DQN-LSTM-SIF Embedding.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import time\n",
    "import pickle\n",
    "import numpy as np\n",
    "from collections import Counter\n",
    "\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "from matplotlib import pyplot as plt\n",
    "\n",
    "%matplotlib inline\n",
    "\n",
    "import torch\n",
    "import torch.autograd as autograd\n",
    "import torch.nn as nn\n",
    "import torch.nn.functional as F\n",
    "import torch.optim as optim"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "def rougeScores(genSummary, refSummary):\n",
    "    genTotal, refTotal, intersection = 0., 0., 0.\n",
    "    for token in list(set(list(refSummary.keys()) + list(genSummary.keys()) )):\n",
    "        intersection += min(refSummary[token], genSummary[token])\n",
    "        refTotal += refSummary[token]\n",
    "        genTotal += genSummary[token]\n",
    "\n",
    "    recall = intersection / refTotal if refTotal > 0. else 0.\n",
    "    prec   = intersection / genTotal if genTotal > 0. else 0.\n",
    "    f1 = (2. * recall * prec) / (recall + prec) if (recall + prec) > 0. else 0.\n",
    "    \n",
    "    return recall, prec, f1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "def buildPredSummary(df, summary, sentence_emb, curr_pred_emb, action, select_index, sent_index):\n",
    "    if action.select(1, select_index).tolist()[0] == 1:\n",
    "        sum_len =  curr_pred_emb.size()[0]\n",
    "        summary = summary + ' ' + df['sentence'][sent_index]\n",
    "        curr_pred_emb = torch.cat([curr_pred_emb[1:sum_len, :], sentence_emb], 0)\n",
    "        return curr_pred_emb, summary\n",
    "    \n",
    "    else:\n",
    "        return curr_pred_emb, summary"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "#sdf = pd.read_csv('/home/francisco/GitHub/DQN-Event-Summarization/data/sif/train_000_0.csv')\n",
    "#sdf = pd.read_csv('/Users/franciscojavierarceo/GitHub/DQN-Event-Summarization/data/sif/train_000_0.csv')\n",
    "#true_summary = sdf['summary'][0]\n",
    "#ts_tokenized = Counter(true_summary.split(\" \"))\n",
    "\n",
    "res = pickle.load(open('/home/francisco/GitHub/DQN-Event-Summarization/data/testsif/sifquerydict_0.pkl', 'rb'))\n",
    "sdf = res['data']\n",
    "true_summary = res['embeddings']['true_summary']\n",
    "ts_tokenized = Counter(true_summary.split(\" \"))\n",
    "nsentences = sdf.index.max()\n",
    "\n",
    "# Initializing stuff\n",
    "SKIP = 0\n",
    "SELECT = 1\n",
    "\n",
    "nepochs = 500\n",
    "lrate = 0.01\n",
    "momentum_rate = 0.8\n",
    "\n",
    "outputdim = 2\n",
    "batchsize = 1\n",
    "sent_size = 1\n",
    "sif_emb_d = 300\n",
    "nhiddenlayers = 1\n",
    "nhiddenunits = 200\n",
    "predsummary_size = 2\n",
    "\n",
    "rand_rate = 1.0\n",
    "decay_rate = 0.05\n",
    "\n",
    "criterion = nn.MSELoss()\n",
    "\n",
    "lossf = {\n",
    "    'loss': [], \n",
    "    'sent_idx': [], \n",
    "    'epoch': [], \n",
    "    'action': [], \n",
    "    'f1': [], \n",
    "    'recall': [],\n",
    "    'precision': [],\n",
    "    'optpred': [], \n",
    "    'rouge_delta':[]\n",
    "}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 108,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'The': 0, 'dog': 1, 'ate': 2, 'the': 3, 'apple': 4, 'Everybody': 5, 'read': 6, 'that': 7, 'book': 8}\n"
     ]
    }
   ],
   "source": [
    "def prepare_sequence(seq, to_ix):\n",
    "    idxs = [to_ix[w] for w in seq]\n",
    "    return torch.tensor(idxs, dtype=torch.long)\n",
    "\n",
    "\n",
    "training_data = [\n",
    "    (\"The dog ate the apple\".split(), [\"DET\", \"NN\", \"V\", \"DET\", \"NN\"]),\n",
    "    (\"Everybody read that book\".split(), [\"NN\", \"V\", \"DET\", \"NN\"])\n",
    "]\n",
    "word_to_ix = {}\n",
    "for sent, tags in training_data:\n",
    "    for word in sent:\n",
    "        if word not in word_to_ix:\n",
    "            word_to_ix[word] = len(word_to_ix)\n",
    "\n",
    "print(word_to_ix)\n",
    "tag_to_ix = {\"DET\": 0, \"NN\": 1, \"V\": 2}\n",
    "\n",
    "# These will usually be more like 32 or 64 dimensional.\n",
    "# We will keep them small, so we can see how the weights change as we train.\n",
    "EMBEDDING_DIM = 6\n",
    "HIDDEN_DIM = 6"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 109,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tensor([[-1.0658, -0.7839, -1.6148],\n",
      "        [-1.1096, -0.7908, -1.5287],\n",
      "        [-1.1140, -0.8332, -1.4392],\n",
      "        [-1.1054, -0.8327, -1.4522],\n",
      "        [-1.1459, -0.7619, -1.5357]])\n"
     ]
    }
   ],
   "source": [
    "model = LSTMTagger(EMBEDDING_DIM, HIDDEN_DIM, len(word_to_ix), len(tag_to_ix))\n",
    "loss_function = nn.NLLLoss()\n",
    "optimizer = optim.SGD(model.parameters(), lr=0.1)\n",
    "\n",
    "# See what the scores are before training\n",
    "# Note that element i,j of the output is the score for tag j for word i.\n",
    "# Here we don't need to train, so the code is wrapped in torch.no_grad()\n",
    "\n",
    "with torch.no_grad():\n",
    "    inputs = prepare_sequence(training_data[0][0], word_to_ix)\n",
    "    tag_scores = model(inputs)\n",
    "    print(tag_scores)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 198,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[[0.0445, 0.2234]]], grad_fn=<CatBackward>)"
      ]
     },
     "execution_count": 198,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "inputsv2 = torch.tensor(np.random.rand(6,), dtype=torch.float)\n",
    "temp_lstm = nn.LSTM(6, 2)\n",
    "\n",
    "o, h = temp_lstm(inputsv2.view(1, 1, -1))\n",
    "\n",
    "o"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 230,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[[0., 0., 0.]]], grad_fn=<LogSoftmaxBackward>)"
      ]
     },
     "execution_count": 230,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "class LSTMTagger(nn.Module):\n",
    "    def __init__(self, embedding_dim, hidden_dim, vocab_size, tagset_size):\n",
    "        super(LSTMTagger, self).__init__()\n",
    "        self.hidden_dim = hidden_dim\n",
    "        self.word_embeddings = nn.Embedding(vocab_size, embedding_dim)\n",
    "        self.lstm = nn.LSTM(embedding_dim, hidden_dim)\n",
    "        self.hidden2tag = nn.Linear(hidden_dim, tagset_size)\n",
    "        self.hidden = self.init_hidden()\n",
    "\n",
    "    def init_hidden(self):\n",
    "        return (torch.zeros(1, 1, self.hidden_dim), torch.zeros(1, 1, self.hidden_dim))\n",
    "\n",
    "    def forward(self, sentence):\n",
    "        embeds = self.word_embeddings(sentence)\n",
    "        lstm_out, self.hidden = self.lstm(embeds.view(len(sentence), 1, -1), self.hidden)\n",
    "        tag_space = self.hidden2tag(lstm_out.view(len(sentence), -1))\n",
    "        tag_scores = F.log_softmax(tag_space, dim=1)\n",
    "        return embeds, tag_space, tag_scores\n",
    "\n",
    "class LSTMTaggerv2(nn.Module):\n",
    "    def __init__(self, embedding_dim, hidden_dim, vocab_size, tagset_size):\n",
    "        super(LSTMTaggerv2, self).__init__()\n",
    "        self.hidden_dim = hidden_dim\n",
    "        #self.word_embeddings = nn.Embedding(vocab_size, embedding_dim)\n",
    "        # The LSTM takes word embeddings as inputs, and outputs hidden states\n",
    "        # with dimensionality hidden_dim\n",
    "        self.lstm = nn.LSTM(embedding_dim, hidden_dim)\n",
    "        # The linear layer that maps from hidden state space to tag space\n",
    "        self.hidden2tag = nn.Linear(hidden_dim, tagset_size)\n",
    "        self.hiddenOther = nn.Linear(hidden_dim, tagset_size)\n",
    "        self.hidden = self.init_hidden()\n",
    "\n",
    "    def init_hidden(self):\n",
    "        # Before we've done anything, we dont have any hidden state.\n",
    "        # Refer to the Pytorch documentation to see exactly why they have this dimensionality.\n",
    "        # The axes semantics are (num_layers, minibatch_size, hidden_dim)\n",
    "        return (torch.zeros(1, 1, self.hidden_dim), torch.zeros(1, 1, self.hidden_dim))\n",
    "\n",
    "    def forward(self, sentence):\n",
    "        #embeds = self.word_embeddings(sentence)\n",
    "        lstm_out, self.hidden = self.lstm(sentence.view(1, 1, -1), self.hidden)\n",
    "        #lstm_out, self.hidden = self.lstm(sentence.view(len(sentence), 1, -1), self.hidden)\n",
    "        #tag_space = self.hidden2tag(lstm_out.view(len(sentence), -1))\n",
    "        tag_space = self.hidden2tag(lstm_out)\n",
    "        tag_scores = F.log_softmax(tag_space, dim=1)\n",
    "        return tag_scores\n",
    "\n",
    "model = LSTMTagger(EMBEDDING_DIM, HIDDEN_DIM, len(word_to_ix), len(tag_to_ix))\n",
    "modelv2 = LSTMTaggerv2(EMBEDDING_DIM, HIDDEN_DIM, len(word_to_ix), len(tag_to_ix))\n",
    "\n",
    "modelv2(inputsv2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 231,
   "metadata": {},
   "outputs": [],
   "source": [
    "inputs_len6 = torch.tensor([0, 1, 2, 3, 4, 5], dtype=torch.long)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 232,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "torch.Size([6, 6])"
      ]
     },
     "execution_count": 232,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model(inputs_len6)[0].shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 233,
   "metadata": {},
   "outputs": [],
   "source": [
    "embeds_tmp, tag_space_tmp, tag_scores_tmp = model(inputs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 222,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(torch.Size([5, 6]), torch.Size([5, 3]), torch.Size([5, 3]))"
      ]
     },
     "execution_count": 222,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "embeds_tmp.shape, tag_space_tmp.shape, tag_scores_tmp.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 223,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([0, 1, 2, 3, 4])"
      ]
     },
     "execution_count": 223,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "inputs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 234,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(tensor([[-0.3333, -1.0985,  0.0903, -1.6879,  0.7927,  2.2558],\n",
       "         [ 0.6505,  0.2082,  0.4561,  0.7229,  0.1207,  0.0498],\n",
       "         [ 0.2246, -0.9083,  0.2103, -1.3785, -1.9157,  0.2116],\n",
       "         [-0.9674,  1.5726,  0.7261,  1.0171,  0.8188,  0.5618],\n",
       "         [ 0.7490,  0.9222, -0.2654, -1.6319, -0.9896, -0.2716]],\n",
       "        grad_fn=<EmbeddingBackward>), torch.Size([5, 6]))"
      ]
     },
     "execution_count": 234,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# 5 records on the bottom\n",
    "embeds_tmp, embeds_tmp.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 214,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[-0.1110, -0.3197,  0.2363],\n",
       "        [-0.1899, -0.2857,  0.2853],\n",
       "        [-0.2019, -0.2547,  0.3657],\n",
       "        [-0.1311, -0.2109,  0.2809],\n",
       "        [-0.1281, -0.2692,  0.2805]], grad_fn=<ThAddmmBackward>)"
      ]
     },
     "execution_count": 214,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tag_space_tmp"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 215,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[-1.1715, -1.3802, -0.8242],\n",
       "        [-1.2576, -1.3533, -0.7824],\n",
       "        [-1.3117, -1.3645, -0.7442],\n",
       "        [-1.2335, -1.3133, -0.8214],\n",
       "        [-1.2158, -1.3569, -0.8072]], grad_fn=<LogSoftmaxBackward>)"
      ]
     },
     "execution_count": 215,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tag_scores_tmp"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 204,
   "metadata": {},
   "outputs": [],
   "source": [
    "temp = model(inputs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 225,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(torch.Size([5, 6]), torch.Size([5, 3]), torch.Size([5, 3]))"
      ]
     },
     "execution_count": 225,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "temp[0].shape, temp[1].shape, temp[2].shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 226,
   "metadata": {},
   "outputs": [],
   "source": [
    "class simpleLSTM(nn.Module):  # inheriting from nn.Module!\n",
    "    # calls the init function of nn.Module.  Dont get confused by syntax, always do it in an nn.Module\n",
    "    def __init__(self, input_size, outputsize, nunits, nlayers, sent_size, batch_size):        \n",
    "        super(simpleLSTM, self).__init__()\n",
    "        self.input_size = input_size\n",
    "        self.nunits = nunits\n",
    "        self.sent_size = sent_size\n",
    "        self.batch_size = batch_size\n",
    "        self.lstm0 = nn.LSTM(input_size, nunits)\n",
    "        self.outputlayer = nn.Linear(nunits, outputsize)\n",
    "        self.hidden = self.init_hidden()\n",
    "\n",
    "    def init_hidden(self):\n",
    "        return (autograd.Variable(torch.zeros(1, self.batch_size, self.nunits)))\n",
    "\n",
    "    def forward(self, sent_embed):\n",
    "        lstm_out, self.hidden = self.lstm0(sent_embed.view(self.batch_size, self.input_size, -1), self.hidden)\n",
    "        #summary_hl = self.linearlayer1(lstm_out0[-1].view(self.sent_size, -1))\n",
    "        return lstm_out #self.outputlayer(summary_hl)\n",
    "        #return self.outputlayer(summary_hl)\n",
    "\n",
    "model = simpleLSTM(input_size=sif_emb_d, outputsize=2, nunits=nhiddenunits,\n",
    "                      nlayers=nhiddenlayers, sent_size=(nsentences+1), batch_size=batchsize)\n",
    "\n",
    "# swap to ADAM\n",
    "optimizer = optim.Adam(model.parameters(), lr=lrate)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 227,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2"
      ]
     },
     "execution_count": 227,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "predsummary_size"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 228,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "torch.Size([122, 300])"
      ]
     },
     "execution_count": 228,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "autograd.Variable(sent_emb).shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 229,
   "metadata": {},
   "outputs": [
    {
     "ename": "RuntimeError",
     "evalue": "input.size(-1) must be equal to input_size. Expected 300, got 122",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mRuntimeError\u001b[0m                              Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-229-dce2f7eaeff7>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mrouge_preds\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mmodel\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mautograd\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mVariable\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0msent_emb\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[0;32m~/anaconda3/lib/python3.6/site-packages/torch/nn/modules/module.py\u001b[0m in \u001b[0;36m__call__\u001b[0;34m(self, *input, **kwargs)\u001b[0m\n\u001b[1;32m    475\u001b[0m             \u001b[0mresult\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_slow_forward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    476\u001b[0m         \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 477\u001b[0;31m             \u001b[0mresult\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mforward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    478\u001b[0m         \u001b[0;32mfor\u001b[0m \u001b[0mhook\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_forward_hooks\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mvalues\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    479\u001b[0m             \u001b[0mhook_result\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mhook\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mresult\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m<ipython-input-226-5593aace0968>\u001b[0m in \u001b[0;36mforward\u001b[0;34m(self, sent_embed)\u001b[0m\n\u001b[1;32m     15\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     16\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0mforward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0msent_embed\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 17\u001b[0;31m         \u001b[0mlstm_out\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mhidden\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlstm0\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0msent_embed\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mview\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbatch_size\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0minput_size\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m-\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mhidden\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     18\u001b[0m         \u001b[0;31m#summary_hl = self.linearlayer1(lstm_out0[-1].view(self.sent_size, -1))\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     19\u001b[0m         \u001b[0;32mreturn\u001b[0m \u001b[0mlstm_out\u001b[0m \u001b[0;31m#self.outputlayer(summary_hl)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/anaconda3/lib/python3.6/site-packages/torch/nn/modules/module.py\u001b[0m in \u001b[0;36m__call__\u001b[0;34m(self, *input, **kwargs)\u001b[0m\n\u001b[1;32m    475\u001b[0m             \u001b[0mresult\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_slow_forward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    476\u001b[0m         \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 477\u001b[0;31m             \u001b[0mresult\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mforward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    478\u001b[0m         \u001b[0;32mfor\u001b[0m \u001b[0mhook\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_forward_hooks\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mvalues\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    479\u001b[0m             \u001b[0mhook_result\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mhook\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mresult\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/anaconda3/lib/python3.6/site-packages/torch/nn/modules/rnn.py\u001b[0m in \u001b[0;36mforward\u001b[0;34m(self, input, hx)\u001b[0m\n\u001b[1;32m    176\u001b[0m             \u001b[0mflat_weight\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    177\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 178\u001b[0;31m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcheck_forward_args\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mhx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mbatch_sizes\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    179\u001b[0m         func = self._backend.RNN(\n\u001b[1;32m    180\u001b[0m             \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmode\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/anaconda3/lib/python3.6/site-packages/torch/nn/modules/rnn.py\u001b[0m in \u001b[0;36mcheck_forward_args\u001b[0;34m(self, input, hidden, batch_sizes)\u001b[0m\n\u001b[1;32m    128\u001b[0m             raise RuntimeError(\n\u001b[1;32m    129\u001b[0m                 'input.size(-1) must be equal to input_size. Expected {}, got {}'.format(\n\u001b[0;32m--> 130\u001b[0;31m                     self.input_size, input.size(-1)))\n\u001b[0m\u001b[1;32m    131\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    132\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0mis_input_packed\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mRuntimeError\u001b[0m: input.size(-1) must be equal to input_size. Expected 300, got 122"
     ]
    }
   ],
   "source": [
    "rouge_preds = model(autograd.Variable(sent_emb))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "ename": "RuntimeError",
     "evalue": "Expected hidden[0] size (1, 1, 200), got (1, 200)",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mRuntimeError\u001b[0m                              Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-42-8b342bb730f0>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m     42\u001b[0m         \u001b[0mautograd\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mVariable\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mquery_emb\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     43\u001b[0m         \u001b[0mautograd\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mVariable\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0msent_emb\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 44\u001b[0;31m         \u001b[0mautograd\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mVariable\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mpredsummary_emb\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;31m#.view(predsummary_size, batchsize, -1))\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     45\u001b[0m )\n",
      "\u001b[0;32m~/anaconda3/lib/python3.6/site-packages/torch/nn/modules/module.py\u001b[0m in \u001b[0;36m__call__\u001b[0;34m(self, *input, **kwargs)\u001b[0m\n\u001b[1;32m    475\u001b[0m             \u001b[0mresult\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_slow_forward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    476\u001b[0m         \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 477\u001b[0;31m             \u001b[0mresult\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mforward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    478\u001b[0m         \u001b[0;32mfor\u001b[0m \u001b[0mhook\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_forward_hooks\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mvalues\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    479\u001b[0m             \u001b[0mhook_result\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mhook\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mresult\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m<ipython-input-42-8b342bb730f0>\u001b[0m in \u001b[0;36mforward\u001b[0;34m(self, query_embed, sent_embed, summ_embed)\u001b[0m\n\u001b[1;32m     26\u001b[0m         \u001b[0mquery_hl\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlinearlayer0\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mquery_embed\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     27\u001b[0m         \u001b[0msentence_hl\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlinearlayer0\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0msent_embed\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 28\u001b[0;31m         \u001b[0mlstm_out\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mhidden\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlstm\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0msumm_embed\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mview\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msent_size\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbatch_size\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m-\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mhidden\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     29\u001b[0m         \u001b[0msummary_hl\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlinearlayer1\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mlstm_out\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m-\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mview\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msent_size\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m-\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     30\u001b[0m         \u001b[0;31m# Concatenates the layers into (nunit * 3, 2)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/anaconda3/lib/python3.6/site-packages/torch/nn/modules/module.py\u001b[0m in \u001b[0;36m__call__\u001b[0;34m(self, *input, **kwargs)\u001b[0m\n\u001b[1;32m    475\u001b[0m             \u001b[0mresult\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_slow_forward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    476\u001b[0m         \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 477\u001b[0;31m             \u001b[0mresult\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mforward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    478\u001b[0m         \u001b[0;32mfor\u001b[0m \u001b[0mhook\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_forward_hooks\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mvalues\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    479\u001b[0m             \u001b[0mhook_result\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mhook\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mresult\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/anaconda3/lib/python3.6/site-packages/torch/nn/modules/rnn.py\u001b[0m in \u001b[0;36mforward\u001b[0;34m(self, input, hx)\u001b[0m\n\u001b[1;32m    176\u001b[0m             \u001b[0mflat_weight\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    177\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 178\u001b[0;31m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcheck_forward_args\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mhx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mbatch_sizes\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    179\u001b[0m         func = self._backend.RNN(\n\u001b[1;32m    180\u001b[0m             \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmode\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/anaconda3/lib/python3.6/site-packages/torch/nn/modules/rnn.py\u001b[0m in \u001b[0;36mcheck_forward_args\u001b[0;34m(self, input, hidden, batch_sizes)\u001b[0m\n\u001b[1;32m    145\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmode\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;34m'LSTM'\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    146\u001b[0m             check_hidden_size(hidden[0], expected_hidden_size,\n\u001b[0;32m--> 147\u001b[0;31m                               'Expected hidden[0] size {}, got {}')\n\u001b[0m\u001b[1;32m    148\u001b[0m             check_hidden_size(hidden[1], expected_hidden_size,\n\u001b[1;32m    149\u001b[0m                               'Expected hidden[1] size {}, got {}')\n",
      "\u001b[0;32m~/anaconda3/lib/python3.6/site-packages/torch/nn/modules/rnn.py\u001b[0m in \u001b[0;36mcheck_hidden_size\u001b[0;34m(hx, expected_hidden_size, msg)\u001b[0m\n\u001b[1;32m    141\u001b[0m         \u001b[0;32mdef\u001b[0m \u001b[0mcheck_hidden_size\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mhx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mexpected_hidden_size\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmsg\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m'Expected hidden size {}, got {}'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    142\u001b[0m             \u001b[0;32mif\u001b[0m \u001b[0mtuple\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mhx\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msize\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m!=\u001b[0m \u001b[0mexpected_hidden_size\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 143\u001b[0;31m                 \u001b[0;32mraise\u001b[0m \u001b[0mRuntimeError\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmsg\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mformat\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mexpected_hidden_size\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtuple\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mhx\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msize\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    144\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    145\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmode\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;34m'LSTM'\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mRuntimeError\u001b[0m: Expected hidden[0] size (1, 1, 200), got (1, 200)"
     ]
    }
   ],
   "source": [
    "class lstmRegressor(nn.Module):  # inheriting from nn.Module!\n",
    "    # calls the init function of nn.Module.  Dont get confused by syntax, always do it in an nn.Module\n",
    "    def __init__(self, input_size, outputsize, nunits, nlayers, sent_size, batch_size):        \n",
    "        super(lstmRegressor, self).__init__()\n",
    "        self.input_size = input_size\n",
    "        self.nunits = nunits\n",
    "        self.nlayers = nlayers\n",
    "        self.sent_size = sent_size\n",
    "        self.batch_size = batch_size\n",
    "        #self.lstmlayer = nn.LSTM(nunits, nunits, num_layers=self.nlayers)\n",
    "        self.lstm0 = nn.LSTM(input_size, nunits)\n",
    "        self.linearlayer0 = nn.Linear(input_size, nunits)\n",
    "        self.linearlayer1 = nn.Linear(nunits, nunits)\n",
    "        self.dense1_bn = nn.BatchNorm1d(nunits * 3)\n",
    "        self.outputlayer = nn.Linear(nunits * 2, outputsize)\n",
    "        self.hidden = self.init_hidden()\n",
    "\n",
    "    def init_hidden(self):\n",
    "        # Before we've done anything, we dont have any hidden state.\n",
    "        # Refer to the Pytorch documentation to see exactly\n",
    "        # why they have this dimensionality.\n",
    "        # The axes semantics are (num_layers, minibatch_size, hidden_dim)\n",
    "        return (autograd.Variable(torch.zeros(1, self.batch_size, self.nunits)))\n",
    "\n",
    "    def forward(self, query_embed, sent_embed, summ_embed):\n",
    "        query_hl = self.linearlayer0(query_embed)\n",
    "        sentence_hl = self.linearlayer0(sent_embed)\n",
    "        lstm_out, self.hidden = self.lstm0(sent_embed.view(sent_embed, self.batch_size, -1), self.hidden)\n",
    "        lstm_out0, self.hidden = self.lstm0(summ_embed.view(self.sent_size, self.batch_size, -1), self.hidden)\n",
    "        summary_hl = self.linearlayer1(lstm_out0[-1].view(self.sent_size, -1))\n",
    "        # Concatenates the layers into (nunit * 3, 2)\n",
    "        catlayer = torch.cat((query_hl, sentence_hl, summary_hl), 1)\n",
    "        return self.outputlayer(catlayer)\n",
    "        #return self.outputlayer(self.dense1_bn(catlayer))\n",
    "\n",
    "model = lstmRegressor(input_size=sif_emb_d, outputsize=2, nunits=nhiddenunits,\n",
    "                      nlayers=nhiddenlayers, sent_size=predsummary_size, batch_size=batchsize)\n",
    "\n",
    "# swap to ADAM\n",
    "optimizer = optim.Adam(model.parameters(), lr=lrate)\n",
    "\n",
    "rouge_preds = model(\n",
    "        autograd.Variable(query_emb),\n",
    "        autograd.Variable(sent_emb),\n",
    "        autograd.Variable(predsummary_emb)#.view(predsummary_size, batchsize, -1))\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Single Iteration"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "dict_keys(['query_embeddings', 'sentence_embeddings', 'true_summary'])"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "res['embeddings']['query_embeddings']\n",
    "res['embeddings']['sentence_embeddings']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "predsummary_size"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "f1_t0 = 0.\n",
    "curr_summary = ''\n",
    "predsummary_emb = torch.from_numpy(np.zeros((predsummary_size, sif_emb_d))).float()\n",
    "sent_index = 0\n",
    "#for sent_index in range(nsentences):        \n",
    "model.zero_grad()\n",
    "\n",
    "# The embeddings start on the 5th column (index 4)\n",
    "sent_emb = torch.FloatTensor(\n",
    "    #sdf[sdf.columns[4:]].values[sent_index, :].reshape(1, sif_emb_d)\n",
    "    res['embeddings']['sentence_embeddings'].values.reshape(nsentences + 1, sif_emb_d)\n",
    ")\n",
    "\n",
    "query_emb = torch.FloatTensor(\n",
    "    res['embeddings']['query_embeddings'].values.reshape(1, sif_emb_d)\n",
    ")\n",
    "\n",
    "train_ys = torch.from_numpy(np.asarray([0]).reshape(1, 1)).float()\n",
    "action = torch.from_numpy(np.asarray([0, 0]).reshape(1, 2)).int()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "ename": "RuntimeError",
     "evalue": "Expected hidden[0] size (1, 1, 300), got (2, 1, 300)",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mRuntimeError\u001b[0m                              Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-21-d2db46bed57e>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      2\u001b[0m         \u001b[0mautograd\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mVariable\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mquery_emb\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      3\u001b[0m         \u001b[0mautograd\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mVariable\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0msent_emb\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 4\u001b[0;31m         \u001b[0mautograd\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mVariable\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mpredsummary_emb\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mview\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mpredsummary_size\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mbatchsize\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m-\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      5\u001b[0m )\n",
      "\u001b[0;32m~/anaconda3/lib/python3.6/site-packages/torch/nn/modules/module.py\u001b[0m in \u001b[0;36m__call__\u001b[0;34m(self, *input, **kwargs)\u001b[0m\n\u001b[1;32m    475\u001b[0m             \u001b[0mresult\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_slow_forward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    476\u001b[0m         \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 477\u001b[0;31m             \u001b[0mresult\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mforward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    478\u001b[0m         \u001b[0;32mfor\u001b[0m \u001b[0mhook\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_forward_hooks\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mvalues\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    479\u001b[0m             \u001b[0mhook_result\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mhook\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mresult\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m<ipython-input-3-2e6f4a526e2b>\u001b[0m in \u001b[0;36mforward\u001b[0;34m(self, query_embed, sent_embed, summ_embed)\u001b[0m\n\u001b[1;32m     28\u001b[0m         \u001b[0mquery_hl\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlinearlayer\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mquery_embed\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     29\u001b[0m         \u001b[0msentence_hl\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlinearlayer\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0msent_embed\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 30\u001b[0;31m         \u001b[0mout\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mhiddenlayer\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlstmlayer\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0msumm_embed\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mhidden\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     31\u001b[0m         \u001b[0msummary_hl\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlinearlayer\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mout\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m-\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     32\u001b[0m         \u001b[0;31m# Concatenates the layers into (900, 2)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/anaconda3/lib/python3.6/site-packages/torch/nn/modules/module.py\u001b[0m in \u001b[0;36m__call__\u001b[0;34m(self, *input, **kwargs)\u001b[0m\n\u001b[1;32m    475\u001b[0m             \u001b[0mresult\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_slow_forward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    476\u001b[0m         \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 477\u001b[0;31m             \u001b[0mresult\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mforward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    478\u001b[0m         \u001b[0;32mfor\u001b[0m \u001b[0mhook\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_forward_hooks\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mvalues\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    479\u001b[0m             \u001b[0mhook_result\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mhook\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mresult\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/anaconda3/lib/python3.6/site-packages/torch/nn/modules/rnn.py\u001b[0m in \u001b[0;36mforward\u001b[0;34m(self, input, hx)\u001b[0m\n\u001b[1;32m    176\u001b[0m             \u001b[0mflat_weight\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    177\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 178\u001b[0;31m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcheck_forward_args\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mhx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mbatch_sizes\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    179\u001b[0m         func = self._backend.RNN(\n\u001b[1;32m    180\u001b[0m             \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmode\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/anaconda3/lib/python3.6/site-packages/torch/nn/modules/rnn.py\u001b[0m in \u001b[0;36mcheck_forward_args\u001b[0;34m(self, input, hidden, batch_sizes)\u001b[0m\n\u001b[1;32m    145\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmode\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;34m'LSTM'\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    146\u001b[0m             check_hidden_size(hidden[0], expected_hidden_size,\n\u001b[0;32m--> 147\u001b[0;31m                               'Expected hidden[0] size {}, got {}')\n\u001b[0m\u001b[1;32m    148\u001b[0m             check_hidden_size(hidden[1], expected_hidden_size,\n\u001b[1;32m    149\u001b[0m                               'Expected hidden[1] size {}, got {}')\n",
      "\u001b[0;32m~/anaconda3/lib/python3.6/site-packages/torch/nn/modules/rnn.py\u001b[0m in \u001b[0;36mcheck_hidden_size\u001b[0;34m(hx, expected_hidden_size, msg)\u001b[0m\n\u001b[1;32m    141\u001b[0m         \u001b[0;32mdef\u001b[0m \u001b[0mcheck_hidden_size\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mhx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mexpected_hidden_size\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmsg\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m'Expected hidden size {}, got {}'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    142\u001b[0m             \u001b[0;32mif\u001b[0m \u001b[0mtuple\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mhx\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msize\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m!=\u001b[0m \u001b[0mexpected_hidden_size\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 143\u001b[0;31m                 \u001b[0;32mraise\u001b[0m \u001b[0mRuntimeError\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmsg\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mformat\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mexpected_hidden_size\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtuple\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mhx\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msize\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    144\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    145\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmode\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;34m'LSTM'\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mRuntimeError\u001b[0m: Expected hidden[0] size (1, 1, 300), got (2, 1, 300)"
     ]
    }
   ],
   "source": [
    "rouge_preds = model(\n",
    "        autograd.Variable(query_emb),\n",
    "        autograd.Variable(sent_emb),\n",
    "        autograd.Variable(predsummary_emb.view(predsummary_size, batchsize, -1))\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "qMax, qIndx = rouge_preds.max(dim=1)\n",
    "\n",
    "if np.random.uniform() < rand_rate and rand_rate > 0:\n",
    "    # Randomly choosing either 0 or 1 some percent of the time\n",
    "    qIndx = np.random.randint(0, 2, 1)[0]\n",
    "\n",
    "action[:, qIndx.data[0]] = 1\n",
    "action[:, abs(qIndx.data[0] - 1)] = 0\n",
    "\n",
    "predQonActions = torch.masked_select(rouge_preds, autograd.Variable(action.byte()))\n",
    "# building the summary and capturing the embedding\n",
    "# without a history model doesn't make a lot of sense\n",
    "# not clear what's happening...\n",
    "# concatenate summary embedding to input or try separate joining layer like before\n",
    "# Worth looking at rougue of each sentence...calculate f1 for each sentence to find out\n",
    "# might help figure out what's going on...\n",
    "predsummary_emb, curr_summary = buildPredSummary(\n",
    "                                     sdf, \n",
    "                                     curr_summary, \n",
    "                                     sent_emb, \n",
    "                                     predsummary_emb, \n",
    "                                     action, \n",
    "                                     SELECT, \n",
    "                                     sent_index\n",
    "                                )\n",
    "recall, prec, f1 = rougeScores(ts_tokenized, Counter(curr_summary.split(\" \")))\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "t0 = time.clock()\n",
    "for epoch in range(nepochs):\n",
    "    # reset embeddings and summary at the start of training\n",
    "    f1_t0 = 0.\n",
    "    curr_summary = ''\n",
    "    predsummary_emb = torch.from_numpy(np.zeros((predsummary_size, sif_emb_d))).float()\n",
    "    \n",
    "    for sent_index in range(nsentences):        \n",
    "        model.zero_grad()\n",
    "        \n",
    "        # The embeddings start on the 5th column (index 4)\n",
    "        sent_emb = torch.FloatTensor(\n",
    "            sdf[sdf.columns[4:]].values[sent_index, :].reshape(1, sif_emb_d)\n",
    "        )\n",
    "\n",
    "        train_ys = torch.from_numpy(np.asarray([0]).reshape(1, 1)).float()\n",
    "        action = torch.from_numpy(np.asarray([0, 0]).reshape(1,2)).int()\n",
    "\n",
    "        rouge_preds = model(\n",
    "                autograd.Variable(sent_emb), \n",
    "                autograd.Variable(predsummary_emb.view(predsummary_size, batchsize, -1)) \n",
    "        )\n",
    "\n",
    "        qMax, qIndx = rouge_preds.max(dim=1)\n",
    "\n",
    "        if np.random.uniform() < rand_rate and rand_rate > 0:\n",
    "            # Randomly choosing either 0 or 1 some percent of the time\n",
    "            qIndx = np.random.randint(0, 2, 1)[0]\n",
    "\n",
    "        action[:, qIndx.data[0]] = 1\n",
    "        action[:, abs(qIndx.data[0] - 1)] = 0\n",
    "        \n",
    "        predQonActions = torch.masked_select(rouge_preds, autograd.Variable(action.byte()))\n",
    "        # building the summary and capturing the embedding\n",
    "        # without a history model doesn't make a lot of sense\n",
    "        # not clear what's happening...\n",
    "        # concatenate summary embedding to input or try separate joining layer like before\n",
    "        # Worth looking at rougue of each sentence...calculate f1 for each sentence to find out\n",
    "        # might help figure out what's going on...\n",
    "        predsummary_emb, curr_summary = buildPredSummary(\n",
    "                                             sdf, \n",
    "                                             curr_summary, \n",
    "                                             sent_emb, \n",
    "                                             predsummary_emb, \n",
    "                                             action, \n",
    "                                             SELECT, \n",
    "                                             sent_index\n",
    "                                        )\n",
    "        recall, prec, f1 = rougeScores(ts_tokenized, Counter(curr_summary.split(\" \")))\n",
    "        \n",
    "        # Backward part\n",
    "        rouge_delta = f1 - f1_t0\n",
    "        train_ys[0] = rouge_delta\n",
    "        loss = criterion(predQonActions, autograd.Variable(train_ys))\n",
    "        \n",
    "        loss.backward()\n",
    "        optimizer.step()\n",
    "        \n",
    "        lossf['loss'].append(loss.data[0])\n",
    "        lossf['sent_idx'].append(sent_index)\n",
    "        lossf['epoch'].append(epoch)\n",
    "        lossf['action'].append(qIndx.data[0])\n",
    "        lossf['f1'].append(f1)\n",
    "        lossf['recall'].append(recall)\n",
    "        lossf['precision'].append(prec)\n",
    "        lossf['optpred'].append(predQonActions.data[0])\n",
    "        lossf['rouge_delta'].append(rouge_delta)\n",
    "\n",
    "        # Storing last round\n",
    "        f1_t0 = f1\n",
    "        \n",
    "    if rand_rate > 0:\n",
    "        rand_rate -= decay_rate\n",
    "        \n",
    "    if (epoch % 100) == 0:\n",
    "        print('epoch %i; f1 = %.3f' % (epoch, f1))\n",
    "\n",
    "print(\"training complete...runtime = %.2f minutes\" %  ( (time.clock() - t0)/(60. * 60)) ) \n",
    "\n",
    "# Making the performance data a dataframe\n",
    "perf = pd.DataFrame(lossf)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "perf[perf['sent_idx'] == perf['sent_idx'].max()].plot(\n",
    "    x='epoch', y='loss', c='red',\n",
    "    grid=True,\n",
    "    figsize=(12, 6),\n",
    "    title='Loss value across training'\n",
    ")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "perf[perf['sent_idx'] == perf['sent_idx'].max()].plot(\n",
    "    x='epoch', y=['f1', 'precision', 'recall'],\n",
    "    grid=True,\n",
    "    figsize=(12, 6),\n",
    "    ylim=[0, 1],\n",
    "    title='f1-score across training'\n",
    ")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "lead3 = ' '.join(sdf['sentence'][0:3])\n",
    "\n",
    "finalsummary = rougeScores(ts_tokenized, Counter(curr_summary.split(\" \")))\n",
    "baseline = rougeScores(ts_tokenized, Counter(lead3.split(\" \")))\n",
    "\n",
    "print(\"lead-3  recall = %.3f; precision = %.3f; f1-score = %.3f \" % (baseline[0], baseline[1], baseline[2]))\n",
    "\n",
    "print(\"learned recall = %.3f; precision = %.3f; f1-score = %.3f \" % (finalsummary[0], finalsummary[1], finalsummary[2]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "lead3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "true_summary"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "curr_summary"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "sdf.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# END"
   ]
  }
 ],
 "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": "3.6.3"
  },
  "latex_envs": {
   "LaTeX_envs_menu_present": true,
   "autocomplete": true,
   "bibliofile": "biblio.bib",
   "cite_by": "apalike",
   "current_citInitial": 1,
   "eqLabelWithNumbers": true,
   "eqNumInitial": 1,
   "hotkeys": {
    "equation": "Ctrl-E",
    "itemize": "Ctrl-I"
   },
   "labels_anchors": false,
   "latex_user_defs": false,
   "report_style_numbering": false,
   "user_envs_cfg": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}