seq2seq.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "c7eb3670-671e-4709-bc95-bc6301474bda",
   "metadata": {},
   "source": [
    "# Import Object"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "id": "d0bf2fa1-eeaf-425c-b843-0b25bd96dc9e",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "from sklearn import preprocessing\n",
    "from sklearn.model_selection import train_test_split\n",
    "import torch\n",
    "from torch import nn\n",
    "from torch import optim\n",
    "from torch.utils.data import TensorDataset, DataLoader\n",
    "\n",
    "from models import Seq2Seq, Encoder, Decoder\n",
    "from train_model import train_seq2seq_net"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "id": "f48de273-0d80-493a-b33f-fa9c3e3a46fc",
   "metadata": {},
   "outputs": [],
   "source": [
    "device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "75cec094-bde7-4ba4-a6e1-e6be278aea5c",
   "metadata": {},
   "source": [
    "# Prepare Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "id": "1ececdbd-d3ae-4162-be3b-cc31419c354e",
   "metadata": {},
   "outputs": [],
   "source": [
    "train_x = pd.read_csv(\"./data/X_train.csv\").values\n",
    "train_y = pd.read_csv(\"./data/Y_train.csv\").values.reshape(-1)\n",
    "\n",
    "ss = preprocessing.StandardScaler()\n",
    "train_x = ss.fit_transform(train_x)\n",
    "\n",
    "num_days = int(train_x.shape[0] / 16)\n",
    "train_x = train_x.reshape(num_days, 16, 16)\n",
    "train_y = train_y.reshape(num_days, 16, 1)\n",
    "train_x, test_x, train_y, test_y = train_test_split(train_x,\n",
    "                                                    train_y,\n",
    "                                                    test_size=0.35,\n",
    "                                                    shuffle=True)\n",
    "\n",
    "train_x = torch.tensor(train_x, dtype=torch.float32)\n",
    "train_y = torch.tensor(train_y, dtype=torch.float32)\n",
    "test_x = torch.tensor(test_x, dtype=torch.float32)\n",
    "test_y = torch.tensor(test_y, dtype=torch.float32)\n",
    "\n",
    "train_x = train_x.to(device)\n",
    "train_y = train_y.to(device)\n",
    "test_x = test_x.to(device)\n",
    "test_y = test_y.to(device)\n",
    "\n",
    "train_ds = TensorDataset(train_x, train_y)\n",
    "train_loader = DataLoader(train_ds, batch_size=32, shuffle=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0c0b59df-be09-45e1-8483-6416194613e4",
   "metadata": {},
   "source": [
    "# Set Free Parameters"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 79,
   "id": "602070e9-c806-4244-af92-2f7e055af31e",
   "metadata": {},
   "outputs": [],
   "source": [
    "input_size_encoder = 16\n",
    "input_size_decoder = 1\n",
    "num_layers = 1\n",
    "hidden_size = 50\n",
    "num_classes = 2\n",
    "learning_rate = 0.001\n",
    "num_epochs = 50\n",
    "dropout_ratio = 0.5"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6ed01196-81be-42a8-9bc7-3c64f3199762",
   "metadata": {},
   "source": [
    "# Instantiate Model, Criterion, Optimizer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "id": "ba9f539d-32e5-43f4-bcad-c04782facca7",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/higuchi-lab/anaconda3/lib/python3.8/site-packages/torch/nn/modules/rnn.py:58: UserWarning: dropout option adds dropout after all but last recurrent layer, so non-zero dropout expects num_layers greater than 1, but got dropout=0.5 and num_layers=1\n",
      "  warnings.warn(\"dropout option adds dropout after all but last \"\n"
     ]
    }
   ],
   "source": [
    "encoder_net = Encoder(input_size_encoder, hidden_size,\n",
    "                      num_layers, dropout_ratio).to(device)\n",
    "decoder_net = Decoder(input_size_decoder, hidden_size,\n",
    "                      num_classes, num_layers, dropout_ratio).to(device)\n",
    "\n",
    "model = Seq2Seq(encoder_net, decoder_net).to(device)\n",
    "criterion = nn.CrossEntropyLoss()\n",
    "optimizer = optim.Adam(model.parameters(), lr=learning_rate)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "cb0fab1a-1768-49e2-be45-6e5abfa62296",
   "metadata": {},
   "source": [
    "# Training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 81,
   "id": "e03573db-9770-481a-ac8a-c0023b711224",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "model = train_seq2seq_net(model, criterion, optimizer,\n",
    "                          num_epochs, train_loader, device)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4248eab2-5ada-477d-926c-3dadae20ac52",
   "metadata": {},
   "source": [
    "# Accuracy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "id": "439af3df-78be-4e6f-9a7c-d30e84389b98",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tensor(0.8121, device='cuda:0')\n"
     ]
    }
   ],
   "source": [
    "model.eval()\n",
    "decoder_y = test_y[:, :-1, :]\n",
    "\n",
    "decoder_t = test_y[:, 1:, :]\n",
    "decoder_t = decoder_t.reshape(-1)\n",
    "\n",
    "scores = model.generate(test_x, decoder_y, decoder_y.shape[1], device)\n",
    "pred_y = scores.argmax(axis=2)\n",
    "pred_y = pred_y.reshape(-1)\n",
    "\n",
    "print(sum(pred_y == decoder_t) / int(decoder_t.shape[0]))"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.8.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}