pytorch_template.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# PyTorch Code Examples\n",
    "\n",
    "## Tensors and Variables"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* pytorch Tensors는 numpy's arrays와 비슷하게 동작한다"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tensor([[1., 2.],\n",
      "        [3., 4.]])\n"
     ]
    }
   ],
   "source": [
    "a = torch.Tensor([[1,2],[3,4]])\n",
    "print(a)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tensor([[ 1.,  4.],\n",
      "        [ 9., 16.]])\n"
     ]
    }
   ],
   "source": [
    "print(a**2)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* pytorch Variables는 a tensor 그리고 record operations 를 동작하게 묶어준다.\n",
    "* backprop 동작을 하게 만들어준다"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "from torch.autograd import Variable\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tensor([[1., 2.],\n",
      "        [3., 4.]], requires_grad=True)\n"
     ]
    }
   ],
   "source": [
    "a = Variable(torch.Tensor([[1,2],[3,4]]), requires_grad=True)\n",
    "print(a) # torch.FloatTensor of size 2x2\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tensor(30., grad_fn=<SumBackward0>)\n"
     ]
    }
   ],
   "source": [
    "y = torch.sum(a**2)\n",
    "print(y) # torch.FloatTensor of size 1\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tensor([[2., 4.],\n",
      "        [6., 8.]])\n"
     ]
    }
   ],
   "source": [
    "y.backward() # compute gradients of y wrt a\n",
    "print(a.grad) # print dy/da_ij = 2 * a_ij for a_11, a_12, a_21, a_22"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Core Traning Step\n",
    "\n",
    "* pass a batch of inputs through the model / batch 인풋을 모델에 넣기\n",
    "* calculate the loss / 로스 계산\n",
    "* calculate backprop / backprop 계산\n",
    "* update the parameters / 파라미터 업데이트"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'model' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-8-b42fa39ec9b5>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m\u001b[0m\n\u001b[1;32m----> 1\u001b[1;33m \u001b[0moutput_batch\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mmodel\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mtrain_batch\u001b[0m\u001b[1;33m)\u001b[0m \u001b[1;31m# compute model output\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m      2\u001b[0m \u001b[0mloss\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mloss_fn\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0moutput_batch\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mlabels_batch\u001b[0m\u001b[1;33m)\u001b[0m \u001b[1;31m# calculate loss\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      3\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      4\u001b[0m \u001b[0moptimizer\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mzero_grad\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m \u001b[1;31m# clear previous gradients\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      5\u001b[0m \u001b[0mloss\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mbackward\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m \u001b[1;31m# comput gradients of all variables wrt loss\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;31mNameError\u001b[0m: name 'model' is not defined"
     ]
    }
   ],
   "source": [
    "output_batch = model(train_batch) # compute model output\n",
    "loss = loss_fn(output_batch, labels_batch) # calculate loss\n",
    "\n",
    "optimizer.zero_grad() # clear previous gradients\n",
    "loss.backward() # comput gradients of all variables wrt loss\n",
    "\n",
    "optimizer.step() # perform updates using claculated gradients"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Models in pytorch\n",
    "* pytorch의 subclass인 torch.nn.Module을 이용해서 모델을 정의한다.\n",
    "* init 과 forward 2가지 단계로 모델을 구성한다.\n",
    "* operations은 학습가능한 파라미터가 아니므로 torch.nn.functional을 사용한다.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch.nn as nn\n",
    "import torch.nn.functional as F\n",
    "\n",
    "class TwoLayerNet(nn.Module):\n",
    "  def __init__(self, D_in, H, D_out):\n",
    "    \"\"\"\n",
    "    constructor에 2개 모듈을 멤버변수로 할당한다.\n",
    "    \n",
    "    D_in : input dimension\n",
    "    H :dimension of hidden layer\n",
    "    D_out : output dimension\n",
    "    \"\"\"\n",
    "    super(TwoLayerNet, self).__init__()\n",
    "    self.linear1 = nn.Linear(D_in, H)\n",
    "    self.linear2 = nn.Linear(H, D_out)\n",
    "    \n",
    "  def forward(self, x):\n",
    "    \"\"\"\n",
    "    input을 받아 pred 값으로 리턴, 모델에 인풋에 대하여 아웃풋을 정의\n",
    "    모델 설계하는 부분 tf2 의 call 부분\n",
    "    \"\"\"\n",
    "    h_relu = F.relu(self.linear1(x))\n",
    "    y_pred = self.linear2(h_relu)\n",
    "    \n",
    "    return y_pred"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* 위의 모델을 이용해서 간단히 모델을 정의해보자"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# N is batch size; D_in is input dimension;\n",
    "# H is the dimension of the hidden layer; D_out is output dimension.\n",
    "N, D_in, H, D_out = 32, 100, 50, 10\n",
    "\n",
    "# Create random Tensors to hold inputs and outputs, and wrap them in Variables\n",
    "x = Variable(torch.randn(N, D_in))  # dim: 32 x 100\n",
    "\n",
    "# Construct our model by instantiating the class defined above\n",
    "model = TwoLayerNet(D_in, H, D_out)\n",
    "\n",
    "# Forward pass: Compute predicted y by passing x to the model\n",
    "y_pred = model(x)   # dim: 32 x 10"
   ]
  },
  {
   "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": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}