demonstration/part2_model_training.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Neural Quantum Dynamics Emulator (Part II)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Scope of this notebook\n",
    "\n",
    "1. We define the configurations for training.\n",
    "2. We define the benchmark models.\n",
    "3. We run training on demo dataset.\n",
    "4. We show the demo models test results."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Standard libraries\n",
    "import os\n",
    "import time\n",
    "import datetime\n",
    "\n",
    "# Computation Libraries\n",
    "import numpy as np\n",
    "import tensorflow as tf\n",
    "from tensorflow import keras\n",
    "from tensorflow.keras import layers\n",
    "from emulator.optimization import create_optimizer\n",
    "\n",
    "#Graphic Libraries\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib.gridspec as gridspec\n",
    "\n",
    "#Custom Libraries\n",
    "from emulator.utils import *"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Section 3: Training Setups"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3.1 Configurations\n",
    "\n",
    "The critical model and training settings are provided bellow."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "class Config:\n",
    "    def __init__(self, **kwargs):\n",
    "        \"\"\" Configuration Settings for Training and Models\"\"\"\n",
    "        \n",
    "        #Read/write Settings\n",
    "        self.qdml_tfrecords = \"../datasets/demo/tfrecords/*\"\n",
    "        self.model_name = \"demo\"\n",
    "        self.models_dir = \"../models_checkpoints/\"\n",
    "        self.log_name = None\n",
    "\n",
    "        #Training Settings\n",
    "        self.train_batch = 128\n",
    "        self.input_frames = 4\n",
    "        self.input_channels = 3\n",
    "        self.output_channels = 2\n",
    "        self.window_size = 23\n",
    "        self.dropout_rate = 0.\n",
    "        self.hidden_size = 69\n",
    "        self.num_train_steps = 900000 #5290000 for full dataset\n",
    "\n",
    "        #Optimizer Settings\n",
    "        self.learning_rate = 1e-3\n",
    "        self.lr_decay_power = 1.0\n",
    "        self.weight_decay_rate = 0.01\n",
    "        self.num_warmup_steps = 9000 # 52900 for full dataset\n",
    "        self.opt_beta_1 = 0.9\n",
    "        self.opt_beta_2 = 0.999\n",
    "        self.end_lr = 1e-6\n",
    "        self.clip_norm = 1.0\n",
    "        \n",
    "        #Misc\n",
    "        self.seed = 711\n",
    "        self.logging_steps = 1000\n",
    "        self.save_ckpt_steps = 1000000\n",
    "        \n",
    "        self.update(kwargs)        \n",
    "        self.model_dir = os.path.join(self.models_dir, self.model_name)\n",
    "        self.log_dir = os.path.join(self.model_dir, \"logs\")\n",
    "        self.ckpts = os.path.join(self.model_dir, \"ckpts\")\n",
    "    \n",
    "    def update(self, kwargs):\n",
    "        for k, v in kwargs.items():\n",
    "            if v is not None:\n",
    "                self.__dict__[k] = v\n",
    "\n",
    "config = Config()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3.2 Load dataset\n",
    "\n",
    "Here we load the windowed-dataset generated from last notebook. In this code snippet, we generate a tensorflow dataset with shuffled and batched training data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "dataset = iter(get_dataset(config))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "All shapes are: (Batch, (Time_steps,) Window_size, Channels)\n",
      "Inputs shape: (128, 4, 23, 3)\n",
      "labels shape: (128, 23, 2)\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1224x360 with 12 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "def plot_windows(config, feature, target):\n",
    "    \"\"\" For any feature and target, plot all input frames and target frame\"\"\"\n",
    "    plt.style.use('seaborn-pastel')\n",
    "    ylabel = [\"Re(Psi)\", \"Im(Psi)\", \"Potential\"]\n",
    "    \n",
    "    #Plot the input frames\n",
    "    fig = plt.figure(figsize=(17, 5))\n",
    "    outer = gridspec.GridSpec(1, config.input_frames)\n",
    "    \n",
    "    for i in range(config.input_frames):\n",
    "        inner = gridspec.GridSpecFromSubplotSpec(config.input_channels, 1, subplot_spec=outer[i], wspace=0.1, hspace=0.1)\n",
    "        for j in range(config.input_channels):\n",
    "            ax = plt.Subplot(fig, inner[j])\n",
    "            y = feature[i, :, j]\n",
    "            ax.plot(y, c=\"blue\")\n",
    "            if j == 0:\n",
    "                ax.set_title(f\"Input Frame #{i+1}\")\n",
    "            if j != 2:\n",
    "                ax.set_xticks([])\n",
    "            if i == 0:\n",
    "                ax.set_ylabel(ylabel[j])\n",
    "                ax.yaxis.set_label_coords(-0.15, 0.5)\n",
    "            fig.add_subplot(ax)\n",
    "    \n",
    "    #Plot the target frame\n",
    "    fig2 = plt.figure()\n",
    "    label = gridspec.GridSpec(config.output_channels,1)\n",
    "    for i in range(2):\n",
    "        ax = plt.Subplot(fig2,label[i])\n",
    "        ax.plot(target[:,i],c='blue')\n",
    "        ax.set_ylabel(ylabel[i])\n",
    "        if i==0:\n",
    "            ax.set_title(\"Target Frame\")\n",
    "            ax.set_xticks([])\n",
    "        fig2.add_subplot(ax)\n",
    "\n",
    "batch_feature, batch_target = next(dataset)\n",
    "print('All shapes are: (Batch, (Time_steps,) Window_size, Channels)')\n",
    "print(f'Inputs shape: {batch_feature.shape}')\n",
    "print(f'labels shape: {batch_target.shape}')\n",
    "plot_windows(config, batch_feature[2], batch_target[7])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Section 4: Models"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The minimal trainable model we can apply to this task is to insert linear transformation between the input and output. In this case, we choose to let the output only depend on the last timestep. This model is treated as the baseline model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"linear_model\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "lambda (Lambda)              multiple                  0         \n",
      "_________________________________________________________________\n",
      "flatten (Flatten)            multiple                  0         \n",
      "_________________________________________________________________\n",
      "dense (Dense)                multiple                  3220      \n",
      "_________________________________________________________________\n",
      "reshape (Reshape)            multiple                  0         \n",
      "=================================================================\n",
      "Total params: 3,220\n",
      "Trainable params: 3,220\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "class LinearModel(tf.keras.Model):\n",
    "    def __init__(self, config):\n",
    "        super().__init__()\n",
    "        self.last = layers.Lambda(lambda x: x[:, -1, :, :])\n",
    "        self.flatten = layers.Flatten()\n",
    "        self.dense = layers.Dense(config.window_size * config.output_channels)\n",
    "        self.reshape = layers.Reshape((config.window_size, config.output_channels))\n",
    "\n",
    "    def call(self, input_tensor, training=False):\n",
    "        x = self.last(input_tensor)\n",
    "        x = self.flatten(x)\n",
    "        x = self.dense(x)\n",
    "        x = self.reshape(x)\n",
    "\n",
    "        return x\n",
    "\n",
    "conf_lin = Config(model_name=\"demo-lin\")\n",
    "m_lin = LinearModel(conf_lin)\n",
    "m_lin.build((None, conf_lin.input_frames, conf_lin.window_size, conf_lin.input_channels))\n",
    "m_lin.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We stack more linear layers to check the performance of a deeper and more expressive model. Moreover, the model takes multiple timesteps to see how the features change overtime before making predictions. ReLU activation function is used to introduce some non-linearities. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"dense_model\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "reshape_1 (Reshape)          multiple                  0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              multiple                  4830      \n",
      "_________________________________________________________________\n",
      "conv1d (Conv1D)              multiple                  19113     \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              multiple                  3220      \n",
      "_________________________________________________________________\n",
      "reshape_2 (Reshape)          multiple                  0         \n",
      "=================================================================\n",
      "Total params: 27,163\n",
      "Trainable params: 27,163\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "class DenseModel(tf.keras.Model):\n",
    "    def __init__(self, config):\n",
    "        super().__init__()\n",
    "        self.flatten = layers.Reshape((config.input_frames, config.window_size * config.input_channels))\n",
    "        self.dense1 = layers.Dense(config.hidden_size, activation=\"relu\")\n",
    "        # conv1d equivlent to a dense layer here\n",
    "        self.conv1d = layers.Conv1D(filters=config.hidden_size, kernel_size=(config.input_frames), activation='relu')\n",
    "        self.dense2 = layers.Dense(config.window_size * config.output_channels)\n",
    "        self.reshape = layers.Reshape((config.window_size, config.output_channels))\n",
    "\n",
    "    def call(self, input_tensor, training=False):\n",
    "        x = self.flatten(input_tensor)\n",
    "        x = self.dense1(x)\n",
    "        x = self.conv1d(x)\n",
    "        x = self.dense2(x)\n",
    "        x = self.reshape(x)\n",
    "        return x\n",
    "\n",
    "\n",
    "conf_dense = Config(model_name=\"demo-dense\")\n",
    "m_dense = DenseModel(conf_dense)\n",
    "m_dense.build((None, conf_dense.input_frames, conf_dense.window_size, conf_dense.input_channels))\n",
    "m_dense.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "A convolution layer can also take multiple time steps as input to make a prediction."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"cnn_model\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "flatten_1 (Flatten)          multiple                  0         \n",
      "_________________________________________________________________\n",
      "conv2d (Conv2D)              multiple                  208       \n",
      "_________________________________________________________________\n",
      "dense_3 (Dense)              multiple                  25461     \n",
      "_________________________________________________________________\n",
      "dense_4 (Dense)              multiple                  3220      \n",
      "_________________________________________________________________\n",
      "reshape_3 (Reshape)          multiple                  0         \n",
      "=================================================================\n",
      "Total params: 28,889\n",
      "Trainable params: 28,889\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "class CNNModel(tf.keras.Model):\n",
    "    def __init__(self, config):\n",
    "        super().__init__()\n",
    "        self.flatten = layers.Flatten()\n",
    "        self.conv1d =layers.Conv2D(filters=16, kernel_size=(config.input_frames, 1), activation='relu')\n",
    "        self.dense1 = layers.Dense(config.hidden_size, activation=\"relu\")\n",
    "        self.dense2 = layers.Dense(config.window_size * config.output_channels)\n",
    "        self.reshape = layers.Reshape((config.window_size, config.output_channels))\n",
    "\n",
    "    def call(self, input_tensor, training=False):\n",
    "        x = self.conv1d(input_tensor)\n",
    "        x = self.flatten(x)\n",
    "        x = self.dense1(x)\n",
    "        x = self.dense2(x)\n",
    "        x = self.reshape(x)\n",
    "        return x\n",
    "    \n",
    "conf_conv = Config(model_name=\"demo-conv\")\n",
    "m_conv = CNNModel(conf_conv)\n",
    "m_conv.build((None, conf_conv.input_frames, conf_conv.window_size, conf_conv.input_channels))\n",
    "m_conv.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "A Recurrent Neural Network (RNN) is a type of neural network well-suited to time series data. RNNs process a time series step by step and maintain an internal state from time-step to time-step. In this model, the recurrent layer warms up its internal state during first few timesteps and only returns the output of the final timestep. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"rnn_model\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "reshape_4 (Reshape)          multiple                  0         \n",
      "_________________________________________________________________\n",
      "dense_5 (Dense)              multiple                  4830      \n",
      "_________________________________________________________________\n",
      "gru (GRU)                    multiple                  28980     \n",
      "_________________________________________________________________\n",
      "dense_6 (Dense)              multiple                  6394      \n",
      "_________________________________________________________________\n",
      "reshape_5 (Reshape)          multiple                  0         \n",
      "=================================================================\n",
      "Total params: 40,204\n",
      "Trainable params: 40,204\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "class RNNModel(tf.keras.Model):\n",
    "    def __init__(self, config):\n",
    "        super().__init__()\n",
    "        self.flatten = layers.Reshape((config.input_frames, config.window_size * config.input_channels)) #Keep 2 dimenional instead of 1 dimensional and have reccurent in 1 dime\n",
    "        self.dense1 = layers.Dense(config.hidden_size, activation=\"relu\") #Try to remove and check\n",
    "        self.gru = tf.keras.layers.GRU(config.hidden_size, return_state=True, activation='relu', dropout=config.dropout_rate)\n",
    "        self.dense2 = layers.Dense(config.window_size * config.output_channels)\n",
    "        self.reshape = layers.Reshape((config.window_size, config.output_channels))\n",
    "\n",
    "    def call(self, input_tensor, training=False):\n",
    "        x = self.flatten(input_tensor)\n",
    "        x = self.dense1(x)\n",
    "        output, final_state = self.gru(x, initial_state=None, training=training)\n",
    "        x = tf.concat((output, final_state), axis=-1)\n",
    "        x = self.dense2(x)\n",
    "        x = self.reshape(x)\n",
    "        return x\n",
    "\n",
    "conf_gru = Config(model_name=\"demo-gru\")\n",
    "m_gru = RNNModel(conf_gru)\n",
    "m_gru.build((None, conf_gru.input_frames, conf_gru.window_size, conf_gru.input_channels))\n",
    "m_gru.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Section 5: Training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "def train(config, model):\n",
    "    \"\"\" Main training loop \"\"\"\n",
    "    \n",
    "    def train_step(feature, target, model, optimizer):\n",
    "        mse = tf.keras.losses.MeanSquaredError()\n",
    "        with tf.GradientTape() as tape:\n",
    "            predict = model(feature, True)\n",
    "            loss = mse(target, predict)\n",
    "        \n",
    "        gradients = tape.gradient(loss, model.trainable_variables)\n",
    "        clipped_gradients = [tf.clip_by_norm(g, config.clip_norm) for g in gradients]\n",
    "        optimizer.apply_gradients(zip(clipped_gradients, model.trainable_variables))\n",
    "\n",
    "        return loss\n",
    "    \n",
    "    #Seed for replication\n",
    "    tf.random.set_seed(config.seed)\n",
    "    \n",
    "    #Logging during training\n",
    "    log_name = datetime.datetime.now().strftime(\"%Y%m%d-%H%M%S\") if config.log_name is None else config.log_name\n",
    "    train_log_dir = config.log_dir + \"/\" + log_name\n",
    "    train_summary_writer = tf.summary.create_file_writer(train_log_dir)\n",
    "    f = open(train_log_dir + \"/\" + \"config.txt\", \"w\")\n",
    "    f.write(str(config.__dict__))\n",
    "    f.close()\n",
    "\n",
    "    #Load dataset from config\n",
    "    dataset = iter(get_dataset(config))\n",
    "    \n",
    "    \n",
    "    #Optimizer\n",
    "    optimizer, lr_schedule = create_optimizer(\n",
    "        init_lr=config.learning_rate,\n",
    "        num_train_steps=config.num_train_steps,\n",
    "        num_warmup_steps=config.num_warmup_steps,\n",
    "        weight_decay_rate=config.weight_decay_rate,\n",
    "        power=config.lr_decay_power,\n",
    "        adam_beta1=config.opt_beta_1,\n",
    "        adam_beta2=config.opt_beta_2,\n",
    "        min_lr_ratio=config.end_lr / config.learning_rate)\n",
    "    \n",
    "    tf.config.optimizer.set_jit(True)\n",
    "\n",
    "    \n",
    "    metrics = {\"train_loss\": tf.keras.metrics.Mean(name=\"train_loss\")}\n",
    "\n",
    "    step = 0\n",
    "    start_time = logging_time = time.time()\n",
    "    \n",
    "    #Training for every step\n",
    "    while step < config.num_train_steps:\n",
    "        feature, target = next(dataset)\n",
    "        loss = train_step(feature, target, model, optimizer)\n",
    "        metrics[\"train_loss\"].update_state(loss)\n",
    "\n",
    "        step += 1\n",
    "        if step % config.logging_steps == 0:\n",
    "            print(\"Step: {:6d}, Loss: {:.8f}, Time per logging: {:.2f}s, Elapsed: {:}, ETA: {:}\".format(\n",
    "                step,\n",
    "                metrics[\"train_loss\"].result(),\n",
    "                time.time() - logging_time,\n",
    "                get_readable_time(time.time() - start_time),\n",
    "                get_readable_time((time.time() - start_time) / step * (config.num_train_steps - step))), flush=True)\n",
    "            with train_summary_writer.as_default():\n",
    "                for k in metrics:\n",
    "                    tf.summary.scalar(k, metrics[k].result(), step=step)\n",
    "                    metrics[k].reset_states()\n",
    "            logging_time = time.time()\n",
    "\n",
    "        if step % config.save_ckpt_steps == 0 and step != config.num_train_steps:\n",
    "            model.save_weights(os.path.join(config.ckpts, \"step_\" + str(step)))\n",
    "            print(\"Model weights saved at \", os.path.join(config.ckpts, \"step_\" + str(step)))\n",
    "\n",
    "    model.save_weights(os.path.join(config.ckpts, \"final_step\"))\n",
    "    print(\"Model Trained!\")\n",
    "    print(\"Model weights saved at \", os.path.join(config.ckpts, \"final_step\"))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Step:   1000, Loss: 0.00536922, Time per logging: 8.58s, Elapsed:  0h 0m 8s, ETA:  2h 8m30s\n",
      "Step:   2000, Loss: 0.00076312, Time per logging: 7.26s, Elapsed:  0h 0m15s, ETA:  1h59m26s\n",
      "Step:   3000, Loss: 0.00025804, Time per logging: 7.33s, Elapsed:  0h 0m23s, ETA:  1h56m21s\n",
      "Step:   4000, Loss: 0.00013423, Time per logging: 7.37s, Elapsed:  0h 0m30s, ETA:  1h55m 7s\n",
      "Step:   5000, Loss: 0.00010109, Time per logging: 7.37s, Elapsed:  0h 0m38s, ETA:  1h54m22s\n",
      "(...)\n",
      "Step: 896000, Loss: 0.00000054, Time per logging: 13.54s, Elapsed:  3h21m51s, ETA:  0h 0m54s\n",
      "Step: 897000, Loss: 0.00000071, Time per logging: 14.01s, Elapsed:  3h22m 5s, ETA:  0h 0m40s\n",
      "Step: 898000, Loss: 0.00000070, Time per logging: 16.26s, Elapsed:  3h22m21s, ETA:  0h 0m27s\n",
      "Step: 899000, Loss: 0.00000054, Time per logging: 14.26s, Elapsed:  3h22m36s, ETA:  0h 0m13s\n",
      "Step: 900000, Loss: 0.00000069, Time per logging: 13.38s, Elapsed:  3h22m49s, ETA:  0h 0m 0s\n",
      "Model Trained!\n",
      "Model weights saved at  ../models/demo-gru\\ckpts\\final_step\n"
     ]
    }
   ],
   "source": [
    "\"\"\"Training GRU model for demo dataset\"\"\"\n",
    "conf_gru = Config(model_name=\"demo-gru\")   #model_name = \"full-gru\" for the full dataset\n",
    "m_gru = RNNModel(conf_gru)\n",
    "train(conf_gru, m_gru)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Section 6: Demo Models Test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "def plot_error(ckpts_list, test_file, pot_scalar=10.0, skip_frames=0, frames=400, norm=False):\n",
    "    \"\"\"Plot loss over time for all four type of models\"\"\"\n",
    "    data = retrieve_data(test_file, pot_scalar)\n",
    "    fig = plt.figure(figsize=(12,8))\n",
    "    plt.title('Error comparison for different models')\n",
    "    plt.xlabel('Time')\n",
    "    plt.ylabel('Error (Log Scale)')\n",
    "    for model, ckpt in ckpts_list:\n",
    "        model.load_weights(ckpt)\n",
    "        result = get_result(model, config, data, skip_frames, frames, normalization=norm)\n",
    "        loss = get_loss(config, data, result, skip_frames, frames)\n",
    "        plt.plot(loss, label=ckpt.split('/')[2])\n",
    "    plt.yscale('log')\n",
    "    plt.legend()\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "\"\"\"Error Comparison between models\"\"\"\n",
    "ckpts_list = [\n",
    "    (LinearModel(config), \"../models_checkpoints/demo-lin/ckpts/final_step\"),\n",
    "    (DenseModel(config), \"../models_checkpoints/demo-dense/ckpts/final_step\"),\n",
    "    (CNNModel(config), \"../models_checkpoints/demo-conv/ckpts/final_step\"),\n",
    "    (RNNModel(config), \"../models_checkpoints/demo-gru/ckpts/final_step\")\n",
    "]\n",
    "\n",
    "\"\"\"This test case has a single potential barrier\"\"\"\n",
    "# Important! First you must download test cases and unpack them in datasets directory\n",
    "# URL: https://drive.google.com/file/d/1lBxIZyfRkTNGhX5sWRJES2qomKKKKywj/view?usp=sharing\n",
    "plot_error(ckpts_list, \"../datasets/test/full_test_E0_0.0to9.0_BH_0.0to14.0/X0_15.1_S0_1.7_E0_6.4_BH_7.4_BW_9.4.txt\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Processed 1/100 (1.0%), ELAPSED:  0h 0m 5s, ETA:  0h 9m25s\n",
      "Processed 2/100 (2.0%), ELAPSED:  0h 0m12s, ETA:  0h10m15s\n",
      "Processed 3/100 (3.0%), ELAPSED:  0h 0m19s, ETA:  0h10m16s\n",
      "Processed 4/100 (4.0%), ELAPSED:  0h 0m33s, ETA:  0h13m21s\n",
      "Processed 5/100 (5.0%), ELAPSED:  0h 0m39s, ETA:  0h12m22s\n",
      "(...)\n",
      "Processed 96/100 (96.0%), ELAPSED:  0h13m30s, ETA:  0h 0m33s\n",
      "Processed 97/100 (97.0%), ELAPSED:  0h13m37s, ETA:  0h 0m25s\n",
      "Processed 98/100 (98.0%), ELAPSED:  0h13m44s, ETA:  0h 0m16s\n",
      "Processed 99/100 (99.0%), ELAPSED:  0h13m50s, ETA:  0h 0m 8s\n",
      "Processed 100/100 (100.0%), ELAPSED:  0h14m 4s, ETA:  0h 0m 0s\n"
     ]
    }
   ],
   "source": [
    "#Save Loss and Correlations for all test cases\n",
    "evaluate(config, ckpts_list, \"../datasets/test/full_test_E0_0.0to9.0_BH_0.0to14.0/*.txt\", \"demo-results.npz\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "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": [
    "#Plot Mean and Standard Deviation of Error over all test files\n",
    "npzfile = np.load(\"demo-results.npz\")\n",
    "x = np.arange(len(npzfile[\"ckpts\"]))\n",
    "all_loss = npzfile[\"all_loss\"]\n",
    "test_mae = np.mean(all_loss, axis = (1, 2))\n",
    "plt.bar(x, test_mae, 0.3)\n",
    "plt.xticks(ticks=x, labels=npzfile[\"ckpts\"],rotation=45)\n",
    "plt.ylabel('MAE (average over all outputs)')\n",
    "plt.yscale('log')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "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": [
    "#Plot Mean and Standard Deviation of Correlation over all test files\n",
    "npzfile = np.load(\"demo-results.npz\")\n",
    "x = np.arange(len(npzfile[\"ckpts\"]))\n",
    "all_corr = npzfile[\"all_corr\"]\n",
    "test_corr = np.mean(all_corr, axis = (1, 2))\n",
    "test_std = np.std(np.mean(all_corr, axis = 2), axis = 1)\n",
    "plt.bar(x, test_corr, 0.3, yerr=test_std, capsize=5)\n",
    "plt.xticks(ticks=x, labels=npzfile[\"ckpts\"],rotation=45)\n",
    "plt.ylabel('Correlations (average over all outputs)')\n",
    "plt.show()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}