4way1shotMAMLClassification.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "import tensorflow.keras as keras\n",
    "import tensorflow.keras.backend as keras_backend\n",
    "from tensorflow.keras.utils import to_categorical\n",
    "\n",
    "import random\n",
    "import sys\n",
    "import time\n",
    "\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "\n",
    "import random\n",
    "import matplotlib.pyplot as plt\n",
    "import itertools\n",
    "import glob\n",
    "\n",
    "tf.keras.backend.set_floatx('float64')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2.0.0\n"
     ]
    }
   ],
   "source": [
    "print(tf.__version__)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0 P0-W0\n",
      "(861, 504)\n",
      "(861,)\n",
      "1 P1-W1\n",
      "(743, 504)\n",
      "(743,)\n",
      "2 P9-W7\n",
      "(819, 504)\n",
      "(819,)\n",
      "3 P14-W0\n",
      "(824, 504)\n",
      "(824,)\n"
     ]
    }
   ],
   "source": [
    "path = r'/home/giang/Lab/Code/Testsets' # use your path\n",
    "classes = {0:'P0-W0', 1:'P1-W1', 2:'P9-W7', 3:'P14-W0'}\n",
    "\n",
    "df, x_np, y_np = {}, {}, {}\n",
    "\n",
    "for i, cls in classes.items():\n",
    "  print(i, cls)\n",
    "  all_files = glob.glob(path + \"/\" + cls + \"/*.csv\")\n",
    "  df[i] = pd.concat((pd.read_csv(f, header=None) for f in all_files))\n",
    "  x_np[i] = np.transpose(np.array(df[i]))\n",
    "  print(x_np[i].shape)\n",
    "   \n",
    "  y_np[i] = np.ones(len(x_np[i])) * i # Target label\n",
    "  print(y_np[i].shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(x_np[0][0][0:55])\n",
    "#plt.plot(x_np[1])\n",
    "#plt.plot(x_np[2])\n",
    "#plt.plot(x_np[3])\n",
    "plt.xlabel(\"Subcarrier\")\n",
    "plt.ylabel(\"Amplitude\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "743\n"
     ]
    }
   ],
   "source": [
    "num_tasks = min(len(x_np[i]) for i in range(len(classes))) \n",
    "print(num_tasks)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(4, 504)\n",
      "(4, 1)\n"
     ]
    }
   ],
   "source": [
    "task_data = {}\n",
    "task_label = {}\n",
    "for i in range(num_tasks):\n",
    "    task_data[i] = np.stack([x_np[k][i] for k in range(len(classes))],axis=0)\n",
    "    task_label[i] = np.stack([y_np[k][i] for k in range(len(classes))],axis=0)\n",
    "    task_label[i] = task_label[i].reshape(-1,1)\n",
    "print(task_data[0].shape)\n",
    "print(task_label[0].shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(4, 504)\n",
      "(4, 4)\n"
     ]
    }
   ],
   "source": [
    "data_label = {}\n",
    "for i in range(num_tasks):\n",
    "    data_label[i] = np.concatenate((task_data[i],task_label[i]),axis=1)\n",
    "    np.random.shuffle(data_label[i])\n",
    "    task_data[i],task_label[i] = np.split(data_label[i],[-1],axis=1)\n",
    "    task_label[i] = to_categorical(task_label[i])\n",
    "print(task_data[0].shape)\n",
    "print(task_label[0].shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "class MLPModel(keras.Model):\n",
    "    def __init__(self):\n",
    "        super().__init__()\n",
    "        self.hidden1 = keras.layers.Dense(50, input_shape=(504,))\n",
    "        self.hidden2 = keras.layers.Dense(30)\n",
    "        self.hidden3 = keras.layers.Dense(30)\n",
    "        self.out = keras.layers.Dense(4)\n",
    "        \n",
    "    def forward(self, x):\n",
    "        x = keras.activations.relu(self.hidden1(x))\n",
    "        x = keras.activations.relu(self.hidden2(x))\n",
    "        x = keras.activations.relu(self.hidden3(x))\n",
    "        x = keras.activations.softmax(self.out(x))\n",
    "        return x"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "def loss_function(y,pred_y):\n",
    "    return keras_backend.mean(keras.losses.categorical_crossentropy(y,pred_y))\n",
    "\n",
    "def np_to_tensor(list_of_numpy_objs):\n",
    "    return (tf.convert_to_tensor(obj) for obj in list_of_numpy_objs)\n",
    "    \n",
    "\n",
    "def compute_loss(model, x, y, loss_fn=loss_function):\n",
    "    logits = model.forward(x)\n",
    "    mse = loss_fn(y, logits)\n",
    "    return mse, logits\n",
    "\n",
    "\n",
    "def compute_gradients(model, x, y, loss_fn=loss_function):\n",
    "    with tf.GradientTape() as tape:\n",
    "        loss, logits = compute_loss(model, x, y, loss_fn)\n",
    "    return tape.gradient(loss, model.trainable_variables), loss, logits\n",
    "\n",
    "\n",
    "def apply_gradients(optimizer, gradients, variables):\n",
    "    optimizer.apply_gradients(zip(gradients, variables))\n",
    "\n",
    "    \n",
    "def train_batch(x, y, model, optimizer):\n",
    "    tensor_x, tensor_y = np_to_tensor((x, y))\n",
    "    gradients, loss, logits = compute_gradients(model, tensor_x, tensor_y)\n",
    "    apply_gradients(optimizer, gradients, model.trainable_variables)\n",
    "    return logits, loss\n",
    "\n",
    "def accuracy_function(y, pred_y):\n",
    "    return keras.metrics.categorical_accuracy(y, pred_y)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "def train_model(model,data,label, epochs=1, lr=0.001, log_steps=1):\n",
    "    model = model\n",
    "    optimizer = keras.optimizers.Adam(learning_rate=lr)\n",
    "\n",
    "    losses = []\n",
    "    accuracies = []\n",
    "    for epoch in range(epochs):\n",
    "        epoch_totalloss = 0\n",
    "        epoch_totalaccuracy = 0\n",
    "        start = time.time()\n",
    "        for i in range(len(data)):\n",
    "            task_totalloss = 0\n",
    "            task_totalaccuracy = 0\n",
    "            for k in range(len(classes)):              \n",
    "                x = data[i][k,:].reshape(1,-1)\n",
    "                y = label[i][k]\n",
    "                logits, task_currloss = train_batch(x, y, model, optimizer)\n",
    "                task_totalloss += task_currloss\n",
    "                task_loss = task_totalloss / (k + 1.0)\n",
    "                task_curraccuracy = accuracy_function(y, logits)\n",
    "                task_totalaccuracy += task_curraccuracy\n",
    "                task_accuracy = task_totalaccuracy / (k + 1.0)\n",
    "            epoch_totalaccuracy += task_accuracy\n",
    "            epoch_totalloss += task_loss\n",
    "        epoch_loss = epoch_totalloss/(i+1)\n",
    "        epoch_accuracy = epoch_totalaccuracy/(i+1)\n",
    "        losses.append(epoch_loss)\n",
    "        accuracies.append(epoch_accuracy)\n",
    "            \n",
    "        if epoch % log_steps == 0:\n",
    "            print('Epoch {}: Loss = {}, Train accuracy = {}, Time to run {} epoch = {:.2f} seconds'.format(\n",
    "                    epoch+1, epoch_loss, epoch_accuracy, log_steps, time.time() - start))\n",
    "            start = time.time()\n",
    "    plt.figure()\n",
    "    plt.plot(losses)\n",
    "    plt.title('Loss Vs Epochs')\n",
    "    plt.xlabel('Epochs')\n",
    "    plt.ylabel('Loss')\n",
    "    plt.show()\n",
    "    plt.figure()\n",
    "    plt.plot(accuracies)\n",
    "    plt.title('Train Accuracy Vs Epochs')\n",
    "    plt.xlabel('Epochs')\n",
    "    plt.ylabel('Accuracy')\n",
    "    plt.show()\n",
    "    return model,losses, accuracies"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1: loss = 0.10215504647800708, train accuracy = [0.96971736], Time to run 1 epoch = 120.54 seconds\n",
      "Epoch 2: loss = 0.03660325317623795, train accuracy = [0.98654105], Time to run 1 epoch = 120.56 seconds\n",
      "Epoch 3: loss = 0.02792938144215788, train accuracy = [0.98990579], Time to run 1 epoch = 119.82 seconds\n",
      "Epoch 4: loss = 0.020823918619923828, train accuracy = [0.99427995], Time to run 1 epoch = 120.14 seconds\n",
      "Epoch 5: loss = 0.02109745033348402, train accuracy = [0.99663526], Time to run 1 epoch = 120.41 seconds\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "model = MLPModel()\n",
    "trained, loss, accuracies = train_model(model,task_data,task_label,5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "def copy_model(model, x):\n",
    "    '''Copy model weights to a new model.\n",
    "    \n",
    "    Args:\n",
    "        model: model to be copied.\n",
    "        x: An input example. This is used to run\n",
    "            a forward pass in order to add the weights of the graph\n",
    "            as variables.\n",
    "    Returns:\n",
    "        A copy of the model.\n",
    "    '''\n",
    "    copied_model = MLPModel()\n",
    "    \n",
    "    # If we don't run this step the weights are not \"initialized\"\n",
    "    # and the gradients will not be computed.\n",
    "    copied_model.forward(tf.convert_to_tensor(x))\n",
    "    \n",
    "    copied_model.set_weights(model.get_weights())\n",
    "    return copied_model\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "def train_maml(model, data, label, epochs, lr_inner=0.01, batch_size=1, log_steps=1):\n",
    "    '''Train using the MAML setup.\n",
    "    \n",
    "    The comments in this function that start with:\n",
    "        \n",
    "        Step X:\n",
    "        \n",
    "    Refer to a step described in the Algorithm 1 of the paper.\n",
    "    \n",
    "    Args:\n",
    "        model: A model.\n",
    "        epochs: Number of epochs used for training.\n",
    "        dataset: A dataset used for training.\n",
    "        lr_inner: Inner learning rate (alpha in Algorithm 1). Default value is 0.01.\n",
    "        batch_size: Batch size. Default value is 1. The paper does not specify\n",
    "            which value they use.\n",
    "        log_steps: At every `log_steps` a log message is printed.\n",
    "    \n",
    "    Returns:\n",
    "        A strong, fully-developed and trained maml.\n",
    "    '''\n",
    "    optimizer = keras.optimizers.Adam()\n",
    "    \n",
    "    # Step 2: instead of checking for convergence, we train for a number\n",
    "    # of epochs\n",
    "    losses = []\n",
    "    accuracies = []\n",
    "    for epoch in range(epochs):\n",
    "        epoch_totalloss = 0\n",
    "        epoch_totalaccuracy = 0\n",
    "        start = time.time()\n",
    "        # Step 3 and 4\n",
    "        for i in range(len(data)):\n",
    "            z = data[i][0,:].reshape(1,-1)\n",
    "            model.forward(z)  # run forward pass to initialize weights\n",
    "            with tf.GradientTape() as test_tape:\n",
    "                # test_tape.watch(model.trainable_variables)\n",
    "                # Step 5\n",
    "                with tf.GradientTape() as train_tape:\n",
    "                    task_totaltrainloss = 0\n",
    "                    for k in range(len(classes)):\n",
    "                        x = data[i][k,:].reshape(1,-1)\n",
    "                        y = label[i][k]\n",
    "                        task_currtrainloss, _ = compute_loss(model, x, y) \n",
    "                        task_totaltrainloss += task_currtrainloss\n",
    "                        task_trainloss = task_totaltrainloss/(k+1)\n",
    "                # Step 6\n",
    "                gradients = train_tape.gradient(task_trainloss, model.trainable_variables)\n",
    "                k = 0\n",
    "                model_copy = copy_model(model, z)\n",
    "                for j in range(len(model_copy.layers)):\n",
    "                    model_copy.layers[j].kernel = tf.subtract(model.layers[j].kernel,\n",
    "                                tf.multiply(lr_inner, gradients[k]))\n",
    "                    model_copy.layers[j].bias = tf.subtract(model.layers[j].bias,\n",
    "                                tf.multiply(lr_inner, gradients[k+1]))\n",
    "                    k += 2\n",
    "                # Step 8\n",
    "                task_totaltestloss = 0\n",
    "                task_totalaccuracy = 0\n",
    "                for k in range(len(classes)):                  \n",
    "                    x = data[i][k,:].reshape(1,-1)\n",
    "                    y = label[i][k]\n",
    "                    task_currtestloss, logits = compute_loss(model_copy, x, y)\n",
    "                    task_totaltestloss += task_currtestloss\n",
    "                    task_testloss = task_totaltestloss/(k+1)\n",
    "                    task_curraccuracy = accuracy_function(y, logits)\n",
    "                    task_totalaccuracy += task_curraccuracy\n",
    "                    task_accuracy = task_totalaccuracy/(k+1)\n",
    "            # Step 8\n",
    "            gradients = test_tape.gradient(task_testloss, model.trainable_variables)\n",
    "            optimizer.apply_gradients(zip(gradients, model.trainable_variables))\n",
    "            \n",
    "            # Logs\n",
    "            epoch_totalloss += task_testloss\n",
    "            epoch_totalaccuracy += task_accuracy\n",
    "        epoch_loss = epoch_totalloss / (i+1.0)\n",
    "        epoch_accuracy = epoch_totalaccuracy / (i+1.0)\n",
    "        losses.append(epoch_loss)\n",
    "        accuracies.append(epoch_accuracy)\n",
    "            \n",
    "        if epoch % log_steps == 0:\n",
    "            print('Epoch {}: Loss = {}, Train accuracy = {}, Time to run {} epoch = {}'.format(epoch+1, epoch_loss, epoch_accuracy, log_steps, time.time() - start))\n",
    "            start = time.time()\n",
    "    plt.figure()\n",
    "    plt.plot(losses)\n",
    "    plt.title('Loss Vs Epochs')\n",
    "    plt.xlabel('Epochs')\n",
    "    plt.ylabel('Loss')\n",
    "    plt.show()\n",
    "    plt.figure()\n",
    "    plt.plot(accuracies)\n",
    "    plt.title('Train Accuracy Vs Epochs')\n",
    "    plt.xlabel('Epochs')\n",
    "    plt.ylabel('Accuracy')\n",
    "    plt.show()\n",
    "    return model, losses, accuracies"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1: Loss = 0.09502734121579631, Train accuracy = [0.97240915], Time to run 1 epoch = 470.40244030952454\n",
      "Epoch 2: Loss = 0.0531642009105087, Train accuracy = [0.993607], Time to run 1 epoch = 468.9064631462097\n",
      "Epoch 3: Loss = 0.05444150798172311, Train accuracy = [0.993607], Time to run 1 epoch = 468.99922013282776\n",
      "Epoch 4: Loss = 0.1782327909244995, Train accuracy = [0.98687752], Time to run 1 epoch = 470.2843177318573\n",
      "Epoch 5: Loss = 0.0038714136874352023, Train accuracy = [0.99899058], Time to run 1 epoch = 469.50508069992065\n",
      "Epoch 6: Loss = 0.019562444013115077, Train accuracy = [0.99663526], Time to run 1 epoch = 469.4703679084778\n",
      "Epoch 7: Loss = 0.0034555836970238717, Train accuracy = [0.99899058], Time to run 1 epoch = 469.404305934906\n",
      "Epoch 8: Loss = 0.004547005071274467, Train accuracy = [0.99932705], Time to run 1 epoch = 470.5746555328369\n",
      "Epoch 9: Loss = 0.0005626518457456801, Train accuracy = [1.], Time to run 1 epoch = 469.1021752357483\n",
      "Epoch 10: Loss = 7.38895577258551e-05, Train accuracy = [1.], Time to run 1 epoch = 468.64230394363403\n",
      "Epoch 11: Loss = 6.72030933325056e-05, Train accuracy = [1.], Time to run 1 epoch = 468.8531484603882\n",
      "Epoch 12: Loss = 3.1828695549611804e-05, Train accuracy = [1.], Time to run 1 epoch = 470.6192572116852\n",
      "Epoch 13: Loss = 0.048852985722797844, Train accuracy = [0.99697174], Time to run 1 epoch = 468.908940076828\n",
      "Epoch 14: Loss = 2.5201720552886337e-05, Train accuracy = [1.], Time to run 1 epoch = 468.5795488357544\n",
      "Epoch 15: Loss = 0.07743758474969271, Train accuracy = [0.99158816], Time to run 1 epoch = 468.21903824806213\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "train = MLPModel()\n",
    "maml, loss, accuracies = train_maml(train,task_data,task_label,15)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0 P0-W0\n",
      "(373, 504)\n",
      "(373,)\n",
      "1 P1-W1\n",
      "(361, 504)\n",
      "(361,)\n",
      "2 P9-W7\n",
      "(476, 504)\n",
      "(476,)\n",
      "3 P14-W0\n",
      "(461, 504)\n",
      "(461,)\n"
     ]
    }
   ],
   "source": [
    "testpath = r'/home/giang/Lab/Code/Trainsets' # use your path\n",
    "testclasses = {0:'P0-W0', 1:'P1-W1', 2:'P9-W7', 3:'P14-W0'}\n",
    "\n",
    "df_test, x_np_test, y_np_test = {}, {}, {}\n",
    "\n",
    "for i, cls in testclasses.items():\n",
    "    print(i, cls)\n",
    "    all_files = glob.glob(testpath + \"/\" + cls + \"/*.csv\")\n",
    "    df_test[i] = pd.concat((pd.read_csv(f, header=None) for f in all_files))\n",
    "    x_np_test[i] = np.transpose(np.array(df_test[i]))\n",
    "    print(x_np_test[i].shape)\n",
    "   \n",
    "    y_np_test[i] = np.ones(len(x_np_test[i])) * i # Target label\n",
    "    print(y_np_test[i].shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "361\n"
     ]
    }
   ],
   "source": [
    "num_tasks_test = min(len(x_np_test[i]) for i in range(len(testclasses))) \n",
    "print(num_tasks_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(4, 504)\n",
      "(4, 1)\n"
     ]
    }
   ],
   "source": [
    "task_data_test = {}\n",
    "task_label_test = {}\n",
    "for i in range(num_tasks_test):\n",
    "    task_data_test[i] = np.stack([x_np_test[k][i] for k in range(len(testclasses))],axis=0)\n",
    "    task_label_test[i] = np.stack([y_np_test[k][i] for k in range(len(testclasses))],axis=0)\n",
    "    task_label_test[i] = task_label_test[i].reshape(-1,1)\n",
    "print(task_data_test[0].shape)\n",
    "print(task_label_test[0].shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(4, 504)\n",
      "(4, 4)\n"
     ]
    }
   ],
   "source": [
    "data_label_test = {}\n",
    "for i in range(num_tasks_test):\n",
    "    data_label_test[i] = np.concatenate((task_data_test[i],task_label_test[i]),axis=1)\n",
    "    np.random.shuffle(data_label_test[i])\n",
    "    task_data_test[i],task_label_test[i] = np.split(data_label_test[i],[-1],axis=1)\n",
    "    task_label_test[i] = to_categorical(task_label_test[i])\n",
    "print(task_data_test[0].shape)\n",
    "print(task_label_test[0].shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "def adaptation(model, optimizer, data_test, label_test, num_steps = 10, log_step =1):\n",
    "    losses = []\n",
    "    accuracies = []\n",
    "    # If 0 in fits we log the loss before any training\n",
    "\n",
    "    totalacc = 0\n",
    "    totalloss = 0\n",
    "    for k in range(len(testclasses)):\n",
    "        curr_loss, curr_logits = compute_loss(model, data_test[k,:].reshape(1,-1), label_test[k])\n",
    "        curr_acc = accuracy_function(label_test[k],curr_logits)\n",
    "        totalloss += curr_loss\n",
    "        task_loss = totalloss / (k+1)\n",
    "        totalacc += curr_acc\n",
    "        task_acc = totalacc / (k+1)   \n",
    "    losses.append(task_loss)\n",
    "    accuracies.append(task_acc)\n",
    "        \n",
    "    for step in range(1, num_steps):\n",
    "        for k in range(len(testclasses)):\n",
    "            train_batch(data_test[k,:].reshape(1,-1), label_test[k], model, optimizer)\n",
    "            \n",
    "        totalacc = 0\n",
    "        totalloss = 0\n",
    "        for k in range(len(testclasses)):\n",
    "            curr_loss, curr_logits = compute_loss(model, data_test[k,:].reshape(1,-1), label_test[k])\n",
    "            curr_acc = accuracy_function(label_test[k],curr_logits)\n",
    "            totalloss += curr_loss\n",
    "            task_loss = totalloss / (k+1)\n",
    "            totalacc += curr_acc\n",
    "            task_acc = totalacc / (k+1)\n",
    "        if step % log_step == 0:\n",
    "            losses.append(task_loss)\n",
    "            accuracies.append(task_acc)\n",
    "    return accuracies, losses\n",
    "\n",
    "\n",
    "def test(model, data, label, k = 2, num_steps=10, lr=0.01):\n",
    "   \n",
    "    x_test = data[k]\n",
    "    y_test = label[k]\n",
    "    \n",
    "    # copy model so we can use the same model multiple times\n",
    "    copied_model = copy_model(model, x_test)\n",
    "    \n",
    "    # use SGD for this part of training as described in the paper\n",
    "    optimizer = keras.optimizers.SGD(learning_rate=lr)\n",
    "    \n",
    "    # run training and log fit results\n",
    "    accuracies, losses = adaptation(copied_model, optimizer, x_test, y_test, num_steps)\n",
    "    \n",
    "    # plot\n",
    "    plt.figure()\n",
    "    plt.plot(losses , '--')\n",
    "    plt.xlabel('Adaptation Steps')\n",
    "    plt.ylabel('Loss')\n",
    "    plt.show()\n",
    "    \n",
    "    plt.figure()\n",
    "    plt.plot(accuracies , '--')\n",
    "    plt.xlabel('Adaptation Steps')\n",
    "    plt.ylabel('Accuracy')\n",
    "    plt.ylim(0, 1.1)\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYIAAAEGCAYAAABo25JHAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy8QZhcZAAAdZklEQVR4nO3debxVdb3/8debwzwIyqQCAiaICE73OKRWKuaUoZY5lGlm+fOmWZqadk39eft502y8mUpaafpztsKyqOs8y0EZBEQRBY4gMs/T4XzuH3uB+wzABs8665yz3s/Hg4d7re93r/0523P2e6/p+1VEYGZm+dUq6wLMzCxbDgIzs5xzEJiZ5ZyDwMws5xwEZmY51zrrArZVjx49YsCAAVmXYWbWrIwbN25BRPSsr63ZBcGAAQOoqKjIugwzs2ZF0szNtfnQkJlZzjkIzMxyzkFgZpZzDgIzs5xzEJiZ5ZyDwMws5xwEZmY55yAwM8s5B4GZWc45CMzMcs5BYGaWcw4CM7OccxCYmeWcg8DMLOccBGZmOZdaEEj6naQPJb2xmXZJ+pWk6ZImSjogrVrMzGzz0twj+ANw3BbajwcGJf/OB25NsRYzM9uM1GYoi4hnJQ3YQpeTgLsjIoCXJXWTtEtEzE2rJrPmYMb8Ffzm6XfqrD/nkwMY3rcrU+cu487n363Tfv6nd2dw7y6Mn72Ee16uOxnVt4/ag/7dO/Hqu4t4sGJ2nfbLjtmTnbu257m35/OX8XPqtP/HCXuxY6e2/GvKPMZM/qBO+/Un7U3Htq3568Q5PD1tfp32G7+4D2WtxCPjKnlpxsIabW3KxH99YR8A7nt1FuNmLq7R3rlda64buTcAf3jhXd6Ys6xGe/dObbnqhL0AuP2Zd3j7wxU12nft2p5Lj9kTgF898TazFq2q0T6wRycuPHIPAH4y5k3mLVtbo33Izl34xqd2B+BHf53CktXra7Tv27crX/3kAAB++Oc3WL1+Q432AwfsyOkH7gbAFQ9PoDpqvjeH79GDk/fvw7qqan7wp0nUdvRevThu2C511jeULKeq7AMU/zZWJuvqBIGk8ynsNbDbbrs1SnFmjWlDdTCxcgn777Yjy9dU8dI7C+v0OXGfwgfB4lXr6m0//cB+ACxcsbbe9q8dOgCAecvW1Nu+8cNr7pL629dWVQPw/uJV9bZXJZ9uMxfW3174zifeXbCyTnvb1h8dnHh73oo67d06ttn0eNq85XXad+naftPjyXOW1QmST/TqvOnxxMqlTJ1bM0hWrKna9Pj1WUuYubBmUFTHR5/c42Yt5sNaQdG+zUf1v/ruIlasrarRvlOntpsevzRjIdXVNZrp063Dptep770bVFR/GhQRW++1vRsv7BH8NSKG1dP2N+C/IuL5ZPkJ4IqIGLelbZaXl4fnLLaW5jdPT+emf0zjsYsOZ3jfrlmXYy2QpHERUV5fW5ZXDVUC/YqW+wJ190fNWripc5fx83+9xQnDd2ZYnx2yLsdyKMsgGA2cnVw9dAiw1OcHLG/WVVVz6YMT6NqhLT86eTiSsi7Jcii1cwSS7gOOAHpIqgSuBdoARMRtwOPACcB0YBVwblq1mDVVv3ziLabOXcYdZ5fXOI5s1pjSvGrozK20B3BhWq9v1hz03qE9Xz2kP0cP7Z11KZZjWV41ZJZ7ZyeXHJplyUNMmGXg5jHT+Mv497MuwwxwEJg1uhemL+DXT01n/OwlWZdiBjgIzBrV0tXrueyhCezesxPfP25I1uWYAT5HYNao/u9jk/lw+Voe+fdDad+mLOtyzADvEZg1mgmzl/Doa+/zrSM+wX79umVdjtkm3iMwayT79uvG7889kMM+0SPrUsxq8B6BWcoigsrFhUHMjtyzV40B1syaAv9GmqXsT6+/z1E3P8MEXyVkTZSDwCxFc5as5trRk9m3X1eG9fGootY0OQjMUhIRfP+RiWyoDm7+0r6UtfKActY0OQjMUnLPyzN57u0F/Mfn9qJ/905Zl2O2WQ4Cs5QsWLGOo4b04ssHeVY9a9p8+ahZSi757GA2VIfnGLAmz3sEZg3svldnbZp31ucFrDlwEJg1oKlzl3HNX97gnldmZl2KWckcBGYNZG3VBi55YDxdO7TlP08alnU5ZiXzOQKzBvLL/3mbNz9Yzp3neNpJa168R2DWAKbOXcZtz7zDaeV9GbGXp5205sV7BGYNYM/eXfjRycP5/L67ZF2K2TZzEJh9TCvWVtG5XWu+fLDvF7DmyYeGzD6G599ewOE3PsnESg8oZ82Xg8BsOy1dvZ7LH55A905tGdy7S9blmG03Hxoy204bp5181NNOWjPnPQKz7TBm8gc8+tr7XHjEJ9jX005aM+cgMNsOL72zkL133YGLjhqUdSlmH5sPDZlth+tG7s2KtVWedtJaBP8Wm22DJ9+cx1vzlgPQuZ2/R1nL4CAwK9GcJav5zn3juf6xKVmXYtagHARmJaiuDq54eCIbIrjhlOFZl2PWoBwEZiW455WZPD+9MO3kbt07Zl2OWYNKNQgkHSdpmqTpkq6sp303SU9Jel3SREknpFmP2fZ4b8FKbnh8Kp8Z3NPTTlqLlNrZLkllwC3AZ4FKYKyk0RFRfID1auDBiLhV0lDgcWBAWjWZbY+du7bnG4fvzlmH9Pe0k9YipXnZw0HA9IiYASDpfuAkoDgIAtghedwVmJNiPWbbbEN10L5NGZcdu2fWpZilJs1DQ32A2UXLlcm6YtcBZ0mqpLA38O36NiTpfEkVkirmz5+fRq1mdUyZs4zP/vwZpsxZlnUpZqlKMwjq24eOWstnAn+IiL7ACcAfJdWpKSJGRUR5RJT37NkzhVLNalpbtYFLHxzPstVV7Ny1fdblmKUqzSCoBPoVLfel7qGf84AHASLiJaA90CPFmsxK8otk2skbvzjc005ai5dmEIwFBkkaKKktcAYwulafWcAIAEl7UQgCH/uxTI2buYjbPe2k5UhqQRARVcBFwBhgKoWrgyZLul7SyKTb94BvSpoA3Ad8LSJqHz4ya1QPjJ3NLl078MMTh2ZdilmjUHP73C0vL4+Kioqsy7AWrLo6mLtsDX26dci6FLMGI2lcRJTX1+Y7i80SE2YvYd6yNbRqJYeA5YqDwIzCtJMX3DOOb937WtalmDU6j6Nrmfpw+Ro+WLqmxjohhvftCsDsRatYvGpdjfZWEsP6FNpnLlzJ0tXra7S3KWvFXrsU7lOcMX8FK9ZW1Whv17qMPXcuzDE8/cPlrFq3gTuee5cPl6/ltrP+reF+OLNmwkFgmYkojOj59LSaF4q1biWm31AYduqXT7zNw+Mqa7R37dCGCdceA8CP//4mf3/jgxrtfbp14IUrjwLgusem8OxbNbc/uHdn/nnJZwD4/iOTGDdzMQAXjxjkaSctl3yy2DK1tmoDT079sMZMXxIcNaRw2ebkOUvr7DG0LmvFZwYXbiycWLmE+cvX1mhv36aMw/Yo3I7y+qzFLFpZc4+iU7vWHLJ7dwAq3lvE0tXr6dSuNQcP3MljCVmLtaWTxQ4Cy8ToCXM4ZOBO9NrBd+2aNQZfNWRNyoTZS7jkgfH86sm3sy7FzHAQWCNbs74whk+vLu24/NghWZdjZvhksTWyn4yZxjvzV/LH8w6ia4c2WZdjZniPwBrRKzMW8rsX3uXsT/bnU4M8iqxZU+E9Ams0g3t34dxDB3LZsYOzLsXMijgIrFFUVwc7dmrLNZ/3QG5mTY0PDVnqnnxzHqf85gXmLVuz9c5m1ugcBJaqxSvX8f1HJrG2qppuHX1y2Kwp8qEhS9XVf3mDJavWcde5B9GudVnW5ZhZPbxHYKkZPWEOf5s4l+8ePZihu+6QdTlmthkOAktFRHD3i++x/27d+D+f3j3rcsxsC3xoyFIhiXu+cTBLV6+ndZm/b5g1Zf4LtQb36ruLWLm2ivZtyujtQeXMmjwHgTWoWQtX8bXfv8q1oydnXYqZlchBYA1mQ3Vw2UMTKJO49LO+e9isuXAQWIO58/kZvPreIq4duTe7evJ3s2bDQWAN4q15y7l5zFscM7Q3XzygT9blmNk22GoQSLpI0o6NUYw1Xx3alHHkkJ7c8IXhnu7RrJkpZY9gZ2CspAclHSf/lVs9+u3Ukdu/Wk6Pzu2yLsXMttFWgyAirgYGAXcCXwPelnSDpE+kXJs1AxNmL+H8uytYuGLt1jubWZNU0jmCKMxw/0HyrwrYEXhY0k0p1mZN3MZpJye9v9Q3jZk1Y1u9s1jSxcA5wALgDuDyiFgvqRXwNnBFuiVaU3XTPzztpFlLUMoQEz2AL0TEzOKVEVEt6cR0yrKm7qV3PO2kWUtRyv7848CijQuSukg6GCAipqZVmDVdEcFNY95kQPeOXHn8kKzLMbOPqZQguBVYUbS8Mlm3VclVRtMkTZd05Wb6nCZpiqTJkv5/Kdu1bEniznMOZNTZ5XRs63ELzZq7Uv6KlZwsBjYdEirl3EIZcAvwWaCSwiWooyNiSlGfQcBVwGERsVhSr23+CaxRvTN/Bf136shOndqyU6e2WZdjZg2glD2CGZIultQm+fcdYEYJzzsImB4RMyJiHXA/cFKtPt8EbomIxQAR8eG2FG+Na9HKdZx++8tc9eikrEsxswZUShBcABwKvE/hm/3BwPklPK8PMLtouTJZV2wwMFjSC5JelnRcfRuSdL6kCkkV8+fPL+GlraFFBD/88xssXb2Orx8+MOtyzKwBbfUQT/It/Yzt2HZ9dyBHreXWFG5WOwLoCzwnaVhELKlVwyhgFEB5eXntbVgjGD1hDn+bNJfLj92TvXbxtJNmLUkpx/rbA+cBewObZhmJiK9v5amVQL+i5b7AnHr6vBwR64F3JU2jEAxjt166NZYPlq7hmr9M9rSTZi1UKYeG/khhvKFjgWcofKAvL+F5Y4FBkgZKakthr2J0rT5/Bo4EkNSDwqGiUs4/WCNatmY9u+3UkZ+dtp/vIDZrgUr5q94jIn4IrIyIu4DPAcO39qSIqAIuAsYAU4EHI2KypOsljUy6jQEWSpoCPEXhruWF2/ODWHoG9+7C6IsOY2CPTlmXYmYpKOXy0fXJf5dIGkZhvKEBpWw8Ih6ncENa8bprih4HcGnyz5qYmQtX8vsX3uPyY/ekUzvfL2DWUpWyRzAqmY/gagqHdqYAN6ZalWVu47STj7xWydLV67f+BDNrtrb4NS8ZWG5Zcp3/s4DPFObEnc/PYOx7i/npl/b1tJNmLdwW9wgioprCcX7LkeJpJ7/gaSfNWrxSDg39S9JlkvpJ2mnjv9Qrs8xc/ec36NK+taedNMuJUs4Abrxf4MKidYEPE7VYPzl1H+YsWeNpJ81yopQ7iz2eQE4sWrmOHTu2oX/3TvTv7ktFzfKilDuLz65vfUTc3fDlWFbWrN/Al257kUN2787/O2Wrt4mYWQtSyqGhA4setwdGAK8BDoIWZOO0k9eN3DvrUsyskZVyaOjbxcuSulIYdsJaiBffWeBpJ81ybHsGjllFYWA4awGWr1nP5Q9NZGCPTp520iynSjlH8BgfDR/dChgKPJhmUdZ4Zi5cxYbq4Ken7etpJ81yqpS//JuLHlcBMyOiMqV6rJEN69OVZ644gnaty7IuxcwyUsqhoVnAKxHxTES8QGG00AGpVmWpW7RyHaOefYeqDdUOAbOcK2WP4CEKU1VutCFZd2D93Zuu6x+bwrhZi2us69utA7d85QAArnp0ElPmLqvRvkfPzvz0tH0BuOSB8cxYsLJG+/A+O/CjkwuXW/77PeOYs3RNjfaDB+7ED07YC4BzfvcqS2oN4PaZQT249Jg9ATjt9pdYW1Vdo/3YvXvzrSP2oLo6OOXWF+v8TCfvtyvnHjaQlWur+PIdr9Rp//JB/Tj9wN1YsGIt591VsWn9whVrmbdsDUcN6cUevbrUeZ6Z5UcpQdA6mXwegIhYl0w00+x0bldGtw5taqzr0r51jcfb2t65XZua7etqtnds+9G37a61ngvQsWh4564d2rCuVhB0bPPR82u/NkCHpF2qv33jt/1WUo32bh3acMVxQxwCZoYKUwJsoYP0L+C/I2J0snwScHFEjGiE+uooLy+PioqKrXc0M7NNJI2LiPL62krZI7gAuFfSr5PlSqDeu43NzKz5KeWGsneAQyR1prAHUcp8xWZm1kxs9aohSTdI6hYRKyJiuaQdJf2oMYozM7P0lXL56PERsWTjQjJb2QnplWRmZo2plCAok7RpYHpJHQAPVG9m1kKUcrL4HuAJSb9Pls8F7kqvJDMza0ylnCy+SdJE4GhAwD+A/mkXZmZmjaPU0Uc/AKqBL1KYj2BqahWZmVmj2uwegaTBwBnAmcBC4AEKl48e2Ui1mZlZI9jSoaE3geeAz0fEdABJlzRKVWZm1mi2dGjoixQOCT0l6beSRlA4R2BmZi3IZoMgIv4UEacDQ4CngUuA3pJulXRMI9VnZmYp2+rJ4ohYGRH3RsSJQF9gPHBl6pWZmVmj2KY5iyNiUUTcHhFHpVWQmZk1ru2ZvN7MzFqQVINA0nGSpkmaLmmzh5MknSopJNU7VraZmaUntSCQVAbcAhwPDAXOlDS0nn5dgIuBuvMsmplZ6tLcIzgImB4RM5KpLu8HTqqn338CNwFr6mkzM7OUpRkEfYDZRcuVybpNJO0P9IuIv25pQ5LOl1QhqWL+/PkNX6mZWY6lGQT13Xy2aYJkSa2AnwPf29qGImJURJRHRHnPnj0bsEQzM0szCCqBfkXLfYE5RctdgGHA05LeAw4BRvuEsZlZ40ozCMYCgyQNlNSWwgB2ozc2RsTSiOgREQMiYgDwMjAyIipSrMnMzGpJLQgiogq4CBhDYdjqByNisqTrJY1M63XNzGzblDJD2XaLiMeBx2utu2YzfY9IsxYzM6uf7yw2M8s5B4GZWc45CMzMcs5BYGaWcw4CM7OccxCYmeWcg8DMLOccBGZmOecgMDPLOQeBmVnOOQjMzHLOQWBmlnMOAjOznHMQmJnlnIPAzCznHARmZjnnIDAzyzkHgZlZzjkIzMxyzkFgZpZzDgIzs5xzEJiZ5ZyDwMws5xwEZmY55yAwM8s5B4GZWc45CMzMcs5BYGaWcw4CM7OccxCYmeVcqkEg6ThJ0yRNl3RlPe2XSpoiaaKkJyT1T7MeMzOrK7UgkFQG3AIcDwwFzpQ0tFa314HyiNgHeBi4Ka16zMysfmnuERwETI+IGRGxDrgfOKm4Q0Q8FRGrksWXgb4p1mNmZvVIMwj6ALOLliuTdZtzHvD3+hoknS+pQlLF/PnzG7BEMzNLMwhUz7qot6N0FlAO/KS+9ogYFRHlEVHes2fPBizRzMxap7jtSqBf0XJfYE7tTpKOBv4D+ExErE2xHjMzq0eaewRjgUGSBkpqC5wBjC7uIGl/4HZgZER8mGItZma2GakFQURUARcBY4CpwIMRMVnS9ZJGJt1+AnQGHpI0XtLozWzOzMxSkuahISLiceDxWuuuKXp8dJqvb2ZmW+c7i83Mcs5BYGaWcw4CM7OccxCYmeWcg8DMLOccBGZmOecgMDPLOQeBmVnOOQjMzHLOQWBmlnMOAjOznHMQmJnlnIPAzCznHARmZjnnIDAzyzkHgZlZzjkIzMxyzkFgZpZzDgIzs5xzEJiZ5ZyDwMws5xwEZmY55yAwM8s5B4GZWc45CMzMcs5BYGaWcw4CM7OccxCYmeWcg8DMLOccBGZmOecgMDPLuVSDQNJxkqZJmi7pynra20l6IGl/RdKANOsxM7O6UgsCSWXALcDxwFDgTElDa3U7D1gcEXsAPwduTKseMzOrX5p7BAcB0yNiRkSsA+4HTqrV5yTgruTxw8AISUqxJjMzq6V1itvuA8wuWq4EDt5cn4iokrQU6A4sKO4k6Xzg/GRxhaRp21lTj9rbzjm/HzX5/fiI34uaWsL70X9zDWkGQX3f7GM7+hARo4BRH7sgqSIiyj/udloKvx81+f34iN+Lmlr6+5HmoaFKoF/Rcl9gzub6SGoNdAUWpViTmZnVkmYQjAUGSRooqS1wBjC6Vp/RwDnJ41OBJyOizh6BmZmlJ7VDQ8kx/4uAMUAZ8LuImCzpeqAiIkYDdwJ/lDSdwp7AGWnVk/jYh5daGL8fNfn9+Ijfi5pa9PshfwE3M8s331lsZpZzDgIzs5zLTRBsbbiLvJDUT9JTkqZKmizpO1nX1BRIKpP0uqS/Zl1L1iR1k/SwpDeT35NPZl1TViRdkvydvCHpPknts64pDbkIghKHu8iLKuB7EbEXcAhwYY7fi2LfAaZmXUQT8UvgHxExBNiXnL4vkvoAFwPlETGMwkUvaV/QkolcBAGlDXeRCxExNyJeSx4vp/BH3ifbqrIlqS/wOeCOrGvJmqQdgE9TuKKPiFgXEUuyrSpTrYEOyX1OHal7L1SLkJcgqG+4i1x/+AEko73uD7ySbSWZ+wVwBVCddSFNwO7AfOD3yaGyOyR1yrqoLETE+8DNwCxgLrA0Iv6ZbVXpyEsQlDSURZ5I6gw8Anw3IpZlXU9WJJ0IfBgR47KupYloDRwA3BoR+wMrgVyeU5O0I4UjBwOBXYFOks7Ktqp05CUIShnuIjcktaEQAvdGxKNZ15Oxw4CRkt6jcMjwKEn3ZFtSpiqByojYuJf4MIVgyKOjgXcjYn5ErAceBQ7NuKZU5CUIShnuIheSYb7vBKZGxM+yridrEXFVRPSNiAEUfi+ejIgW+a2vFBHxATBb0p7JqhHAlAxLytIs4BBJHZO/mxG00BPnaY4+2mRsbriLjMvKymHAV4FJksYn634QEY9nWJM1Ld8G7k2+NM0Azs24nkxExCuSHgZeo3C13eu00KEmPMSEmVnO5eXQkJmZbYaDwMws5xwEZmY55yAwM8s5B4GZWc45CKzZkHSKpJA0ZAt9/iDp1O3c/hGStnrDUO1+ki6QdPb2vGat7baS9KtkpMtJksZKGpi0/eDjbt9scxwE1pycCTxPeiNAHkFpd47W6BcRt0XE3Q3w+qdTGMpgn4gYDpwCbBzwzUFgqXEQWLOQjI10GHAeRUGggl9LmiLpb0CvorZrkm/Vb0galdwdiqSnJf1C0otJ20HJAHwXAJdIGi/pU5I+L+mVZPC1/5HUezP9rpN0WbLt/SS9LGmipD8l49VsfM0bJb0q6S1Jn6rnx9wFmBsR1QARURkRiyX9mMIImOMl3Zts76xkW+Ml3Z4MtY6kFZJ+Kuk1SU9I6pmsvzh5jyZKur/B/sdYi+AgsObiZApj5L8FLJK0cfybU4A9geHAN6n5jf7XEXFgMpZ8B+DEorZOEXEo8C0Kd5q/B9wG/Dwi9ouI5yjsfRySDL52P3DFZvoVuxv4fkTsA0wCri1qax0RBwHfrbV+oweBzycf7j+VtD9ARFwJrE5e7yuS9qKw93BYROwHbAC+svHnAl6LiAOAZ4pe50pg/6SuC+p7gy2/HATWXJxJ4cOY5L9nJo8/DdwXERsiYg7wZNFzjky+0U8CjgL2Lmq7DyAingV2kNStntfsC4xJnn95refXIakr0C0inklW3ZXUt9HGAf7GAQNqPz8iKimE2lUUhsR+QtKIel5qBPBvwNhkmJARFIaPJnneA8nje4DDk8cTKQwbcRaF4RLMNsnFWEPWvEnqTuGDfJikoDBeVEi6IulSZ5yUZErB31CYXWq2pOuA4mkGaz+nvrFW/hv4WUSMlnQEcN3H+TmAtcl/N7CZv72IWAv8Hfi7pHkU9oSeqNVNwF0RcVUJr7nx5/ochVAaCfxQ0t4R4UAwwHsE1jycCtwdEf0jYkBE9APepfBt91ngDBXmHN4FODJ5zsYP/QXJ+YXaVxKdDiDpcAoTjiwFlgNdivp0Bd5PHp9TtL52PwCSbSwuOv7/VQqHZ0oi6QBJuyaPWwH7ADOT5vXJ8OFQCIZTJfVK+u4kqX/S1qroZ/0y8HyyrX4R8RSFCXi6AZ1LrctaPu8RWHNwJvDjWuseofBB9y0KewuTgLdIPngjYomk3ybr36MwFHmxxZJeBHYAvp6sewx4WNJJFEbgvA54SNL7wMsUJiipr1+xc4DbJHVk20fu7AX8VlK7ZPlV4NfJ41HAREmvJecJrgb+mXzIrwcupBAaK4G9JY0DllIIvDLgnuTQlSic38jz9JNWi0cftdyR9DRwWURUZF1LQ5O0IiL8bd+2iQ8NmZnlnPcIzMxyznsEZmY55yAwM8s5B4GZWc45CMzMcs5BYGaWc/8LMXLNFB3cdskAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "for i in np.random.randint(0, len(task_data_test), size=3):\n",
    "    test(maml,task_data_test,task_label_test,i)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}