0786 - mae: 0.1932 - val_loss: 0.6175 - val_mae: 0.5711\n",
      "Epoch 428/500\n",
      "2/2 [==============================] - 2s 919ms/step - loss: 0.0564 - mae: 0.1591 - val_loss: 0.5720 - val_mae: 0.5369\n",
      "Epoch 429/500\n",
      "2/2 [==============================] - 2s 917ms/step - loss: 0.0453 - mae: 0.1450 - val_loss: 0.5797 - val_mae: 0.5270\n",
      "Epoch 430/500\n",
      "2/2 [==============================] - 2s 966ms/step - loss: 0.0459 - mae: 0.1508 - val_loss: 0.5659 - val_mae: 0.5166\n",
      "Epoch 431/500\n",
      "2/2 [==============================] - 2s 922ms/step - loss: 0.0411 - mae: 0.1411 - val_loss: 0.5390 - val_mae: 0.5149\n",
      "Epoch 432/500\n",
      "2/2 [==============================] - 2s 933ms/step - loss: 0.0413 - mae: 0.1424 - val_loss: 0.5043 - val_mae: 0.5007\n",
      "Epoch 433/500\n",
      "2/2 [==============================] - 2s 927ms/step - loss: 0.0451 - mae: 0.1515 - val_loss: 0.5050 - val_mae: 0.4909\n",
      "Epoch 434/500\n",
      "2/2 [==============================] - 2s 957ms/step - loss: 0.0537 - mae: 0.1602 - val_loss: 0.5195 - val_mae: 0.5222\n",
      "Epoch 435/500\n",
      "2/2 [==============================] - 2s 974ms/step - loss: 0.0768 - mae: 0.1883 - val_loss: 0.4494 - val_mae: 0.4611\n",
      "Epoch 436/500\n",
      "2/2 [==============================] - 2s 947ms/step - loss: 0.1594 - mae: 0.2757 - val_loss: 0.3630 - val_mae: 0.4261\n",
      "Epoch 437/500\n",
      "2/2 [==============================] - 2s 900ms/step - loss: 0.4638 - mae: 0.4366 - val_loss: 0.4790 - val_mae: 0.4853\n",
      "Epoch 438/500\n",
      "2/2 [==============================] - 2s 915ms/step - loss: 0.7227 - mae: 0.5074 - val_loss: 0.5704 - val_mae: 0.5882\n",
      "Epoch 439/500\n",
      "2/2 [==============================] - 2s 895ms/step - loss: 0.6455 - mae: 0.5493 - val_loss: 0.9091 - val_mae: 0.7538\n",
      "Epoch 440/500\n",
      "2/2 [==============================] - 2s 939ms/step - loss: 0.5726 - mae: 0.5304 - val_loss: 0.4248 - val_mae: 0.4936\n",
      "Epoch 441/500\n",
      "2/2 [==============================] - 2s 953ms/step - loss: 0.3439 - mae: 0.4012 - val_loss: 0.5579 - val_mae: 0.5218\n",
      "Epoch 442/500\n",
      "2/2 [==============================] - 2s 1s/step - loss: 0.1605 - mae: 0.2649 - val_loss: 0.5862 - val_mae: 0.5678\n",
      "Epoch 443/500\n",
      "2/2 [==============================] - 2s 910ms/step - loss: 0.1099 - mae: 0.2153 - val_loss: 0.5629 - val_mae: 0.5573\n",
      "Epoch 444/500\n",
      "2/2 [==============================] - 2s 965ms/step - loss: 0.1153 - mae: 0.2173 - val_loss: 0.5128 - val_mae: 0.5338\n",
      "Epoch 445/500\n",
      "2/2 [==============================] - 2s 987ms/step - loss: 0.1560 - mae: 0.2346 - val_loss: 0.4835 - val_mae: 0.4684\n",
      "Epoch 446/500\n",
      "2/2 [==============================] - 2s 1s/step - loss: 0.1597 - mae: 0.2621 - val_loss: 0.4824 - val_mae: 0.4888\n",
      "Epoch 447/500\n",
      "2/2 [==============================] - 2s 976ms/step - loss: 0.1171 - mae: 0.2334 - val_loss: 0.4479 - val_mae: 0.4576\n",
      "Epoch 448/500\n",
      "2/2 [==============================] - 2s 995ms/step - loss: 0.1000 - mae: 0.2181 - val_loss: 0.4737 - val_mae: 0.4882\n",
      "Epoch 449/500\n",
      "2/2 [==============================] - 2s 1s/step - loss: 0.2042 - mae: 0.2890 - val_loss: 0.3794 - val_mae: 0.4268\n",
      "Epoch 450/500\n",
      "2/2 [==============================] - 2s 993ms/step - loss: 0.3550 - mae: 0.3551 - val_loss: 0.5343 - val_mae: 0.5230\n",
      "Epoch 451/500\n",
      "2/2 [==============================] - 2s 909ms/step - loss: 0.0775 - mae: 0.1997 - val_loss: 0.6164 - val_mae: 0.5771\n",
      "Epoch 452/500\n",
      "2/2 [==============================] - 2s 958ms/step - loss: 0.0655 - mae: 0.1819 - val_loss: 0.5764 - val_mae: 0.5421\n",
      "Epoch 453/500\n",
      "2/2 [==============================] - 2s 900ms/step - loss: 0.0675 - mae: 0.1763 - val_loss: 0.5727 - val_mae: 0.5386\n",
      "Epoch 454/500\n",
      "2/2 [==============================] - 2s 894ms/step - loss: 0.1091 - mae: 0.2141 - val_loss: 0.6232 - val_mae: 0.5813\n",
      "Epoch 455/500\n",
      "2/2 [==============================] - 2s 949ms/step - loss: 0.2552 - mae: 0.3088 - val_loss: 0.4748 - val_mae: 0.5199\n",
      "Epoch 456/500\n",
      "2/2 [==============================] - 2s 978ms/step - loss: 0.1393 - mae: 0.2249 - val_loss: 0.5186 - val_mae: 0.5344\n",
      "Epoch 457/500\n",
      "2/2 [==============================] - 2s 908ms/step - loss: 0.2294 - mae: 0.2575 - val_loss: 0.4287 - val_mae: 0.4288\n",
      "Epoch 458/500\n",
      "2/2 [==============================] - 2s 926ms/step - loss: 0.1882 - mae: 0.2777 - val_loss: 0.4787 - val_mae: 0.5271\n",
      "Epoch 459/500\n",
      "2/2 [==============================] - 2s 916ms/step - loss: 0.0866 - mae: 0.2005 - val_loss: 0.5080 - val_mae: 0.5084\n",
      "Epoch 460/500\n",
      "2/2 [==============================] - 2s 918ms/step - loss: 0.0628 - mae: 0.1663 - val_loss: 0.4759 - val_mae: 0.4773\n",
      "Epoch 461/500\n",
      "2/2 [==============================] - 2s 941ms/step - loss: 0.0651 - mae: 0.1573 - val_loss: 0.5213 - val_mae: 0.5125\n",
      "Epoch 462/500\n",
      "2/2 [==============================] - 2s 944ms/step - loss: 0.0511 - mae: 0.1606 - val_loss: 0.5173 - val_mae: 0.5114\n",
      "Epoch 463/500\n",
      "2/2 [==============================] - 2s 931ms/step - loss: 0.0394 - mae: 0.1376 - val_loss: 0.5197 - val_mae: 0.5089\n",
      "Epoch 464/500\n",
      "2/2 [==============================] - 2s 973ms/step - loss: 0.0594 - mae: 0.1525 - val_loss: 0.5191 - val_mae: 0.5051\n",
      "Epoch 465/500\n",
      "2/2 [==============================] - 2s 930ms/step - loss: 0.0495 - mae: 0.1491 - val_loss: 0.4683 - val_mae: 0.4866\n",
      "Epoch 466/500\n",
      "2/2 [==============================] - 2s 935ms/step - loss: 0.0409 - mae: 0.1376 - val_loss: 0.5608 - val_mae: 0.5229\n",
      "Epoch 467/500\n",
      "2/2 [==============================] - 2s 987ms/step - loss: 0.0350 - mae: 0.1353 - val_loss: 0.5222 - val_mae: 0.5044\n",
      "Epoch 468/500\n",
      "2/2 [==============================] - 2s 939ms/step - loss: 0.0297 - mae: 0.1241 - val_loss: 0.5384 - val_mae: 0.5160\n",
      "Epoch 469/500\n",
      "2/2 [==============================] - 2s 946ms/step - loss: 0.0284 - mae: 0.1214 - val_loss: 0.5694 - val_mae: 0.5347\n",
      "Epoch 470/500\n",
      "2/2 [==============================] - 2s 989ms/step - loss: 0.0357 - mae: 0.1360 - val_loss: 0.5841 - val_mae: 0.5398\n",
      "Epoch 471/500\n",
      "2/2 [==============================] - 2s 955ms/step - loss: 0.0403 - mae: 0.1429 - val_loss: 0.5984 - val_mae: 0.5422\n",
      "Epoch 472/500\n",
      "2/2 [==============================] - 2s 900ms/step - loss: 0.0611 - mae: 0.1721 - val_loss: 0.5569 - val_mae: 0.5280\n",
      "Epoch 473/500\n",
      "2/2 [==============================] - 2s 956ms/step - loss: 0.0956 - mae: 0.2052 - val_loss: 0.5867 - val_mae: 0.5542\n",
      "Epoch 474/500\n",
      "2/2 [==============================] - 2s 975ms/step - loss: 0.0852 - mae: 0.2114 - val_loss: 0.5690 - val_mae: 0.5387\n",
      "Epoch 475/500\n",
      "2/2 [==============================] - 2s 952ms/step - loss: 0.0918 - mae: 0.2091 - val_loss: 0.6005 - val_mae: 0.5662\n",
      "Epoch 476/500\n",
      "2/2 [==============================] - 2s 959ms/step - loss: 0.2912 - mae: 0.3271 - val_loss: 0.5856 - val_mae: 0.5747\n",
      "Epoch 477/500\n",
      "2/2 [==============================] - 2s 936ms/step - loss: 0.0669 - mae: 0.1929 - val_loss: 0.5255 - val_mae: 0.5413\n",
      "Epoch 478/500\n",
      "2/2 [==============================] - 2s 904ms/step - loss: 0.0642 - mae: 0.1796 - val_loss: 0.5358 - val_mae: 0.5243\n",
      "Epoch 479/500\n",
      "2/2 [==============================] - 2s 927ms/step - loss: 0.0500 - mae: 0.1551 - val_loss: 0.4902 - val_mae: 0.4973\n",
      "Epoch 480/500\n",
      "2/2 [==============================] - 2s 975ms/step - loss: 0.0391 - mae: 0.1353 - val_loss: 0.4878 - val_mae: 0.4928\n",
      "Epoch 481/500\n",
      "2/2 [==============================] - 2s 924ms/step - loss: 0.0524 - mae: 0.1544 - val_loss: 0.4544 - val_mae: 0.4622\n",
      "Epoch 482/500\n",
      "2/2 [==============================] - 2s 952ms/step - loss: 0.0499 - mae: 0.1488 - val_loss: 0.4756 - val_mae: 0.4819\n",
      "Epoch 483/500\n",
      "2/2 [==============================] - 2s 1s/step - loss: 0.0285 - mae: 0.1240 - val_loss: 0.4771 - val_mae: 0.4804\n",
      "Epoch 484/500\n",
      "2/2 [==============================] - 2s 986ms/step - loss: 0.0270 - mae: 0.1195 - val_loss: 0.4982 - val_mae: 0.4918\n",
      "Epoch 485/500\n",
      "2/2 [==============================] - 2s 1s/step - loss: 0.0296 - mae: 0.1303 - val_loss: 0.4968 - val_mae: 0.4943\n",
      "Epoch 486/500\n",
      "2/2 [==============================] - 2s 939ms/step - loss: 0.0347 - mae: 0.1376 - val_loss: 0.4785 - val_mae: 0.4860\n",
      "Epoch 487/500\n",
      "2/2 [==============================] - 2s 1s/step - loss: 0.0371 - mae: 0.1361 - val_loss: 0.4358 - val_mae: 0.4516\n",
      "Epoch 488/500\n",
      "2/2 [==============================] - 2s 1s/step - loss: 0.0269 - mae: 0.1212 - val_loss: 0.4480 - val_mae: 0.4591\n",
      "Epoch 489/500\n",
      "2/2 [==============================] - 2s 979ms/step - loss: 0.0275 - mae: 0.1241 - val_loss: 0.4875 - val_mae: 0.4841\n",
      "Epoch 490/500\n",
      "2/2 [==============================] - 2s 969ms/step - loss: 0.0292 - mae: 0.1330 - val_loss: 0.4674 - val_mae: 0.4688\n",
      "Epoch 491/500\n",
      "2/2 [==============================] - 2s 988ms/step - loss: 0.0279 - mae: 0.1261 - val_loss: 0.4689 - val_mae: 0.4688\n",
      "Epoch 492/500\n",
      "2/2 [==============================] - 2s 961ms/step - loss: 0.0235 - mae: 0.1156 - val_loss: 0.4619 - val_mae: 0.4691\n",
      "Epoch 493/500\n",
      "2/2 [==============================] - 2s 937ms/step - loss: 0.0210 - mae: 0.1058 - val_loss: 0.4724 - val_mae: 0.4812\n",
      "Epoch 494/500\n",
      "2/2 [==============================] - 2s 971ms/step - loss: 0.0195 - mae: 0.1016 - val_loss: 0.4812 - val_mae: 0.4876\n",
      "Epoch 495/500\n",
      "2/2 [==============================] - 2s 915ms/step - loss: 0.0211 - mae: 0.1078 - val_loss: 0.4403 - val_mae: 0.4519\n",
      "Epoch 496/500\n",
      "2/2 [==============================] - 2s 923ms/step - loss: 0.0270 - mae: 0.1214 - val_loss: 0.4536 - val_mae: 0.4620\n",
      "Epoch 497/500\n",
      "2/2 [==============================] - 2s 959ms/step - loss: 0.0345 - mae: 0.1377 - val_loss: 0.4397 - val_mae: 0.4544\n",
      "Epoch 498/500\n",
      "2/2 [==============================] - 2s 950ms/step - loss: 0.0449 - mae: 0.1535 - val_loss: 0.4244 - val_mae: 0.4423\n",
      "Epoch 499/500\n",
      "2/2 [==============================] - 2s 923ms/step - loss: 0.0388 - mae: 0.1480 - val_loss: 0.4312 - val_mae: 0.4444\n",
      "Epoch 500/500\n",
      "2/2 [==============================] - 2s 966ms/step - loss: 0.0332 - mae: 0.1363 - val_loss: 0.4457 - val_mae: 0.4564\n"
     ]
    }
   ],
   "source": [
    "\n",
    "class DisplayProgress(keras.callbacks.Callback):\n",
    "    def on_epoch_end(self, epoch, logs=None):\n",
    "        print('Epoch {}/{}'.format(epoch + 1, eps), end='\\r')\n",
    "\n",
    "def loss_fn(y_true, y_pred):\n",
    "    return tf.reduce_mean(tf.square(y_true - y_pred))\n",
    "\n",
    "\n",
    "opt = optimizers.SGD(learning_rate=0.01)\n",
    "gpicl.compile(optimizer=opt, loss=loss_fn, metrics=['mae'])\n",
    "\n",
    "eps = 500\n",
    "\n",
    "log_dir = \"logs/fit/\" + datetime.datetime.now().strftime(\"%Y%m%d-%H%M%S\")\n",
    "tensorboard_callback = keras.callbacks.TensorBoard(log_dir=log_dir)\n",
    "\n",
    "#with tf.device('/cpu:0'):\n",
    "gpicl.fit(train_s, train_l, epochs=eps, batch_size=200, validation_data=(valid_s, valid_l), callbacks=[tensorboard_callback, DisplayProgress()], \n",
    "#verbose=0\n",
    ")\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 194,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "loss & MAE [0.0754275694489479, 0.20824034512043]\n",
      "0.05643733\n",
      "1.9868584\n",
      "-0.059952535\n",
      "0.3786287\n",
      "1/1 [==============================] - 0s 166ms/step\n",
      "[-3.186183]\n",
      "1/1 [==============================] - 0s 178ms/step\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1/1 [==============================] - 0s 326ms/step\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "test_s, test_l = datagen(400, sampl)\n",
    "\n",
    "# get loss on test set\n",
    "print('loss & MAE', gpicl.evaluate(test_s, test_l, verbose=0))\n",
    "\n",
    "print(tf.math.reduce_mean(test_l).numpy())\n",
    "print(tf.math.reduce_std(test_l).numpy())\n",
    "print(tf.math.reduce_mean(valid_l).numpy())\n",
    "print(tf.math.reduce_std(valid_l).numpy())\n",
    "\n",
    "num = np.random.randint(0, 99) \n",
    "# plot data points, no lines\n",
    "plt.plot(test_s[num, :-1, 0], test_s[num, :-1, -1], 'o')\n",
    "# plot the last 10 points as red points\n",
    "\n",
    "print(gpicl.predict(test_s[num:num+1, :, :])[0])\n",
    "plt.plot(test_s[num, -1:, 0], gpicl.predict(test_s[num:num+1, :, :])[0], 'ro')\n",
    "plt.show()\n",
    "\n",
    "plt.plot(valid_s[num, :-1, 0], valid_s[num, :-1, -1], 'o')\n",
    "# plot the last 10 points as red points\n",
    "plt.plot(valid_s[num, -1:, 0], gpicl.predict(valid_s[num:num+1, :, :])[0], 'ro')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 189,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "\n",
       "      <iframe id=\"tensorboard-frame-c161c4ff386a0ca\" width=\"100%\" height=\"800\" frameborder=\"0\">\n",
       "      </iframe>\n",
       "      <script>\n",
       "        (function() {\n",
       "          const frame = document.getElementById(\"tensorboard-frame-c161c4ff386a0ca\");\n",
       "          const url = new URL(\"http://localhost\");\n",
       "          const port = 6006;\n",
       "          if (port) {\n",
       "            url.port = port;\n",
       "          }\n",
       "          frame.src = url;\n",
       "        })();\n",
       "      </script>\n",
       "    "
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%tensorboard --logdir logs/fit"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "base",
   "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": "3.10.9"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}