simplecnn-vs-pigpen.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Pigpen PyTorch Example\n",
    "by Joe Norton"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "import torchvision\n",
    "import torchvision.transforms as transforms\n",
    "import torch.nn as nn\n",
    "import torch.nn.functional as F\n",
    "import torch.optim as optim"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "import random\n",
    "from PigpenDataset import PigpenDatasetClass\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "transform = transforms.Compose([\n",
    "    transforms.ToTensor(),\n",
    "    #transforms.Normalize((0.5,), (0.5,))  # Example normalization\n",
    "])\n",
    "\n",
    "train_dataset = PigpenDatasetClass('data/train', transform=transform)\n",
    "test_dataset = PigpenDatasetClass('data/test', transform=transform)\n",
    "valid_dataset = PigpenDatasetClass('data/validation', transform=transform)\n",
    "\n",
    "train_loader = torch.utils.data.DataLoader(train_dataset, batch_size=27, shuffle=True)\n",
    "test_loader = torch.utils.data.DataLoader(test_dataset, batch_size=27, shuffle=False)\n",
    "valid_loader = torch.utils.data.DataLoader(valid_dataset, batch_size=27, shuffle=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(54000, 10800, 10800)"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(train_dataset), len(test_dataset), len(valid_dataset)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Using cuda device\n",
      "SimpleCNN(\n",
      "  (flatten): Flatten(start_dim=1, end_dim=-1)\n",
      "  (conv_relu_stack): Sequential(\n",
      "    (0): Conv2d(1, 32, kernel_size=(5, 5), stride=(1, 1), padding=same)\n",
      "    (1): ReLU()\n",
      "    (2): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
      "    (3): Conv2d(32, 64, kernel_size=(5, 5), stride=(1, 1), padding=same)\n",
      "    (4): ReLU()\n",
      "    (5): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
      "  )\n",
      "  (linear_layers): Sequential(\n",
      "    (0): Linear(in_features=3136, out_features=256, bias=True)\n",
      "    (1): ReLU()\n",
      "    (2): Linear(in_features=256, out_features=28, bias=True)\n",
      "  )\n",
      ")\n"
     ]
    }
   ],
   "source": [
    "# Get cpu, gpu or mps device for training.\n",
    "device = (\n",
    "    \"cuda\"\n",
    "    if torch.cuda.is_available()\n",
    "    else \"mps\"\n",
    "    if torch.backends.mps.is_available()\n",
    "    else \"cpu\"\n",
    ")\n",
    "print(f\"Using {device} device\")\n",
    "\n",
    "class SimpleCNN(nn.Module):\n",
    "    def __init__(self):\n",
    "        super(SimpleCNN, self).__init__()\n",
    "        self.flatten = nn.Flatten()\n",
    "        self.conv_relu_stack = nn.Sequential(\n",
    "            nn.Conv2d(1, 32, kernel_size=5, stride=1, padding=\"same\"),\n",
    "            nn.ReLU(),\n",
    "            nn.MaxPool2d(2),\n",
    "            nn.Conv2d(32, 64, kernel_size=5, stride=1, padding=\"same\"),\n",
    "            nn.ReLU(),\n",
    "            nn.MaxPool2d(2)\n",
    "        )\n",
    "        # Adjust these sizes based on your actual output from convolutions\n",
    "        self.linear_layers = nn.Sequential(\n",
    "            nn.Linear(3136, 256),  # Adjust size according to output\n",
    "            nn.ReLU(),\n",
    "            nn.Linear(256, 28)  # 28 output classes\n",
    "        )\n",
    "\n",
    "    def forward(self, x):\n",
    "        x = self.conv_relu_stack(x)\n",
    "        x = self.flatten(x)  # Flatten the tensor\n",
    "        x = self.linear_layers(x)\n",
    "        return x\n",
    "\n",
    "model = SimpleCNN().to(device)\n",
    "print(model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "loss_fn = nn.CrossEntropyLoss()\n",
    "optimizer = optim.SGD(model.parameters(), lr=0.01, momentum=0.9)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "#stat accumulators\n",
    "epoch_val_loss = []\n",
    "epoch_val_acc = []"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "def train(dataloader, model, loss_fn, optimizer):\n",
    "    size = len(dataloader.dataset)\n",
    "    model.train()\n",
    "    for batch, (X, y) in enumerate(dataloader):\n",
    "        X, y = X.to(device), y.to(device)\n",
    "\n",
    "        # Compute prediction error\n",
    "        pred = model(X)\n",
    "        loss = loss_fn(pred, y)\n",
    "\n",
    "        # Backpropagation\n",
    "        loss.backward()\n",
    "        optimizer.step()\n",
    "        optimizer.zero_grad()\n",
    "\n",
    "        if batch % 100 == 0:\n",
    "            loss, current = loss.item(), (batch + 1) * len(X)\n",
    "            print(f\"loss: {loss:>7f}  [{current:>5d}/{size:>5d}]\")\n",
    "        \n",
    "def test(dataloader, model, loss_fn):\n",
    "    size = len(dataloader.dataset)\n",
    "    num_batches = len(dataloader)\n",
    "    model.eval()\n",
    "    test_loss, correct = 0, 0\n",
    "    with torch.no_grad():\n",
    "        for X, y in dataloader:\n",
    "            X, y = X.to(device), y.to(device)\n",
    "            pred = model(X)\n",
    "            test_loss += loss_fn(pred, y).item()\n",
    "            correct += (pred.argmax(1) == y).type(torch.float).sum().item()\n",
    "    test_loss /= num_batches\n",
    "    correct /= size\n",
    "    print(f\"Test Error: \\n Accuracy: {(100*correct):>0.1f}%, Avg loss: {test_loss:>8f} \\n\")\n",
    "    epoch_val_loss.append(test_loss)\n",
    "    epoch_val_acc.append((100*correct))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1\n",
      "-------------------------------\n",
      "loss: 3.321844  [   27/54000]\n",
      "loss: 3.108101  [ 2727/54000]\n",
      "loss: 0.775904  [ 5427/54000]\n",
      "loss: 0.538664  [ 8127/54000]\n",
      "loss: 0.405825  [10827/54000]\n",
      "loss: 0.458463  [13527/54000]\n",
      "loss: 0.248602  [16227/54000]\n",
      "loss: 0.096899  [18927/54000]\n",
      "loss: 0.176965  [21627/54000]\n",
      "loss: 0.009742  [24327/54000]\n",
      "loss: 0.014251  [27027/54000]\n",
      "loss: 0.017586  [29727/54000]\n",
      "loss: 0.017192  [32427/54000]\n",
      "loss: 0.164853  [35127/54000]\n",
      "loss: 0.000760  [37827/54000]\n",
      "loss: 0.004742  [40527/54000]\n",
      "loss: 0.005191  [43227/54000]\n",
      "loss: 0.000192  [45927/54000]\n",
      "loss: 0.037472  [48627/54000]\n",
      "loss: 0.098222  [51327/54000]\n",
      "Test Error: \n",
      " Accuracy: 90.7%, Avg loss: 0.307231 \n",
      "\n",
      "Epoch 2\n",
      "-------------------------------\n",
      "loss: 0.014284  [   27/54000]\n",
      "loss: 0.003541  [ 2727/54000]\n",
      "loss: 0.000394  [ 5427/54000]\n",
      "loss: 0.000692  [ 8127/54000]\n",
      "loss: 0.001378  [10827/54000]\n",
      "loss: 0.010891  [13527/54000]\n",
      "loss: 0.003209  [16227/54000]\n",
      "loss: 0.000144  [18927/54000]\n",
      "loss: 0.001022  [21627/54000]\n",
      "loss: 0.000866  [24327/54000]\n",
      "loss: 0.001434  [27027/54000]\n",
      "loss: 0.010385  [29727/54000]\n",
      "loss: 0.001434  [32427/54000]\n",
      "loss: 0.000703  [35127/54000]\n",
      "loss: 0.000487  [37827/54000]\n",
      "loss: 0.003930  [40527/54000]\n",
      "loss: 0.000729  [43227/54000]\n",
      "loss: 0.019933  [45927/54000]\n",
      "loss: 0.004690  [48627/54000]\n",
      "loss: 0.000776  [51327/54000]\n",
      "Test Error: \n",
      " Accuracy: 93.4%, Avg loss: 0.227085 \n",
      "\n",
      "Epoch 3\n",
      "-------------------------------\n",
      "loss: 0.000018  [   27/54000]\n",
      "loss: 0.001233  [ 2727/54000]\n",
      "loss: 0.000085  [ 5427/54000]\n",
      "loss: 0.000072  [ 8127/54000]\n",
      "loss: 0.000168  [10827/54000]\n",
      "loss: 0.000367  [13527/54000]\n",
      "loss: 0.000514  [16227/54000]\n",
      "loss: 0.000092  [18927/54000]\n",
      "loss: 0.001416  [21627/54000]\n",
      "loss: 0.000854  [24327/54000]\n",
      "loss: 0.000126  [27027/54000]\n",
      "loss: 0.000382  [29727/54000]\n",
      "loss: 0.000981  [32427/54000]\n",
      "loss: 0.001556  [35127/54000]\n",
      "loss: 0.003509  [37827/54000]\n",
      "loss: 0.000061  [40527/54000]\n",
      "loss: 0.000017  [43227/54000]\n",
      "loss: 0.002943  [45927/54000]\n",
      "loss: 0.000126  [48627/54000]\n",
      "loss: 0.000167  [51327/54000]\n",
      "Test Error: \n",
      " Accuracy: 95.4%, Avg loss: 0.161854 \n",
      "\n",
      "Epoch 4\n",
      "-------------------------------\n",
      "loss: 0.000150  [   27/54000]\n",
      "loss: 0.000022  [ 2727/54000]\n",
      "loss: 0.000086  [ 5427/54000]\n",
      "loss: 0.000052  [ 8127/54000]\n",
      "loss: 0.000082  [10827/54000]\n",
      "loss: 0.000248  [13527/54000]\n",
      "loss: 0.000313  [16227/54000]\n",
      "loss: 0.000014  [18927/54000]\n",
      "loss: 0.000313  [21627/54000]\n",
      "loss: 0.000020  [24327/54000]\n",
      "loss: 0.000029  [27027/54000]\n",
      "loss: 0.000216  [29727/54000]\n",
      "loss: 0.000008  [32427/54000]\n",
      "loss: 0.000017  [35127/54000]\n",
      "loss: 0.000060  [37827/54000]\n",
      "loss: 0.000040  [40527/54000]\n",
      "loss: 0.000047  [43227/54000]\n",
      "loss: 0.000098  [45927/54000]\n",
      "loss: 0.000097  [48627/54000]\n",
      "loss: 0.000097  [51327/54000]\n",
      "Test Error: \n",
      " Accuracy: 95.6%, Avg loss: 0.161923 \n",
      "\n",
      "Epoch 5\n",
      "-------------------------------\n",
      "loss: 0.000048  [   27/54000]\n",
      "loss: 0.000092  [ 2727/54000]\n",
      "loss: 0.000065  [ 5427/54000]\n",
      "loss: 0.000017  [ 8127/54000]\n",
      "loss: 0.000055  [10827/54000]\n",
      "loss: 0.000031  [13527/54000]\n",
      "loss: 0.000015  [16227/54000]\n",
      "loss: 0.000029  [18927/54000]\n",
      "loss: 0.000006  [21627/54000]\n",
      "loss: 0.000019  [24327/54000]\n",
      "loss: 0.000060  [27027/54000]\n",
      "loss: 0.000014  [29727/54000]\n",
      "loss: 0.000039  [32427/54000]\n",
      "loss: 0.000005  [35127/54000]\n",
      "loss: 0.000008  [37827/54000]\n",
      "loss: 0.000578  [40527/54000]\n",
      "loss: 0.000025  [43227/54000]\n",
      "loss: 0.000018  [45927/54000]\n",
      "loss: 0.000028  [48627/54000]\n",
      "loss: 0.000141  [51327/54000]\n",
      "Test Error: \n",
      " Accuracy: 95.5%, Avg loss: 0.162137 \n",
      "\n",
      "Epoch 6\n",
      "-------------------------------\n",
      "loss: 0.000015  [   27/54000]\n",
      "loss: 0.000048  [ 2727/54000]\n",
      "loss: 0.000251  [ 5427/54000]\n",
      "loss: 0.000013  [ 8127/54000]\n",
      "loss: 0.000264  [10827/54000]\n",
      "loss: 0.000880  [13527/54000]\n",
      "loss: 0.000010  [16227/54000]\n",
      "loss: 0.000378  [18927/54000]\n",
      "loss: 0.000070  [21627/54000]\n",
      "loss: 0.000029  [24327/54000]\n",
      "loss: 0.000022  [27027/54000]\n",
      "loss: 0.000007  [29727/54000]\n",
      "loss: 0.000004  [32427/54000]\n",
      "loss: 0.000153  [35127/54000]\n",
      "loss: 0.000010  [37827/54000]\n",
      "loss: 0.003565  [40527/54000]\n",
      "loss: 0.000036  [43227/54000]\n",
      "loss: 0.000355  [45927/54000]\n",
      "loss: 0.000068  [48627/54000]\n",
      "loss: 0.000217  [51327/54000]\n",
      "Test Error: \n",
      " Accuracy: 95.2%, Avg loss: 0.171455 \n",
      "\n",
      "Epoch 7\n",
      "-------------------------------\n",
      "loss: 0.000005  [   27/54000]\n",
      "loss: 0.000013  [ 2727/54000]\n",
      "loss: 0.000018  [ 5427/54000]\n",
      "loss: 0.000059  [ 8127/54000]\n",
      "loss: 0.000019  [10827/54000]\n",
      "loss: 0.000004  [13527/54000]\n",
      "loss: 0.000048  [16227/54000]\n",
      "loss: 0.000048  [18927/54000]\n",
      "loss: 0.000021  [21627/54000]\n",
      "loss: 0.000030  [24327/54000]\n",
      "loss: 0.000022  [27027/54000]\n",
      "loss: 0.000098  [29727/54000]\n",
      "loss: 0.000003  [32427/54000]\n",
      "loss: 0.000008  [35127/54000]\n",
      "loss: 0.000020  [37827/54000]\n",
      "loss: 0.000011  [40527/54000]\n",
      "loss: 0.000009  [43227/54000]\n",
      "loss: 0.000018  [45927/54000]\n",
      "loss: 0.000005  [48627/54000]\n",
      "loss: 0.000075  [51327/54000]\n",
      "Test Error: \n",
      " Accuracy: 95.4%, Avg loss: 0.170096 \n",
      "\n",
      "Epoch 8\n",
      "-------------------------------\n",
      "loss: 0.000004  [   27/54000]\n",
      "loss: 0.000040  [ 2727/54000]\n",
      "loss: 0.000049  [ 5427/54000]\n",
      "loss: 0.000005  [ 8127/54000]\n",
      "loss: 0.000036  [10827/54000]\n",
      "loss: 0.000007  [13527/54000]\n",
      "loss: 0.000019  [16227/54000]\n",
      "loss: 0.000032  [18927/54000]\n",
      "loss: 0.000338  [21627/54000]\n",
      "loss: 0.000032  [24327/54000]\n",
      "loss: 0.000001  [27027/54000]\n",
      "loss: 0.000007  [29727/54000]\n",
      "loss: 0.000038  [32427/54000]\n",
      "loss: 0.000008  [35127/54000]\n",
      "loss: 0.000014  [37827/54000]\n",
      "loss: 0.000090  [40527/54000]\n",
      "loss: 0.000027  [43227/54000]\n",
      "loss: 0.000001  [45927/54000]\n",
      "loss: 0.000026  [48627/54000]\n",
      "loss: 0.000042  [51327/54000]\n",
      "Test Error: \n",
      " Accuracy: 95.3%, Avg loss: 0.174897 \n",
      "\n",
      "Epoch 9\n",
      "-------------------------------\n",
      "loss: 0.000013  [   27/54000]\n",
      "loss: 0.000007  [ 2727/54000]\n",
      "loss: 0.000131  [ 5427/54000]\n",
      "loss: 0.000002  [ 8127/54000]\n",
      "loss: 0.000107  [10827/54000]\n",
      "loss: 0.000138  [13527/54000]\n",
      "loss: 0.000024  [16227/54000]\n",
      "loss: 0.000029  [18927/54000]\n",
      "loss: 0.000005  [21627/54000]\n",
      "loss: 0.000042  [24327/54000]\n",
      "loss: 0.000002  [27027/54000]\n",
      "loss: 0.000004  [29727/54000]\n",
      "loss: 0.000002  [32427/54000]\n",
      "loss: 0.000043  [35127/54000]\n",
      "loss: 0.000017  [37827/54000]\n",
      "loss: 0.000044  [40527/54000]\n",
      "loss: 0.000002  [43227/54000]\n",
      "loss: 0.000005  [45927/54000]\n",
      "loss: 0.000003  [48627/54000]\n",
      "loss: 0.000006  [51327/54000]\n",
      "Test Error: \n",
      " Accuracy: 95.3%, Avg loss: 0.172402 \n",
      "\n",
      "Epoch 10\n",
      "-------------------------------\n",
      "loss: 0.000004  [   27/54000]\n",
      "loss: 0.000034  [ 2727/54000]\n",
      "loss: 0.000003  [ 5427/54000]\n",
      "loss: 0.000024  [ 8127/54000]\n",
      "loss: 0.000042  [10827/54000]\n",
      "loss: 0.000074  [13527/54000]\n",
      "loss: 0.000015  [16227/54000]\n",
      "loss: 0.000006  [18927/54000]\n",
      "loss: 0.000060  [21627/54000]\n",
      "loss: 0.000004  [24327/54000]\n",
      "loss: 0.000003  [27027/54000]\n",
      "loss: 0.000024  [29727/54000]\n",
      "loss: 0.000008  [32427/54000]\n",
      "loss: 0.000006  [35127/54000]\n",
      "loss: 0.000005  [37827/54000]\n",
      "loss: 0.000012  [40527/54000]\n",
      "loss: 0.000032  [43227/54000]\n",
      "loss: 0.000024  [45927/54000]\n",
      "loss: 0.000009  [48627/54000]\n",
      "loss: 0.000003  [51327/54000]\n",
      "Test Error: \n",
      " Accuracy: 95.3%, Avg loss: 0.176877 \n",
      "\n",
      "Epoch 11\n",
      "-------------------------------\n",
      "loss: 0.000004  [   27/54000]\n",
      "loss: 0.000002  [ 2727/54000]\n",
      "loss: 0.000005  [ 5427/54000]\n",
      "loss: 0.000010  [ 8127/54000]\n",
      "loss: 0.000082  [10827/54000]\n",
      "loss: 0.000010  [13527/54000]\n",
      "loss: 0.000037  [16227/54000]\n",
      "loss: 0.000138  [18927/54000]\n",
      "loss: 0.000115  [21627/54000]\n",
      "loss: 0.000013  [24327/54000]\n",
      "loss: 0.000005  [27027/54000]\n",
      "loss: 0.000070  [29727/54000]\n",
      "loss: 0.000002  [32427/54000]\n",
      "loss: 0.000008  [35127/54000]\n",
      "loss: 0.000012  [37827/54000]\n",
      "loss: 0.000004  [40527/54000]\n",
      "loss: 0.000003  [43227/54000]\n",
      "loss: 0.000286  [45927/54000]\n",
      "loss: 0.000009  [48627/54000]\n",
      "loss: 0.000014  [51327/54000]\n",
      "Test Error: \n",
      " Accuracy: 95.2%, Avg loss: 0.177529 \n",
      "\n",
      "Epoch 12\n",
      "-------------------------------\n",
      "loss: 0.000128  [   27/54000]\n",
      "loss: 0.000381  [ 2727/54000]\n",
      "loss: 0.000014  [ 5427/54000]\n",
      "loss: 0.000003  [ 8127/54000]\n",
      "loss: 0.000028  [10827/54000]\n",
      "loss: 0.000003  [13527/54000]\n",
      "loss: 0.000003  [16227/54000]\n",
      "loss: 0.000002  [18927/54000]\n",
      "loss: 0.000009  [21627/54000]\n",
      "loss: 0.000415  [24327/54000]\n",
      "loss: 0.000026  [27027/54000]\n",
      "loss: 0.000019  [29727/54000]\n",
      "loss: 0.000094  [32427/54000]\n",
      "loss: 0.000000  [35127/54000]\n",
      "loss: 0.000015  [37827/54000]\n",
      "loss: 0.000002  [40527/54000]\n",
      "loss: 0.000013  [43227/54000]\n",
      "loss: 0.000060  [45927/54000]\n",
      "loss: 0.000008  [48627/54000]\n",
      "loss: 0.000006  [51327/54000]\n",
      "Test Error: \n",
      " Accuracy: 95.2%, Avg loss: 0.178745 \n",
      "\n",
      "Epoch 13\n",
      "-------------------------------\n",
      "loss: 0.000010  [   27/54000]\n",
      "loss: 0.000006  [ 2727/54000]\n",
      "loss: 0.000031  [ 5427/54000]\n",
      "loss: 0.000014  [ 8127/54000]\n",
      "loss: 0.000011  [10827/54000]\n",
      "loss: 0.000048  [13527/54000]\n",
      "loss: 0.000145  [16227/54000]\n",
      "loss: 0.000012  [18927/54000]\n",
      "loss: 0.000001  [21627/54000]\n",
      "loss: 0.000002  [24327/54000]\n",
      "loss: 0.000010  [27027/54000]\n",
      "loss: 0.000003  [29727/54000]\n",
      "loss: 0.000088  [32427/54000]\n",
      "loss: 0.000023  [35127/54000]\n",
      "loss: 0.000001  [37827/54000]\n",
      "loss: 0.000005  [40527/54000]\n",
      "loss: 0.000010  [43227/54000]\n",
      "loss: 0.000003  [45927/54000]\n",
      "loss: 0.000088  [48627/54000]\n",
      "loss: 0.000001  [51327/54000]\n",
      "Test Error: \n",
      " Accuracy: 95.2%, Avg loss: 0.180974 \n",
      "\n",
      "Epoch 14\n",
      "-------------------------------\n",
      "loss: 0.000012  [   27/54000]\n",
      "loss: 0.000001  [ 2727/54000]\n",
      "loss: 0.000478  [ 5427/54000]\n",
      "loss: 0.000098  [ 8127/54000]\n",
      "loss: 0.000001  [10827/54000]\n",
      "loss: 0.000037  [13527/54000]\n",
      "loss: 0.000001  [16227/54000]\n",
      "loss: 0.000002  [18927/54000]\n",
      "loss: 0.000001  [21627/54000]\n",
      "loss: 0.000002  [24327/54000]\n",
      "loss: 0.000001  [27027/54000]\n",
      "loss: 0.000052  [29727/54000]\n",
      "loss: 0.000068  [32427/54000]\n",
      "loss: 0.000009  [35127/54000]\n",
      "loss: 0.000012  [37827/54000]\n",
      "loss: 0.000037  [40527/54000]\n",
      "loss: 0.000006  [43227/54000]\n",
      "loss: 0.000002  [45927/54000]\n",
      "loss: 0.000004  [48627/54000]\n",
      "loss: 0.000007  [51327/54000]\n",
      "Test Error: \n",
      " Accuracy: 95.1%, Avg loss: 0.184248 \n",
      "\n",
      "Epoch 15\n",
      "-------------------------------\n",
      "loss: 0.000001  [   27/54000]\n",
      "loss: 0.000002  [ 2727/54000]\n",
      "loss: 0.000002  [ 5427/54000]\n",
      "loss: 0.000006  [ 8127/54000]\n",
      "loss: 0.000009  [10827/54000]\n",
      "loss: 0.000005  [13527/54000]\n",
      "loss: 0.000011  [16227/54000]\n",
      "loss: 0.000005  [18927/54000]\n",
      "loss: 0.000001  [21627/54000]\n",
      "loss: 0.000015  [24327/54000]\n",
      "loss: 0.000001  [27027/54000]\n",
      "loss: 0.000035  [29727/54000]\n",
      "loss: 0.000191  [32427/54000]\n",
      "loss: 0.000115  [35127/54000]\n",
      "loss: 0.000001  [37827/54000]\n",
      "loss: 0.000142  [40527/54000]\n",
      "loss: 0.000001  [43227/54000]\n",
      "loss: 0.000007  [45927/54000]\n",
      "loss: 0.000104  [48627/54000]\n",
      "loss: 0.000037  [51327/54000]\n",
      "Test Error: \n",
      " Accuracy: 95.1%, Avg loss: 0.184514 \n",
      "\n",
      "Epoch 16\n",
      "-------------------------------\n",
      "loss: 0.000004  [   27/54000]\n",
      "loss: 0.000011  [ 2727/54000]\n",
      "loss: 0.000002  [ 5427/54000]\n",
      "loss: 0.000005  [ 8127/54000]\n",
      "loss: 0.000013  [10827/54000]\n",
      "loss: 0.000007  [13527/54000]\n",
      "loss: 0.000006  [16227/54000]\n",
      "loss: 0.000001  [18927/54000]\n",
      "loss: 0.000006  [21627/54000]\n",
      "loss: 0.000028  [24327/54000]\n",
      "loss: 0.000021  [27027/54000]\n",
      "loss: 0.000027  [29727/54000]\n",
      "loss: 0.000010  [32427/54000]\n",
      "loss: 0.000005  [35127/54000]\n",
      "loss: 0.000026  [37827/54000]\n",
      "loss: 0.000004  [40527/54000]\n",
      "loss: 0.000043  [43227/54000]\n",
      "loss: 0.000000  [45927/54000]\n",
      "loss: 0.000004  [48627/54000]\n",
      "loss: 0.000033  [51327/54000]\n",
      "Test Error: \n",
      " Accuracy: 95.1%, Avg loss: 0.186315 \n",
      "\n",
      "Epoch 17\n",
      "-------------------------------\n",
      "loss: 0.000018  [   27/54000]\n",
      "loss: 0.000001  [ 2727/54000]\n",
      "loss: 0.000004  [ 5427/54000]\n",
      "loss: 0.000010  [ 8127/54000]\n",
      "loss: 0.000054  [10827/54000]\n",
      "loss: 0.000016  [13527/54000]\n",
      "loss: 0.000007  [16227/54000]\n",
      "loss: 0.000005  [18927/54000]\n",
      "loss: 0.000005  [21627/54000]\n",
      "loss: 0.000193  [24327/54000]\n",
      "loss: 0.000018  [27027/54000]\n",
      "loss: 0.000013  [29727/54000]\n",
      "loss: 0.000003  [32427/54000]\n",
      "loss: 0.000003  [35127/54000]\n",
      "loss: 0.000014  [37827/54000]\n",
      "loss: 0.000023  [40527/54000]\n",
      "loss: 0.000001  [43227/54000]\n",
      "loss: 0.000010  [45927/54000]\n",
      "loss: 0.000010  [48627/54000]\n",
      "loss: 0.000012  [51327/54000]\n",
      "Test Error: \n",
      " Accuracy: 95.1%, Avg loss: 0.185929 \n",
      "\n",
      "Epoch 18\n",
      "-------------------------------\n",
      "loss: 0.000035  [   27/54000]\n",
      "loss: 0.000018  [ 2727/54000]\n",
      "loss: 0.000009  [ 5427/54000]\n",
      "loss: 0.000003  [ 8127/54000]\n",
      "loss: 0.000005  [10827/54000]\n",
      "loss: 0.000001  [13527/54000]\n",
      "loss: 0.000003  [16227/54000]\n",
      "loss: 0.000004  [18927/54000]\n",
      "loss: 0.000024  [21627/54000]\n",
      "loss: 0.000003  [24327/54000]\n",
      "loss: 0.000176  [27027/54000]\n",
      "loss: 0.000029  [29727/54000]\n",
      "loss: 0.000181  [32427/54000]\n",
      "loss: 0.000010  [35127/54000]\n",
      "loss: 0.000006  [37827/54000]\n",
      "loss: 0.000001  [40527/54000]\n",
      "loss: 0.000483  [43227/54000]\n",
      "loss: 0.000003  [45927/54000]\n",
      "loss: 0.000031  [48627/54000]\n",
      "loss: 0.000045  [51327/54000]\n",
      "Test Error: \n",
      " Accuracy: 95.2%, Avg loss: 0.184470 \n",
      "\n",
      "Epoch 19\n",
      "-------------------------------\n",
      "loss: 0.000010  [   27/54000]\n",
      "loss: 0.000022  [ 2727/54000]\n",
      "loss: 0.000012  [ 5427/54000]\n",
      "loss: 0.000007  [ 8127/54000]\n",
      "loss: 0.000003  [10827/54000]\n",
      "loss: 0.000021  [13527/54000]\n",
      "loss: 0.000001  [16227/54000]\n",
      "loss: 0.000005  [18927/54000]\n",
      "loss: 0.000003  [21627/54000]\n",
      "loss: 0.000008  [24327/54000]\n",
      "loss: 0.000050  [27027/54000]\n",
      "loss: 0.000143  [29727/54000]\n",
      "loss: 0.000002  [32427/54000]\n",
      "loss: 0.000004  [35127/54000]\n",
      "loss: 0.000001  [37827/54000]\n",
      "loss: 0.000003  [40527/54000]\n",
      "loss: 0.000004  [43227/54000]\n",
      "loss: 0.000010  [45927/54000]\n",
      "loss: 0.000012  [48627/54000]\n",
      "loss: 0.000012  [51327/54000]\n",
      "Test Error: \n",
      " Accuracy: 95.2%, Avg loss: 0.184891 \n",
      "\n",
      "Epoch 20\n",
      "-------------------------------\n",
      "loss: 0.000040  [   27/54000]\n",
      "loss: 0.000025  [ 2727/54000]\n",
      "loss: 0.000001  [ 5427/54000]\n",
      "loss: 0.000170  [ 8127/54000]\n",
      "loss: 0.000040  [10827/54000]\n",
      "loss: 0.000049  [13527/54000]\n",
      "loss: 0.000039  [16227/54000]\n",
      "loss: 0.000001  [18927/54000]\n",
      "loss: 0.000011  [21627/54000]\n",
      "loss: 0.000005  [24327/54000]\n",
      "loss: 0.000003  [27027/54000]\n",
      "loss: 0.000002  [29727/54000]\n",
      "loss: 0.000025  [32427/54000]\n",
      "loss: 0.000000  [35127/54000]\n",
      "loss: 0.000008  [37827/54000]\n",
      "loss: 0.000002  [40527/54000]\n",
      "loss: 0.000006  [43227/54000]\n",
      "loss: 0.000008  [45927/54000]\n",
      "loss: 0.000013  [48627/54000]\n",
      "loss: 0.000022  [51327/54000]\n",
      "Test Error: \n",
      " Accuracy: 95.1%, Avg loss: 0.187494 \n",
      "\n",
      "Done!\n"
     ]
    }
   ],
   "source": [
    "epochs = 5\n",
    "for t in range(epochs):\n",
    "    print(f\"Epoch {t+1}\\n-------------------------------\")\n",
    "    train(train_loader, model, loss_fn, optimizer)\n",
    "    test(valid_loader, model, loss_fn)\n",
    "print(\"Done!\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<function matplotlib.pyplot.show(close=None, block=None)>"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from matplotlib import pyplot as plt\n",
    "plt.plot(epoch_val_loss, label='val_loss')\n",
    "plt.xlabel('Epoch')\n",
    "plt.ylabel('Loss')\n",
    "plt.title('Val. Loss over Epochs')\n",
    "plt.show"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<function matplotlib.pyplot.show(close=None, block=None)>"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(epoch_val_acc,label='val_acc')\n",
    "plt.xlabel('Epoch')\n",
    "plt.ylabel('Acc')\n",
    "plt.title('Val. Acc over Epochs')\n",
    "plt.show"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test Error: \n",
      " Accuracy: 95.4%, Avg loss: 0.193437 \n",
      "\n"
     ]
    }
   ],
   "source": [
    "test(test_loader, model, loss_fn)"
   ]
  }
 ],
 "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": "3.11.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}