proto_variational.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import math\n",
    "import torch\n",
    "import torch.nn as nn\n",
    "import torch.nn.functional as F\n",
    "import torch.optim as optim\n",
    "import torch.autograd as autograd\n",
    "import torchvision.transforms as transforms\n",
    "from torch.autograd import Variable\n",
    "from torch.nn.parameter import Parameter\n",
    "%matplotlib inline\n",
    "import pylab as pl\n",
    "from IPython import display\n",
    "import time"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "class Block(nn.Module):\n",
    "    def __init__(self, insize, outsize):\n",
    "        super(Block, self).__init__()\n",
    "        self.layers = nn.Sequential(\n",
    "            nn.Conv2d(insize, outsize, kernel_size=3, padding=1),\n",
    "            nn.BatchNorm2d(outsize),\n",
    "            nn.ReLU(),\n",
    "            nn.MaxPool2d(2)\n",
    "        )\n",
    "        \n",
    "    def forward(self, inp):\n",
    "        return self.layers(inp)\n",
    "\n",
    "class ENCODER(nn.Module):\n",
    "    def __init__(self):\n",
    "        super(ENCODER, self).__init__()\n",
    "        self.block1 = Block(3,64)\n",
    "        self.block2 = Block(64,64)\n",
    "        self.block3 = Block(64,64)\n",
    "        self.block4 = Block(64,64)\n",
    "        self.final = nn.Conv2d(64,64, kernel_size=1, padding=0)\n",
    "        \n",
    "    def forward(self, inp, samples):\n",
    "        out = self.block1(inp)\n",
    "        out = self.block2(out)\n",
    "        out = self.block3(out)\n",
    "        out = self.block4(out)\n",
    "        out = self.final(out)\n",
    "        out = out.view(out.size(0),-1)\n",
    "        mu = out[:,:out.size(1)//2]\n",
    "        sigma = out[:,out.size(1)//2:]\n",
    "        noise = Variable(torch.randn(sigma.size(0),samples,sigma.size(1)), requires_grad=False).cuda()\n",
    "        samples = noise*torch.exp(sigma).unsqueeze(1).expand_as(noise)\n",
    "        results = samples+mu.unsqueeze(1).expand_as(noise)\n",
    "        return results.view(-1,mu.size(1)),mu,sigma\n",
    "    \n",
    "class PREDICTOR(nn.Module):\n",
    "    def __init__(self):\n",
    "        super(PREDICTOR, self).__init__()\n",
    "        self.sm = nn.LogSoftmax()\n",
    "        \n",
    "    def forward(self, support, inp, way):\n",
    "        assert support.size(0)%way == 0,\"Error: classes are not the same size!\"\n",
    "        support = support.view(way,support.size(0)//way,-1)\n",
    "        centroids = torch.mean(support, 1).view(support.size(0),-1)\n",
    "#         temp1 = torch.stack([centroids]*inp.size(0))\n",
    "#         temp2 = torch.stack([inp]*centroids.size(0)).transpose(0,1)\n",
    "#         print(temp1.size(),temp2.size())\n",
    "        distmat = torch.sum((torch.stack([centroids]*inp.size(0))-torch.stack([inp]*centroids.size(0))\n",
    "                             .transpose(0,1))**2,2).squeeze()*-1\n",
    "        probs = self.sm(distmat)\n",
    "        return probs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "trainset = torch.load('mIN_train.pth')\n",
    "testset = torch.load('mIN_test.pth')\n",
    "valset = torch.load('mIN_val.pth')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "standardize = transforms.Compose([\n",
    "            transforms.ToTensor(),\n",
    "            transforms.Normalize(mean=[.485,.456,.406],std=[.229,.224,.225])\n",
    "            ])\n",
    "alter = transforms.Compose([\n",
    "            transforms.ToPILImage(),\n",
    "            transforms.RandomHorizontalFlip(),\n",
    "            transforms.RandomCrop(84,padding=10),\n",
    "            standardize\n",
    "            ])\n",
    "\n",
    "def batchmaker(way,trainshot,testshot,theset,alterful=False):\n",
    "    classes = np.random.choice(len(theset),way)\n",
    "    if alterful:\n",
    "        li = [torch.cat([alter(theset[cl][i]).view(1,3,84,84) for i in \n",
    "                         np.random.choice(600,trainshot+testshot)],dim=0).float()\n",
    "              for cl in classes]\n",
    "    else:\n",
    "        li = [torch.cat([standardize(theset[cl][i]).view(1,3,84,84) for i in \n",
    "                         np.random.choice(600,trainshot+testshot)],dim=0).float()\n",
    "              for cl in classes]\n",
    "    support = torch.cat([t[:trainshot,:,:,:] for t in li],dim=0)\n",
    "    stargs = torch.LongTensor([i//trainshot for i in range(trainshot*way)])\n",
    "    query = torch.cat([t[trainshot:,:,:,:] for t in li],dim=0)\n",
    "    qtargs = torch.LongTensor([i//testshot for i in range(testshot*way)])\n",
    "#     print(support.size(),query.size())\n",
    "    return(Variable(support).cuda(),\n",
    "           Variable(query, volatile=(not alterful)).cuda(),\n",
    "           Variable(qtargs, volatile=(not alterful)).cuda(),\n",
    "           Variable(stargs).cuda()\n",
    "          )\n",
    "\n",
    "def evaluate(embed,model,criterion,way,trainshot,testshot,samples,reps,theset):\n",
    "    losstracker = 0\n",
    "    acctracker = 0\n",
    "    acclist = []\n",
    "    for i in range(reps):\n",
    "        s,q,qt,st = batchmaker(way,trainshot,testshot,theset)\n",
    "        embeds,_,_ = embed(s, samples)\n",
    "        _,qembeds,_ = embed(q, 1)\n",
    "        preds = model(embeds,qembeds,way)\n",
    "        \n",
    "        # Val loss\n",
    "        loss = criterion(preds,qt)\n",
    "        losstracker += loss.data[0]\n",
    "        \n",
    "        # Val accuracy\n",
    "        _,bins = torch.max(preds,1)\n",
    "        acc = torch.sum(torch.eq(bins,qt)).data[0]/testshot/way\n",
    "        acctracker += acc\n",
    "        acclist.append(acc)\n",
    "    return losstracker/reps,acctracker/reps,np.var(acclist)\n",
    "\n",
    "way = 20\n",
    "evalway = 5\n",
    "trainshot = 5\n",
    "testshot = 15\n",
    "reps = 100\n",
    "\n",
    "trainsamples = 10\n",
    "testsamples = 50"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# torch.cuda.set_device(1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "117248  parameters in neural net.\n"
     ]
    }
   ],
   "source": [
    "embed = ENCODER().cuda()\n",
    "model = PREDICTOR().cuda()\n",
    "losstracker,regtracker,evallosstracker,evalacctracker = [],[],[],[]\n",
    "optimizer = None\n",
    "nweights = sum([i.numel() for i in list(embed.parameters())])\n",
    "print(nweights,\" parameters in neural net.\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA20AAAE/CAYAAADVKysfAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n2P8DCoc2GYibNOUAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7ff9a3ad1b00>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train loss is: 1.4914665117263795\n",
      "Regularizer loss is: 0.07471243998408318\n",
      "Validation accuracy is: 0.6406666666666667\n",
      "Validation loss is: 0.8654298713803291\n",
      "\n",
      "9988.952450990677\n"
     ]
    }
   ],
   "source": [
    "# TRAIN!\n",
    "criterion = nn.NLLLoss().cuda()\n",
    "vbity = 250\n",
    "epoch = 2000\n",
    "start = time.time()\n",
    "\n",
    "runningloss = 0\n",
    "runningreg = 0\n",
    "for it in range(10*epoch):\n",
    "    if it%10==0:\n",
    "        print(it)\n",
    "    \n",
    "    # Build batch\n",
    "    support, query, targs, _ = batchmaker(way,trainshot,testshot,trainset,alterful=True)\n",
    "    \n",
    "    # Predict\n",
    "    embed.zero_grad()\n",
    "    model.zero_grad()\n",
    "    embeds,mu,sigma = embed(support, trainsamples)\n",
    "    _,qembeds,_ = embed(query, 1)\n",
    "    preds = model(embeds,qembeds,way)\n",
    "\n",
    "    # Calculate Loss\n",
    "    loss = criterion(preds, targs)\n",
    "    reg = .5*torch.mean(torch.exp(sigma)+mu**2-1-sigma)\n",
    "    runningloss += loss.data[0]\n",
    "    runningreg += reg.data[0]\n",
    "    totalloss = loss+reg\n",
    "    \n",
    "    # Backprop\n",
    "    if it%epoch == 0:\n",
    "        optimizer = optim.Adam(embed.parameters(),lr=.001/(2**(it//epoch)))\n",
    "    totalloss.backward()\n",
    "#     nn.utils.clip_grad_norm(model.parameters(), 1)\n",
    "    optimizer.step()\n",
    "    \n",
    "    # Report\n",
    "    if it%vbity == vbity-1:\n",
    "        display.clear_output(wait=True)\n",
    "\n",
    "        losstracker.append(runningloss/vbity)\n",
    "        regtracker.append(runningreg/vbity)\n",
    "        embed = embed.eval()\n",
    "        evalloss, evalacc, _ = evaluate(embed,model,criterion,evalway,trainshot,testshot,testsamples,\n",
    "                                        reps,testset)\n",
    "        embed = embed.train()\n",
    "        evallosstracker.append(evalloss)\n",
    "        evalacctracker.append(evalacc)\n",
    "        \n",
    "        pl.figure(1,figsize=(15,5))\n",
    "        pl.subplot(1,2,1)\n",
    "        pl.plot(losstracker)\n",
    "        pl.plot(evallosstracker)\n",
    "        pl.ylim((0.5,3))\n",
    "        pl.title(\"Loss: Training Blue, Validation Gold\")\n",
    "        pl.subplot(1,2,2)\n",
    "        pl.plot(evalacctracker[::-1])\n",
    "        pl.ylim((0.3,.8))\n",
    "        pl.title(\"Validation Acc\")\n",
    "        pl.show()\n",
    "        \n",
    "        print(\"Train loss is: \"+str(runningloss/vbity)+\n",
    "              \"\\nRegularizer loss is: \"+str(runningreg/vbity)+\n",
    "              \"\\nValidation accuracy is: \"+str(evalacc)+\n",
    "              \"\\nValidation loss is: \"+str(evalloss)+\"\\n\")\n",
    "        runningloss = 0\n",
    "        runningreg = 0\n",
    "        print(time.time()-start)\n",
    "\n",
    "# MORE RECENT TEST\n",
    "        "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(0.6057993148267269, 0.7525999999999996, 0.0090524992592592619)\n"
     ]
    }
   ],
   "source": [
    "criterion = nn.NLLLoss().cuda()\n",
    "embed = embed.eval()\n",
    "print(evaluate(embed,model,criterion,evalway,trainshot,testshot*3,200,testset))\n",
    "embed = embed.train()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.01318637251104336"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "1.96*np.sqrt(.0090525/200)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.64839733333333338"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.mean(evalacctracker[-50:])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/dww78/anaconda3/lib/python3.6/site-packages/torch/serialization.py:147: UserWarning: Couldn't retrieve source code for container of type ENCODER. It won't be checked for correctness upon loading.\n",
      "  \"type \" + obj.__name__ + \". It won't be checked \"\n",
      "/home/dww78/anaconda3/lib/python3.6/site-packages/torch/serialization.py:147: UserWarning: Couldn't retrieve source code for container of type Block. It won't be checked for correctness upon loading.\n",
      "  \"type \" + obj.__name__ + \". It won't be checked \"\n",
      "/home/dww78/anaconda3/lib/python3.6/site-packages/torch/serialization.py:147: UserWarning: Couldn't retrieve source code for container of type PREDICTOR. It won't be checked for correctness upon loading.\n",
      "  \"type \" + obj.__name__ + \". It won't be checked \"\n"
     ]
    }
   ],
   "source": [
    "name = 'variational'\n",
    "torch.save((embed,model,losstracker,evallosstracker,evalacctracker),'saved_models/'+name+'.pth')\n",
    "torch.save(embed.cpu().state_dict(),'saved_models/'+name+'_cpu.pth')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "embed,model,losstracker,evallosstracker,evalacctracker = torch.load('saved_models/naive_4.pth')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "Paper model achieves (over last 1000):\n",
    "    0.71156711111111126\n",
    "after 188500 iters"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 213,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Variable containing:\n",
       "-6.7075e-03 -5.0079e+00 -2.7911e+01\n",
       "-2.9057e+00 -5.6263e-02 -4.5704e+01\n",
       "-1.8272e+01 -4.3870e+01 -1.1600e-08\n",
       "[torch.FloatTensor of size 3x3]"
      ]
     },
     "execution_count": 213,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "support = supportmaker(nclass,shot)\n",
    "inp = inpmaker(nclass)\n",
    "embeds = embed(support)\n",
    "query = embed(inp)\n",
    "model(embeds,query)"
   ]
  }
 ],
 "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": "3.6.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}