selective_classification.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import math\n",
    "import os\n",
    "import random\n",
    "import time\n",
    "\n",
    "import numpy as np\n",
    "import matplotlib\n",
    "import matplotlib.pyplot as plt\n",
    "import torch\n",
    "\n",
    "from models import TextLSTM, TextTCN\n",
    "from models_variational import TextLSTM as TextLSTMVariational\n",
    "from models_variational import TextTCN as TextTCNVariational\n",
    "from torchtext import data, datasets\n",
    "from torchtext.vocab import GloVe\n",
    "from variational import ELBO\n",
    "from yelp_reviews import YELP"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "SEED = 1234\n",
    "torch.manual_seed(SEED)\n",
    "torch.cuda.manual_seed(SEED)\n",
    "random.seed(SEED)\n",
    "rand_state = random.getstate()\n",
    "\n",
    "torch.backends.cudnn.deterministic = True\n",
    "torch.backends.cudnn.benchmark = False\n",
    "\n",
    "device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')\n",
    "batch_size = 128"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pylab as pylab\n",
    "params = {'legend.fontsize': 'x-large',\n",
    "          'figure.figsize': (5, 2),\n",
    "         'axes.labelsize': 'x-large',\n",
    "         'axes.titlesize':'x-large',\n",
    "         'xtick.labelsize':'x-large',\n",
    "         'ytick.labelsize':'x-large'}\n",
    "pylab.rcParams.update(params)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "def plot(coverages, risks_sr, risks_var, dims):\n",
    "    plt.figure(figsize=dims)\n",
    "    plt.plot(coverages, risks_sr, label='SR')\n",
    "    plt.plot(coverages, risks_var, label='BV')\n",
    "    plt.legend(loc='lower right')\n",
    "    plt.xlabel('coverage', fontsize=20)\n",
    "    plt.ylabel('risk', fontsize=20)\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Dataset loading"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "def load_imdb():\n",
    "    TEXT = data.Field(fix_length=400, lower=True,\n",
    "                      tokenize='spacy', batch_first=False)\n",
    "    LABEL = data.LabelField(dtype=torch.long)\n",
    "    train_data, test_data = datasets.IMDB.splits(\n",
    "        TEXT, LABEL, root='./data/imdb/')\n",
    "    train_data, valid_data = train_data.split(\n",
    "        split_ratio=0.8, stratified=False, random_state=rand_state)\n",
    "    TEXT.build_vocab(train_data, vectors=GloVe(\n",
    "    name='6B', dim=100, cache='./glove/'))\n",
    "    print(len(TEXT.vocab.itos))\n",
    "    print()\n",
    "    LABEL.build_vocab(train_data)\n",
    "    return train_data, valid_data, test_data\n",
    "\n",
    "def load_sst():\n",
    "    TEXT = data.Field(fix_length=30, lower=True,\n",
    "                      tokenize='spacy', batch_first=False)\n",
    "    LABEL = data.LabelField(dtype=torch.long)\n",
    "    train_data, valid_data, test_data = datasets.SST.splits(TEXT, LABEL,\n",
    "                                                            root='./data/sst/',\n",
    "                                                            fine_grained=False,\n",
    "                                                            filter_pred=lambda ex:\n",
    "                                                            ex.label != 'neutral')\n",
    "    TEXT.build_vocab(train_data, vectors=GloVe(\n",
    "    name='6B', dim=100, cache='./glove/'))\n",
    "    LABEL.build_vocab(train_data)\n",
    "    return train_data, valid_data, test_data\n",
    "\n",
    "def load_yelp():\n",
    "    TEXT = data.Field(fix_length=300, lower=True,\n",
    "                      tokenize='spacy', batch_first=False)\n",
    "    LABEL = data.LabelField(dtype=torch.long)\n",
    "    train_data, test_data = YELP.splits(\n",
    "        TEXT, LABEL, root='./data/yelp/')\n",
    "    train_data, valid_data = train_data.split(\n",
    "        split_ratio=0.9, stratified=False, random_state=rand_state)\n",
    "    TEXT.build_vocab(train_data, vectors=GloVe(\n",
    "    name='6B', dim=100, cache='./glove/'))\n",
    "    LABEL.build_vocab(train_data)\n",
    "    return train_data, valid_data, test_data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Predictions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "def predict(model, iterator):\n",
    "    def predict_on_batch(model, batch):\n",
    "        model.eval()\n",
    "        with torch.no_grad():\n",
    "            x, y = batch.text, batch.label\n",
    "            y_pred = model(x)\n",
    "        return y_pred, y\n",
    "    \n",
    "    preds = []\n",
    "    scores = []\n",
    "    targets = []\n",
    "    \n",
    "    for batch in iterator:\n",
    "        batch_scores, batch_targets = predict_on_batch(model, batch)\n",
    "        batch_preds = torch.argmax(batch_scores, dim=1)      \n",
    "        preds += [batch_preds.cpu().numpy()]\n",
    "        scores += [batch_scores.cpu().numpy()]\n",
    "        targets += [batch_targets.cpu().numpy()]\n",
    "    \n",
    "    preds = np.hstack(preds)\n",
    "    scores = np.vstack(scores)\n",
    "    targets = np.hstack(targets)\n",
    "    print(preds.shape)\n",
    "    print(scores.shape)\n",
    "    print(targets.shape)\n",
    "    \n",
    "    return preds, scores, targets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "def bbb_predict(model, iterator, T=10):\n",
    "    def bbb_predict_on_batch(model, batch, T):\n",
    "        model.eval()\n",
    "        preds = []\n",
    "        x, y = batch.text, batch.label\n",
    "        targets = y\n",
    "        for _ in range(T):\n",
    "            with torch.no_grad():\n",
    "                y_pred = model(x)\n",
    "                preds += [y_pred]\n",
    "        preds = torch.stack(preds)\n",
    "        pred_means = torch.mean(preds, dim=0)\n",
    "        pred_vars = torch.var(preds, 0, False, False)\n",
    "        pred_vars = torch.mean(pred_vars, dim=-1)\n",
    "        return pred_means, targets, pred_vars\n",
    "    \n",
    "    preds = []\n",
    "    scores = []\n",
    "    targets = []\n",
    "    variances = []\n",
    "    \n",
    "    for batch in iterator:\n",
    "        batch_scores, batch_targets, batch_vars = bbb_predict_on_batch(model, batch, T)\n",
    "        batch_preds = torch.argmax(batch_scores, dim=1)      \n",
    "        preds += [batch_preds.cpu().numpy()]\n",
    "        scores += [batch_scores.cpu().numpy()]\n",
    "        targets += [batch_targets.cpu().numpy()]\n",
    "        variances += [batch_vars.cpu().numpy()]\n",
    "    \n",
    "    preds = np.hstack(preds)\n",
    "    scores = np.vstack(scores)\n",
    "    targets = np.hstack(targets)\n",
    "    variances = np.hstack(variances)\n",
    "    print(preds.shape)\n",
    "    print(scores.shape)\n",
    "    print(targets.shape)\n",
    "    print(variances.shape)\n",
    "    \n",
    "    return preds, scores, targets, variances"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Selective risk"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "def selective_risk_at_coverage(coverage, kappas, preds, targets):\n",
    "    assert coverage >= 0 and coverage <= 1\n",
    "    if coverage == 0:\n",
    "        return 0\n",
    "    assert preds.shape[0] == targets.shape[0]\n",
    "    assert kappas.shape[0] == targets.shape[0]\n",
    "    kappas_sorted = np.sort(kappas)\n",
    "    if coverage == 1:\n",
    "        threshold = kappas_sorted[0]\n",
    "    else:\n",
    "        threshold = kappas_sorted[-int(coverage * preds.size)]\n",
    "    covered_idx = kappas > threshold\n",
    "    if not np.any(covered_idx):\n",
    "        return 0\n",
    "    selective_risk = 1 - np.mean(preds[covered_idx] == targets[covered_idx])\n",
    "    return selective_risk"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0.  , 0.02, 0.04, 0.06, 0.08, 0.1 , 0.12, 0.14, 0.16, 0.18, 0.2 ,\n",
       "       0.22, 0.24, 0.26, 0.28, 0.3 , 0.32, 0.34, 0.36, 0.38, 0.4 , 0.42,\n",
       "       0.44, 0.46, 0.48, 0.5 , 0.52, 0.54, 0.56, 0.58, 0.6 , 0.62, 0.64,\n",
       "       0.66, 0.68, 0.7 , 0.72, 0.74, 0.76, 0.78, 0.8 , 0.82, 0.84, 0.86,\n",
       "       0.88, 0.9 , 0.92, 0.94, 0.96, 0.98, 1.  ])"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "coverages = np.linspace(0,1,51)\n",
    "coverages"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## SST-2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "_, valid_data, _ = load_sst()\n",
    "iterator = data.BucketIterator(valid_data, batch_size=batch_size, device=device)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['best_lstm_5_val_acc=0.8038991.pth',\n",
       " 'best_lr_NO_drop_lstm_26_val_acc=0.8256881.pth',\n",
       " 'best_lr_NO_drop_tcn_23_val_acc=0.8188073.pth',\n",
       " 'best_lr_lstm_variational_laplace_27_val_acc=0.8084862.pth',\n",
       " 'best_tcn_variational_laplace_14_val_acc=0.809633.pth',\n",
       " 'best_lr_lstm_variational_normal_27_val_acc=0.8084862.pth',\n",
       " 'best_tcn_4_val_acc=0.8188073.pth',\n",
       " 'best_lr_tcn_variational_laplace_28_val_acc=0.7821101.pth',\n",
       " 'best_tcn_variational_normal_14_val_acc=0.809633.pth',\n",
       " 'best_NO_drop_tcn_4_val_acc=0.8199541.pth',\n",
       " 'best_lstm_variational_normal_4_val_acc=0.8107798.pth',\n",
       " 'best_lr_tcn_variational_normal_28_val_acc=0.7821101.pth',\n",
       " 'best_lr_tcn_21_val_acc=0.8245413.pth',\n",
       " 'best_NO_drop_lstm_5_val_acc=0.8084862.pth',\n",
       " 'best_lr_lstm_26_val_acc=0.8188073.pth',\n",
       " 'best_lstm_variational_laplace_4_val_acc=0.8107798.pth']"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "saved_folder = './models_final/sst/'\n",
    "os.listdir(saved_folder)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['best_lr_NO_drop_tcn_23_val_acc=0.8188073.pth',\n",
       " 'best_tcn_variational_laplace_14_val_acc=0.809633.pth',\n",
       " 'best_tcn_4_val_acc=0.8188073.pth',\n",
       " 'best_lr_tcn_variational_laplace_28_val_acc=0.7821101.pth',\n",
       " 'best_tcn_variational_normal_14_val_acc=0.809633.pth',\n",
       " 'best_NO_drop_tcn_4_val_acc=0.8199541.pth',\n",
       " 'best_lr_tcn_variational_normal_28_val_acc=0.7821101.pth',\n",
       " 'best_lr_tcn_21_val_acc=0.8245413.pth']"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "name = 'tcn'\n",
    "[l for l in os.listdir(saved_folder) if name in l]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "TextTCN(\n",
       "  (embedding): Embedding(13962, 100, padding_idx=1)\n",
       "  (tcn): TemporalConvNet(\n",
       "    (network): Sequential(\n",
       "      (0): TemporalBlock(\n",
       "        (conv1): Conv1dPathwise(100, 100, kernel_size=(3,), stride=(1,), padding=(2,))\n",
       "        (chomp1): Chomp1d()\n",
       "        (relu1): ReLU()\n",
       "        (conv2): Conv1dPathwise(100, 100, kernel_size=(3,), stride=(1,), padding=(2,))\n",
       "        (chomp2): Chomp1d()\n",
       "        (relu2): ReLU()\n",
       "        (net): Sequential(\n",
       "          (0): Conv1dPathwise(100, 100, kernel_size=(3,), stride=(1,), padding=(2,))\n",
       "          (1): Chomp1d()\n",
       "          (2): ReLU()\n",
       "          (3): Conv1dPathwise(100, 100, kernel_size=(3,), stride=(1,), padding=(2,))\n",
       "          (4): Chomp1d()\n",
       "          (5): ReLU()\n",
       "        )\n",
       "        (relu): ReLU()\n",
       "      )\n",
       "      (1): TemporalBlock(\n",
       "        (conv1): Conv1dPathwise(100, 100, kernel_size=(3,), stride=(1,), padding=(4,), dilation=(2,))\n",
       "        (chomp1): Chomp1d()\n",
       "        (relu1): ReLU()\n",
       "        (conv2): Conv1dPathwise(100, 100, kernel_size=(3,), stride=(1,), padding=(4,), dilation=(2,))\n",
       "        (chomp2): Chomp1d()\n",
       "        (relu2): ReLU()\n",
       "        (net): Sequential(\n",
       "          (0): Conv1dPathwise(100, 100, kernel_size=(3,), stride=(1,), padding=(4,), dilation=(2,))\n",
       "          (1): Chomp1d()\n",
       "          (2): ReLU()\n",
       "          (3): Conv1dPathwise(100, 100, kernel_size=(3,), stride=(1,), padding=(4,), dilation=(2,))\n",
       "          (4): Chomp1d()\n",
       "          (5): ReLU()\n",
       "        )\n",
       "        (relu): ReLU()\n",
       "      )\n",
       "    )\n",
       "  )\n",
       "  (fc): LinearPathwise(in_features=100, out_features=2, bias=True)\n",
       "  (activation): LogSoftmax()\n",
       ")"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model = torch.load(saved_folder + 'best_tcn_variational_normal_14_val_acc=0.809633.pth')\n",
    "model.to(device)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(872,)\n",
      "(872, 2)\n",
      "(872,)\n",
      "(872,)\n"
     ]
    }
   ],
   "source": [
    "preds, scores, targets, variances = bbb_predict(model, iterator, T=10)\n",
    "kappas_sr = np.max(scores, 1)\n",
    "kappas_var = -variances"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "risks_sr = [selective_risk_at_coverage(c, kappas_sr, preds, targets) for c in coverages]\n",
    "risks_var = [selective_risk_at_coverage(c, kappas_var, preds, targets) for c in coverages]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot(coverages, risks_sr, risks_var, (10,6))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot(coverages, risks_sr, risks_var, (10,6))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## IMDb"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "91572\n",
      "\n"
     ]
    }
   ],
   "source": [
    "_, valid_data, _ = load_imdb()\n",
    "iterator = data.BucketIterator(valid_data, batch_size=batch_size, device=device)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['best_lr_lstm_4_val_acc=0.5366.pth',\n",
       " 'best_lr_tcn_variational_normal_26_val_acc=0.882.pth',\n",
       " 'best_lr_NO_drop_tcn_28_val_acc=0.8838.pth',\n",
       " 'best_lr_tcn_28_val_acc=0.8926.pth',\n",
       " 'best_NO_drop_tcn_3_val_acc=0.8974.pth',\n",
       " 'best_lstm_variational_normal_15_val_acc=0.8278.pth',\n",
       " 'best_tcn_3_val_acc=0.8962.pth',\n",
       " 'best_lstm_variational_laplace_15_val_acc=0.8278.pth',\n",
       " 'best_tcn_variational_laplace_4_val_acc=0.8904.pth',\n",
       " 'best_lstm_19_val_acc=0.861.pth',\n",
       " 'best_lr_tcn_variational_laplace_26_val_acc=0.882.pth',\n",
       " 'best_lr_NO_drop_lstm_21_val_acc=0.6754.pth',\n",
       " 'best_tcn_variational_normal_4_val_acc=0.8904.pth',\n",
       " 'best_lr_lstm_variational_normal_30_val_acc=0.7436.pth',\n",
       " 'best_lr_lstm_variational_laplace_30_val_acc=0.7436.pth',\n",
       " 'best_NO_drop_lstm_14_val_acc=0.8596.pth']"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "saved_folder = './models_final/imdb/'\n",
    "os.listdir(saved_folder)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['best_lr_tcn_variational_normal_26_val_acc=0.882.pth',\n",
       " 'best_tcn_variational_laplace_4_val_acc=0.8904.pth',\n",
       " 'best_lr_tcn_variational_laplace_26_val_acc=0.882.pth',\n",
       " 'best_tcn_variational_normal_4_val_acc=0.8904.pth']"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "name = 'tcn_variational'\n",
    "[l for l in os.listdir(saved_folder) if name in l]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "TextTCN(\n",
       "  (embedding): Embedding(91572, 100, padding_idx=1)\n",
       "  (tcn): TemporalConvNet(\n",
       "    (network): Sequential(\n",
       "      (0): TemporalBlock(\n",
       "        (conv1): Conv1dPathwise(100, 100, kernel_size=(3,), stride=(1,), padding=(2,))\n",
       "        (chomp1): Chomp1d()\n",
       "        (relu1): ReLU()\n",
       "        (conv2): Conv1dPathwise(100, 100, kernel_size=(3,), stride=(1,), padding=(2,))\n",
       "        (chomp2): Chomp1d()\n",
       "        (relu2): ReLU()\n",
       "        (net): Sequential(\n",
       "          (0): Conv1dPathwise(100, 100, kernel_size=(3,), stride=(1,), padding=(2,))\n",
       "          (1): Chomp1d()\n",
       "          (2): ReLU()\n",
       "          (3): Conv1dPathwise(100, 100, kernel_size=(3,), stride=(1,), padding=(2,))\n",
       "          (4): Chomp1d()\n",
       "          (5): ReLU()\n",
       "        )\n",
       "        (relu): ReLU()\n",
       "      )\n",
       "      (1): TemporalBlock(\n",
       "        (conv1): Conv1dPathwise(100, 100, kernel_size=(3,), stride=(1,), padding=(4,), dilation=(2,))\n",
       "        (chomp1): Chomp1d()\n",
       "        (relu1): ReLU()\n",
       "        (conv2): Conv1dPathwise(100, 100, kernel_size=(3,), stride=(1,), padding=(4,), dilation=(2,))\n",
       "        (chomp2): Chomp1d()\n",
       "        (relu2): ReLU()\n",
       "        (net): Sequential(\n",
       "          (0): Conv1dPathwise(100, 100, kernel_size=(3,), stride=(1,), padding=(4,), dilation=(2,))\n",
       "          (1): Chomp1d()\n",
       "          (2): ReLU()\n",
       "          (3): Conv1dPathwise(100, 100, kernel_size=(3,), stride=(1,), padding=(4,), dilation=(2,))\n",
       "          (4): Chomp1d()\n",
       "          (5): ReLU()\n",
       "        )\n",
       "        (relu): ReLU()\n",
       "      )\n",
       "    )\n",
       "  )\n",
       "  (fc): LinearPathwise(in_features=100, out_features=2, bias=True)\n",
       "  (activation): LogSoftmax()\n",
       ")"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model = torch.load(saved_folder + 'best_tcn_variational_normal_4_val_acc=0.8904.pth')\n",
    "model.to(device)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(5000,)\n",
      "(5000, 2)\n",
      "(5000,)\n",
      "(5000,)\n"
     ]
    }
   ],
   "source": [
    "preds, scores, targets, variances = bbb_predict(model, iterator, T=100)\n",
    "kappas_sr = np.max(scores, 1)\n",
    "kappas_var = -variances"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "risks_sr = [selective_risk_at_coverage(c, kappas_sr, preds, targets) for c in coverages]\n",
    "risks_var = [selective_risk_at_coverage(c, kappas_var, preds, targets) for c in coverages]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot(coverages, risks_sr, risks_var, (10,6))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot(coverages, risks_sr, risks_var, (10,6))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Yelp-f"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "_, valid_data, _ = load_yelp()\n",
    "iterator = data.BucketIterator(valid_data, batch_size=batch_size, device=device)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['best_tcn_variational_normal_1_val_acc=0.6227769.pth',\n",
       " 'best_lr_tcn_variational_laplace_8_val_acc=0.6158385.pth',\n",
       " 'best_lstm_variational_normal_2_val_acc=0.6295308.pth',\n",
       " 'best_lr_tcn_variational_normal_8_val_acc=0.6176769.pth',\n",
       " 'best_lstm_variational_laplace_2_val_acc=0.6319385.pth',\n",
       " 'best_lr_lstm_variational_normal_16_val_acc=0.6126.pth',\n",
       " 'best_lr_lstm_variational_laplace_15_val_acc=0.6104.pth',\n",
       " 'best_tcn_variational_laplace_2_val_acc=0.6246154.pth']"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "saved_folder = './models_final/yelp/'\n",
    "os.listdir(saved_folder)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = torch.load(saved_folder + 'best_lr_tcn_variational_normal_8_val_acc=0.6176769.pth')\n",
    "model.to(device)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "preds, scores, targets, variances = bbb_predict(model, iterator, T=10)\n",
    "kappas_sr = np.max(scores, 1)\n",
    "kappas_var = -variances"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "risks_sr = [selective_risk_at_coverage(c, kappas_sr, preds, targets) for c in coverages]\n",
    "risks_var = [selective_risk_at_coverage(c, kappas_var, preds, targets) for c in coverages]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "plt.plot(coverages, risks_sr, label='sr')\n",
    "plt.plot(coverages, risks_var, label='var')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "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": "3.6.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}