{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "import torchvision\n",
    "from torch import nn, optim\n",
    "from torch.autograd import Variable\n",
    "import numpy as np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "mnist_data = torchvision.datasets.MNIST(\n",
    "    \"mnist_data\", train=True, transform=torchvision.transforms.ToTensor(), download=True\n",
    ")\n",
    "mnist_dataloader = torch.utils.data.DataLoader(mnist_data, batch_size=50)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "class Mnet(nn.Module):\n",
    "    def __init__(self):\n",
    "        super(Mnet, self).__init__()\n",
    "        self.linear1 = nn.Linear(28 * 28, 400)\n",
    "        self.linear2 = nn.Linear(400, 200)\n",
    "        self.linear3 = nn.Linear(200, 100)\n",
    "        self.linear4 = nn.Linear(100, 50)\n",
    "        self.linear5 = nn.Linear(50, 25)\n",
    "        self.final_linear = nn.Linear(25, 10)\n",
    "\n",
    "        self.relu = nn.ReLU()\n",
    "\n",
    "    def forward(self, images):\n",
    "        x = images.view(-1, 28 * 28)\n",
    "        x = self.relu(self.linear1(x))\n",
    "        x = self.relu(self.linear2(x))\n",
    "        x = self.relu(self.linear3(x))\n",
    "        x = self.relu(self.linear4(x))\n",
    "        x = self.relu(self.linear5(x))\n",
    "        x = self.final_linear(x)\n",
    "        return x"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 50/50 [21:18<00:00, 25.57s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "final loss: 1.1586851087486139e-06\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "from tqdm import tqdm\n",
    "model = Mnet()\n",
    "cec_loss = nn.CrossEntropyLoss()\n",
    "params = model.parameters()\n",
    "optimizer = optim.Adam(params=params, lr=0.001)\n",
    "\n",
    "n_epochs = 50\n",
    "n_iterations = 0\n",
    "\n",
    "for e in tqdm(range(n_epochs)):\n",
    "    for i, (images, labels) in enumerate(mnist_dataloader):\n",
    "        output = model(images)\n",
    "\n",
    "        model.zero_grad()\n",
    "        loss = cec_loss(output, labels)\n",
    "        loss.backward()\n",
    "\n",
    "        optimizer.step()\n",
    "        n_iterations+=1\n",
    "\n",
    "print(f'final loss: {loss.item()}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "torch.save(model, \"mnistmodel.pt\")"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": ".venv",
   "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": "3.10.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}