RNN_building_blocks.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true,
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "# RNN Building Blocks: injecting recursivity to supervised learner"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 262,
   "outputs": [],
   "source": [
    "import torch\n",
    "from torch.utils.data import Dataset\n",
    "from torch import optim\n",
    "from torch.autograd import Variable\n",
    "from torchvision import datasets, transforms\n",
    "import torch.nn.functional as F\n",
    "from torch import nn\n",
    "torch.manual_seed(0)\n",
    "torch.backends.cudnn.deterministic = True\n",
    "torch.backends.cudnn.benchmark = False\n",
    "\n",
    "import numpy as np\n",
    "np.random.seed(0)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 263,
   "outputs": [],
   "source": [
    "from torch import nn\n",
    "\n",
    "\n",
    "class mySeries(Dataset):\n",
    "    \"\"\"Series for RNN dataset.\"\"\"\n",
    "\n",
    "    def __init__(self,  length=1000, transform=None):\n",
    "        self.transform = transform\n",
    "        self.length = length\n",
    "\n",
    "    def __len__(self):\n",
    "        return self.length\n",
    "\n",
    "    def __getitem__(self, idx):\n",
    "        if torch.is_tensor(idx):\n",
    "            idx = idx.tolist()\n",
    "\n",
    "        X = idx+2\n",
    "        y = idx+4\n",
    "\n",
    "        return X,y\n",
    "\n",
    "# mySeries dataset\n",
    "trainset = mySeries()\n",
    "trainloader = torch.utils.data.DataLoader(trainset, batch_size=64)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 264,
   "outputs": [],
   "source": [
    "class RSuperviseL(nn.Module):\n",
    "  def __init__(self):\n",
    "    super().__init__()\n",
    "    self.fc1 = nn.Linear(1, 1)\n",
    "\n",
    "  def forward(self, x):\n",
    "\n",
    "    x= self.fc1(x)\n",
    "    return x\n",
    "\n",
    "model = RSuperviseL()"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 265,
   "outputs": [],
   "source": [
    "criterion = nn.MSELoss()\n",
    "optimizer = optim.RMSprop(model.parameters(), lr=0.001)\n",
    "num_epochs = 1000\n",
    "train_tracker, test_tracker, accuracy_tracker = [], [], []"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 266,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch(1000/1000) | Training loss: 0.40327949904312854 | \n"
     ]
    }
   ],
   "source": [
    "for i in range(num_epochs):\n",
    "    cum_loss = 0\n",
    "\n",
    "    for batch, (X, y) in enumerate(trainloader,1):\n",
    "        X = X.type(torch.float32)\n",
    "        y = y.type(torch.float32)\n",
    "        optimizer.zero_grad()\n",
    "        output = model(X.view(-1,1))\n",
    "        loss = criterion(output, y.view(-1,1))\n",
    "        loss.backward()\n",
    "        optimizer.step()\n",
    "\n",
    "        cum_loss += loss.item()\n",
    "\n",
    "    train_tracker.append(cum_loss/len(trainloader))\n",
    "print(f\"Epoch({i+1}/{num_epochs}) | Training loss: {cum_loss/len(trainloader)} | \")\n"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 267,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Parameter containing:\n",
      "tensor([[1.0000]], requires_grad=True)\n",
      "Parameter containing:\n",
      "tensor([1.9918], requires_grad=True)\n"
     ]
    }
   ],
   "source": [
    "for param in model.parameters():\n",
    "    print(param)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}