A1.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "cd838a5b",
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "import numpy as np\n",
    "from torchvision.datasets import MNIST\n",
    "import torchvision.transforms as tf\n",
    "import torch.nn as nn\n",
    "import matplotlib.pyplot as plt\n",
    "from tqdm import tqdm\n",
    "import imageio\n",
    "from torch.utils.data import DataLoader as DL\n",
    "import os"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "c9150198",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<torch._C.Generator at 0x7b751816c270>"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "torch.cuda.is_available()\n",
    "torch.manual_seed(1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "eb31d537",
   "metadata": {},
   "outputs": [],
   "source": [
    "device = \"cuda\" if torch.cuda.is_available() else \"cpu\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "4feb6547",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'cuda'"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "device"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2e839cb3",
   "metadata": {},
   "source": [
    "MNIST Dataset\n",
    "\n",
    "---\n",
    "\n",
    "Import from `datasets`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "7c0167b9",
   "metadata": {},
   "outputs": [],
   "source": [
    "transform = tf.Compose([tf.ToTensor(), tf.Normalize((0.1307), (0.3081,))])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "46ecc6b2",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Downloading http://yann.lecun.com/exdb/mnist/train-images-idx3-ubyte.gz\n",
      "Failed to download (trying next):\n",
      "HTTP Error 403: Forbidden\n",
      "\n",
      "Downloading https://ossci-datasets.s3.amazonaws.com/mnist/train-images-idx3-ubyte.gz\n",
      "Downloading https://ossci-datasets.s3.amazonaws.com/mnist/train-images-idx3-ubyte.gz to ./data/MNIST/raw/train-images-idx3-ubyte.gz\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "389fc9e1a9c74244b9ffc610b9c495c1",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/9912422 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Extracting ./data/MNIST/raw/train-images-idx3-ubyte.gz to ./data/MNIST/raw\n",
      "\n",
      "Downloading http://yann.lecun.com/exdb/mnist/train-labels-idx1-ubyte.gz\n",
      "Failed to download (trying next):\n",
      "HTTP Error 403: Forbidden\n",
      "\n",
      "Downloading https://ossci-datasets.s3.amazonaws.com/mnist/train-labels-idx1-ubyte.gz\n",
      "Downloading https://ossci-datasets.s3.amazonaws.com/mnist/train-labels-idx1-ubyte.gz to ./data/MNIST/raw/train-labels-idx1-ubyte.gz\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "de8f51a879724d778de3f22daf4e3123",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/28881 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Extracting ./data/MNIST/raw/train-labels-idx1-ubyte.gz to ./data/MNIST/raw\n",
      "\n",
      "Downloading http://yann.lecun.com/exdb/mnist/t10k-images-idx3-ubyte.gz\n",
      "Failed to download (trying next):\n",
      "HTTP Error 403: Forbidden\n",
      "\n",
      "Downloading https://ossci-datasets.s3.amazonaws.com/mnist/t10k-images-idx3-ubyte.gz\n",
      "Downloading https://ossci-datasets.s3.amazonaws.com/mnist/t10k-images-idx3-ubyte.gz to ./data/MNIST/raw/t10k-images-idx3-ubyte.gz\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "ccaadb998e794e8eb53a2657f89f4f6d",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/1648877 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Extracting ./data/MNIST/raw/t10k-images-idx3-ubyte.gz to ./data/MNIST/raw\n",
      "\n",
      "Downloading http://yann.lecun.com/exdb/mnist/t10k-labels-idx1-ubyte.gz\n",
      "Failed to download (trying next):\n",
      "HTTP Error 403: Forbidden\n",
      "\n",
      "Downloading https://ossci-datasets.s3.amazonaws.com/mnist/t10k-labels-idx1-ubyte.gz\n",
      "Downloading https://ossci-datasets.s3.amazonaws.com/mnist/t10k-labels-idx1-ubyte.gz to ./data/MNIST/raw/t10k-labels-idx1-ubyte.gz\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "b879589ff95f455d8a2e1eb045b9c23e",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/4542 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Extracting ./data/MNIST/raw/t10k-labels-idx1-ubyte.gz to ./data/MNIST/raw\n",
      "\n"
     ]
    }
   ],
   "source": [
    "train_ds = DL(MNIST(root=\"./data\", train=True, download = True, transform= transform) ,batch_size=35)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "447acb5e",
   "metadata": {},
   "outputs": [],
   "source": [
    "test_ds = DL(MNIST(root=\"./data\", train=False, download = True, transform= transform) ,batch_size = 35)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "a75603e4",
   "metadata": {},
   "outputs": [],
   "source": [
    "ind = iter(train_ds)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "86359e9c",
   "metadata": {},
   "outputs": [],
   "source": [
    "train_batches = len(ind)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "2bec2152",
   "metadata": {},
   "outputs": [],
   "source": [
    "class encoder(torch.nn.Module):\n",
    "    def __init__(self, input_size, bottleneck):\n",
    "        super().__init__()\n",
    "        self.bn = bottleneck\n",
    "        self.input_size = input_size\n",
    "        self.fc1 = nn.Linear(input_size,bottleneck*2)\n",
    "        self.fc2 = nn.Linear(bottleneck*2, bottleneck)\n",
    "    \n",
    "    def forward(self,x):\n",
    "        m = torch.flatten(x, 1)\n",
    "        return(self.fc2(self.fc1(m)))      "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "8126fb5e",
   "metadata": {},
   "outputs": [],
   "source": [
    "class decoder(torch.nn.Module):\n",
    "    def __init__(self,bottleneck,output_size):\n",
    "        super().__init__()\n",
    "        self.bn = bottleneck\n",
    "        self.fc1 = nn.Linear(bottleneck, bottleneck*2)\n",
    "        self.fc2 = nn.Linear(bottleneck*2, output_size)\n",
    "    def forward(self, x):\n",
    "        return (self.fc2(self.fc1(x)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "2e356251",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "decoder(\n",
       "  (fc1): Linear(in_features=20, out_features=40, bias=True)\n",
       "  (fc2): Linear(in_features=40, out_features=784, bias=True)\n",
       ")"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ec = encoder(784,20)\n",
    "dc =  decoder(20, 784)\n",
    "dc"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "abbd9630",
   "metadata": {},
   "outputs": [],
   "source": [
    "class Autoencoder(torch.nn.Module):\n",
    "    def __init__(self, input_size, bottleneck):\n",
    "        super().__init__()\n",
    "        self.bn = bottleneck\n",
    "        self.input_size = input_size\n",
    "        self.encoder = encoder(self.input_size, self.bn)\n",
    "        self.decoder = decoder(self.bn, self.input_size)\n",
    "    def forward(self,x):\n",
    "        return self.decoder(self.encoder(x))\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "fc18cde3",
   "metadata": {},
   "outputs": [],
   "source": [
    "loss_fn = torch.nn.MSELoss()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "52d3d4e2",
   "metadata": {},
   "outputs": [],
   "source": [
    "y = next(ind)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "91ab04b1",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "torch.Size([35, 1, 28, 28])"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y[0].shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "dd86649b",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([5, 0, 4, 1, 9, 2, 1, 3, 1, 4, 3, 5, 3, 6, 1, 7, 2, 8, 6, 9, 4, 0, 9, 1,\n",
       "        1, 2, 4, 3, 2, 7, 3, 8, 6, 9, 0])"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y[1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "07af8037",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "torch.Size([35, 20])\n"
     ]
    }
   ],
   "source": [
    "img = torch.randn(35, 1, 28, 28)\n",
    "op = ec(img)\n",
    "print(op.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "548d44c7",
   "metadata": {},
   "outputs": [],
   "source": [
    "ae = Autoencoder(784,32)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "184889ad",
   "metadata": {},
   "outputs": [],
   "source": [
    "train_losses = []\n",
    "test_losses = []\n",
    "opt =  torch.optim.Adam(ae.parameters(), lr  = 0.001)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "5a21a5f7",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Autoencoder(\n",
       "  (encoder): encoder(\n",
       "    (fc1): Linear(in_features=784, out_features=64, bias=True)\n",
       "    (fc2): Linear(in_features=64, out_features=32, bias=True)\n",
       "  )\n",
       "  (decoder): decoder(\n",
       "    (fc1): Linear(in_features=32, out_features=64, bias=True)\n",
       "    (fc2): Linear(in_features=64, out_features=784, bias=True)\n",
       "  )\n",
       ")"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ae.to(device)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "2e3de03d",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "422912"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "torch.cuda.memory_allocated()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "1b3f531e",
   "metadata": {},
   "outputs": [],
   "source": [
    "def calc_loss(dl,model):\n",
    "    model.eval()\n",
    "    loss = 0\n",
    "    count = 0\n",
    "    for index,(image, target) in enumerate(dl):\n",
    "        image = image.to(device)\n",
    "        flat_img = torch.flatten(image, 1)\n",
    "        loss = loss + loss_fn(model(image), flat_img).item()\n",
    "        count = count + 1\n",
    "    return loss/count"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "6cb274d9",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1.019876085877766\n"
     ]
    }
   ],
   "source": [
    "calc_loss(test_ds, ae)\n",
    "print(calc_loss(train_ds, ae))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "bd9b7e59",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "60000"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(train_ds.dataset)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "3dfc8b4a",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1715/1715 [00:17<00:00, 98.72it/s] \n",
      "100%|██████████| 1715/1715 [00:18<00:00, 94.93it/s] \n",
      "100%|██████████| 1715/1715 [00:20<00:00, 83.34it/s] \n",
      "100%|██████████| 3/3 [00:56<00:00, 18.68s/it]\n"
     ]
    }
   ],
   "source": [
    "for epoch in tqdm(range(3)):\n",
    "    for index,(image, target) in (enumerate(tqdm(train_ds, leave=True))):\n",
    "        ae.train()\n",
    "        opt.zero_grad()\n",
    "        ae = ae.to(device)\n",
    "        image = image.to(device)\n",
    "        flat_img = torch.flatten(image, 1)\n",
    "        op = ae(image)\n",
    "        loss = loss_fn(op, flat_img)\n",
    "        loss.backward()\n",
    "        opt.step()\n",
    "        if(index %600 == 0):\n",
    "            train_losses.append(loss.item())\n",
    "            test_loss = calc_loss(test_ds, ae)\n",
    "            test_losses.append(test_loss)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "acc2d6cd",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x7b746cd10e80>"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(train_losses, label='train_losses')\n",
    "plt.plot(test_losses, label= 'test_losses')\n",
    "plt.xlabel('Index per 300 iterations')\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "4220860f",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Autoencoder(\n",
       "  (encoder): encoder(\n",
       "    (fc1): Linear(in_features=784, out_features=64, bias=True)\n",
       "    (fc2): Linear(in_features=64, out_features=32, bias=True)\n",
       "  )\n",
       "  (decoder): decoder(\n",
       "    (fc1): Linear(in_features=32, out_features=64, bias=True)\n",
       "    (fc2): Linear(in_features=64, out_features=784, bias=True)\n",
       "  )\n",
       ")"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test_losses[-1]\n",
    "ae = ae.to(\"cpu\")\n",
    "ae.eval()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "f590a01a",
   "metadata": {},
   "outputs": [],
   "source": [
    "batch  = next(ind)\n",
    "img = batch[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "16a315f2",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x7b74614c7f10>"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.imshow(img[0].permute(1, 2, 0))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "108ee1d7",
   "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": [
    "plt.imshow(ae(img[0]).reshape(-1,28,28).permute(1,2,0).detach().numpy())\n",
    "latent = ae.encoder(img[0])"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d482fb63",
   "metadata": {},
   "source": [
    "#### Using the model to visualize the Manifold structure in the latent space"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "df713865",
   "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": [
    "batch_next = next(ind)\n",
    "image_next = batch_next[0]\n",
    "plt.imshow(image_next[0].permute(1, 2, 0))\n",
    "latent_next = ae.encoder(image_next[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "09c453ff",
   "metadata": {},
   "outputs": [],
   "source": [
    "# obtain the latent representations of both of the images\n",
    "detacher = lambda x: torch.reshape(x, (-1, 28, 28)).detach().numpy()\n",
    "imager = lambda x: torch.reshape(x, (-1, 28, 28)).permute(1, 2, 0).detach().numpy()\n",
    "det_trans = tf.Lambda(detacher) \n",
    "image_trans = tf.Lambda(imager)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "id": "431d4a03",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "mkdir: cannot create directory ‘images’: File exists\n"
     ]
    }
   ],
   "source": [
    "!mkdir images"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "id": "4c510a14",
   "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": [
    "filenames = []\n",
    "count = 0\n",
    "for alpha in np.linspace(0,1, 60):\n",
    "    count = count + 1\n",
    "    latent_mid = alpha*latent + (1 - alpha)*latent_next\n",
    "    plt.imshow(image_trans(ae.decoder(latent_mid)))\n",
    "    plt.savefig('images/alpha_' + str(count) + '.png')\n",
    "    filenames.append('images/alpha_' + str(count) + '.png')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "id": "4279b638",
   "metadata": {},
   "outputs": [],
   "source": [
    "images = []\n",
    "for file in filenames:\n",
    "    images.append(imageio.v2.imread(file))\n",
    "    \n",
    "imageio.mimsave('ae.gif', images)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f862e474",
   "metadata": {},
   "source": [
    "![](ae.gif)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "id": "44c3c5b0",
   "metadata": {},
   "outputs": [],
   "source": [
    "for file in filenames:\n",
    "    os.remove(file)"
   ]
  }
 ],
 "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": "3.9.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}