wehub-resource-sync
545 lines
21 KiB
Plaintext
545 lines
21 KiB
Plaintext
{
|
|
"cells": [
|
|
{
|
|
"cell_type": "markdown",
|
|
"metadata": {},
|
|
"source": [
|
|
"Deep Learning Models -- A collection of various deep learning architectures, models, and tips for TensorFlow and PyTorch in Jupyter Notebooks.\n",
|
|
"- Author: Sebastian Raschka\n",
|
|
"- GitHub Repository: https://github.com/rasbt/deeplearning-models"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": 1,
|
|
"metadata": {},
|
|
"outputs": [
|
|
{
|
|
"name": "stdout",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"Sebastian Raschka \n",
|
|
"\n",
|
|
"CPython 3.6.8\n",
|
|
"IPython 7.2.0\n",
|
|
"\n",
|
|
"torch 1.0.0\n"
|
|
]
|
|
}
|
|
],
|
|
"source": [
|
|
"%load_ext watermark\n",
|
|
"%watermark -a 'Sebastian Raschka' -v -p torch"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "markdown",
|
|
"metadata": {},
|
|
"source": [
|
|
"- Runs on CPU or GPU (if available)"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "markdown",
|
|
"metadata": {},
|
|
"source": [
|
|
"# Model Zoo -- Multilayer Perceptron"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "markdown",
|
|
"metadata": {},
|
|
"source": [
|
|
"## Imports"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": 2,
|
|
"metadata": {},
|
|
"outputs": [],
|
|
"source": [
|
|
"import time\n",
|
|
"import numpy as np\n",
|
|
"from torchvision import datasets\n",
|
|
"from torchvision import transforms\n",
|
|
"from torch.utils.data import DataLoader\n",
|
|
"import torch.nn.functional as F\n",
|
|
"import torch\n",
|
|
"\n",
|
|
"\n",
|
|
"if torch.cuda.is_available():\n",
|
|
" torch.backends.cudnn.deterministic = True"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "markdown",
|
|
"metadata": {},
|
|
"source": [
|
|
"## Settings and Dataset"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": 3,
|
|
"metadata": {},
|
|
"outputs": [
|
|
{
|
|
"name": "stdout",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"Image batch dimensions: torch.Size([64, 1, 28, 28])\n",
|
|
"Image label dimensions: torch.Size([64])\n"
|
|
]
|
|
}
|
|
],
|
|
"source": [
|
|
"##########################\n",
|
|
"### SETTINGS\n",
|
|
"##########################\n",
|
|
"\n",
|
|
"# Device\n",
|
|
"device = torch.device(\"cuda:3\" if torch.cuda.is_available() else \"cpu\")\n",
|
|
"\n",
|
|
"# Hyperparameters\n",
|
|
"random_seed = 1\n",
|
|
"learning_rate = 0.1\n",
|
|
"num_epochs = 10\n",
|
|
"batch_size = 64\n",
|
|
"\n",
|
|
"# Architecture\n",
|
|
"num_features = 784\n",
|
|
"num_hidden_1 = 128\n",
|
|
"num_hidden_2 = 256\n",
|
|
"num_classes = 10\n",
|
|
"\n",
|
|
"\n",
|
|
"##########################\n",
|
|
"### MNIST DATASET\n",
|
|
"##########################\n",
|
|
"\n",
|
|
"# Note transforms.ToTensor() scales input images\n",
|
|
"# to 0-1 range\n",
|
|
"train_dataset = datasets.MNIST(root='data', \n",
|
|
" train=True, \n",
|
|
" transform=transforms.ToTensor(),\n",
|
|
" download=True)\n",
|
|
"\n",
|
|
"test_dataset = datasets.MNIST(root='data', \n",
|
|
" train=False, \n",
|
|
" transform=transforms.ToTensor())\n",
|
|
"\n",
|
|
"\n",
|
|
"train_loader = DataLoader(dataset=train_dataset, \n",
|
|
" batch_size=batch_size, \n",
|
|
" shuffle=True)\n",
|
|
"\n",
|
|
"test_loader = DataLoader(dataset=test_dataset, \n",
|
|
" batch_size=batch_size, \n",
|
|
" shuffle=False)\n",
|
|
"\n",
|
|
"# Checking the dataset\n",
|
|
"for images, labels in train_loader: \n",
|
|
" print('Image batch dimensions:', images.shape)\n",
|
|
" print('Image label dimensions:', labels.shape)\n",
|
|
" break"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": 4,
|
|
"metadata": {},
|
|
"outputs": [],
|
|
"source": [
|
|
"##########################\n",
|
|
"### MODEL\n",
|
|
"##########################\n",
|
|
"\n",
|
|
"class MultilayerPerceptron(torch.nn.Module):\n",
|
|
"\n",
|
|
" def __init__(self, num_features, num_classes):\n",
|
|
" super(MultilayerPerceptron, self).__init__()\n",
|
|
" \n",
|
|
" ### 1st hidden layer\n",
|
|
" self.linear_1 = torch.nn.Linear(num_features, num_hidden_1)\n",
|
|
" # The following two lines are not necessary, \n",
|
|
" # but used here to demonstrate how to access the weights\n",
|
|
" # and use a different weight initialization.\n",
|
|
" # By default, PyTorch uses Xavier/Glorot initialization, which\n",
|
|
" # should usually be preferred.\n",
|
|
" self.linear_1.weight.detach().normal_(0.0, 0.1)\n",
|
|
" self.linear_1.bias.detach().zero_()\n",
|
|
" \n",
|
|
" ### 2nd hidden layer\n",
|
|
" self.linear_2 = torch.nn.Linear(num_hidden_1, num_hidden_2)\n",
|
|
" self.linear_2.weight.detach().normal_(0.0, 0.1)\n",
|
|
" self.linear_2.bias.detach().zero_()\n",
|
|
" \n",
|
|
" ### Output layer\n",
|
|
" self.linear_out = torch.nn.Linear(num_hidden_2, num_classes)\n",
|
|
" self.linear_out.weight.detach().normal_(0.0, 0.1)\n",
|
|
" self.linear_out.bias.detach().zero_()\n",
|
|
" \n",
|
|
" def forward(self, x):\n",
|
|
" out = self.linear_1(x)\n",
|
|
" out = F.relu(out)\n",
|
|
" out = self.linear_2(out)\n",
|
|
" out = F.relu(out)\n",
|
|
" logits = self.linear_out(out)\n",
|
|
" probas = F.log_softmax(logits, dim=1)\n",
|
|
" return logits, probas\n",
|
|
"\n",
|
|
" \n",
|
|
"torch.manual_seed(random_seed)\n",
|
|
"model = MultilayerPerceptron(num_features=num_features,\n",
|
|
" num_classes=num_classes)\n",
|
|
"\n",
|
|
"model = model.to(device)\n",
|
|
"\n",
|
|
"optimizer = torch.optim.SGD(model.parameters(), lr=learning_rate) "
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": 5,
|
|
"metadata": {},
|
|
"outputs": [
|
|
{
|
|
"name": "stdout",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"Epoch: 001/010 | Batch 000/938 | Cost: 2.4231\n",
|
|
"Epoch: 001/010 | Batch 050/938 | Cost: 0.7213\n",
|
|
"Epoch: 001/010 | Batch 100/938 | Cost: 0.3137\n",
|
|
"Epoch: 001/010 | Batch 150/938 | Cost: 0.4774\n",
|
|
"Epoch: 001/010 | Batch 200/938 | Cost: 0.3311\n",
|
|
"Epoch: 001/010 | Batch 250/938 | Cost: 0.3082\n",
|
|
"Epoch: 001/010 | Batch 300/938 | Cost: 0.3940\n",
|
|
"Epoch: 001/010 | Batch 350/938 | Cost: 0.2292\n",
|
|
"Epoch: 001/010 | Batch 400/938 | Cost: 0.0837\n",
|
|
"Epoch: 001/010 | Batch 450/938 | Cost: 0.2238\n",
|
|
"Epoch: 001/010 | Batch 500/938 | Cost: 0.2501\n",
|
|
"Epoch: 001/010 | Batch 550/938 | Cost: 0.1543\n",
|
|
"Epoch: 001/010 | Batch 600/938 | Cost: 0.2353\n",
|
|
"Epoch: 001/010 | Batch 650/938 | Cost: 0.4138\n",
|
|
"Epoch: 001/010 | Batch 700/938 | Cost: 0.2955\n",
|
|
"Epoch: 001/010 | Batch 750/938 | Cost: 0.3486\n",
|
|
"Epoch: 001/010 | Batch 800/938 | Cost: 0.1391\n",
|
|
"Epoch: 001/010 | Batch 850/938 | Cost: 0.0675\n",
|
|
"Epoch: 001/010 | Batch 900/938 | Cost: 0.2827\n",
|
|
"Epoch: 001/010 training accuracy: 94.90%\n",
|
|
"Time elapsed: 0.22 min\n",
|
|
"Epoch: 002/010 | Batch 000/938 | Cost: 0.3237\n",
|
|
"Epoch: 002/010 | Batch 050/938 | Cost: 0.2792\n",
|
|
"Epoch: 002/010 | Batch 100/938 | Cost: 0.0915\n",
|
|
"Epoch: 002/010 | Batch 150/938 | Cost: 0.2269\n",
|
|
"Epoch: 002/010 | Batch 200/938 | Cost: 0.2248\n",
|
|
"Epoch: 002/010 | Batch 250/938 | Cost: 0.1607\n",
|
|
"Epoch: 002/010 | Batch 300/938 | Cost: 0.0551\n",
|
|
"Epoch: 002/010 | Batch 350/938 | Cost: 0.2369\n",
|
|
"Epoch: 002/010 | Batch 400/938 | Cost: 0.0538\n",
|
|
"Epoch: 002/010 | Batch 450/938 | Cost: 0.0951\n",
|
|
"Epoch: 002/010 | Batch 500/938 | Cost: 0.2423\n",
|
|
"Epoch: 002/010 | Batch 550/938 | Cost: 0.2082\n",
|
|
"Epoch: 002/010 | Batch 600/938 | Cost: 0.1623\n",
|
|
"Epoch: 002/010 | Batch 650/938 | Cost: 0.1743\n",
|
|
"Epoch: 002/010 | Batch 700/938 | Cost: 0.1987\n",
|
|
"Epoch: 002/010 | Batch 750/938 | Cost: 0.0871\n",
|
|
"Epoch: 002/010 | Batch 800/938 | Cost: 0.1662\n",
|
|
"Epoch: 002/010 | Batch 850/938 | Cost: 0.1087\n",
|
|
"Epoch: 002/010 | Batch 900/938 | Cost: 0.1574\n",
|
|
"Epoch: 002/010 training accuracy: 96.62%\n",
|
|
"Time elapsed: 0.45 min\n",
|
|
"Epoch: 003/010 | Batch 000/938 | Cost: 0.1088\n",
|
|
"Epoch: 003/010 | Batch 050/938 | Cost: 0.0994\n",
|
|
"Epoch: 003/010 | Batch 100/938 | Cost: 0.0771\n",
|
|
"Epoch: 003/010 | Batch 150/938 | Cost: 0.1926\n",
|
|
"Epoch: 003/010 | Batch 200/938 | Cost: 0.1231\n",
|
|
"Epoch: 003/010 | Batch 250/938 | Cost: 0.2532\n",
|
|
"Epoch: 003/010 | Batch 300/938 | Cost: 0.0637\n",
|
|
"Epoch: 003/010 | Batch 350/938 | Cost: 0.1193\n",
|
|
"Epoch: 003/010 | Batch 400/938 | Cost: 0.0839\n",
|
|
"Epoch: 003/010 | Batch 450/938 | Cost: 0.0830\n",
|
|
"Epoch: 003/010 | Batch 500/938 | Cost: 0.0689\n",
|
|
"Epoch: 003/010 | Batch 550/938 | Cost: 0.0504\n",
|
|
"Epoch: 003/010 | Batch 600/938 | Cost: 0.1999\n",
|
|
"Epoch: 003/010 | Batch 650/938 | Cost: 0.1190\n",
|
|
"Epoch: 003/010 | Batch 700/938 | Cost: 0.2022\n",
|
|
"Epoch: 003/010 | Batch 750/938 | Cost: 0.0598\n",
|
|
"Epoch: 003/010 | Batch 800/938 | Cost: 0.0216\n",
|
|
"Epoch: 003/010 | Batch 850/938 | Cost: 0.0174\n",
|
|
"Epoch: 003/010 | Batch 900/938 | Cost: 0.0660\n",
|
|
"Epoch: 003/010 training accuracy: 97.70%\n",
|
|
"Time elapsed: 0.67 min\n",
|
|
"Epoch: 004/010 | Batch 000/938 | Cost: 0.0067\n",
|
|
"Epoch: 004/010 | Batch 050/938 | Cost: 0.0344\n",
|
|
"Epoch: 004/010 | Batch 100/938 | Cost: 0.1624\n",
|
|
"Epoch: 004/010 | Batch 150/938 | Cost: 0.0572\n",
|
|
"Epoch: 004/010 | Batch 200/938 | Cost: 0.0306\n",
|
|
"Epoch: 004/010 | Batch 250/938 | Cost: 0.1564\n",
|
|
"Epoch: 004/010 | Batch 300/938 | Cost: 0.2233\n",
|
|
"Epoch: 004/010 | Batch 350/938 | Cost: 0.0229\n",
|
|
"Epoch: 004/010 | Batch 400/938 | Cost: 0.1174\n",
|
|
"Epoch: 004/010 | Batch 450/938 | Cost: 0.1853\n",
|
|
"Epoch: 004/010 | Batch 500/938 | Cost: 0.1418\n",
|
|
"Epoch: 004/010 | Batch 550/938 | Cost: 0.1071\n",
|
|
"Epoch: 004/010 | Batch 600/938 | Cost: 0.0354\n",
|
|
"Epoch: 004/010 | Batch 650/938 | Cost: 0.0487\n",
|
|
"Epoch: 004/010 | Batch 700/938 | Cost: 0.1886\n",
|
|
"Epoch: 004/010 | Batch 750/938 | Cost: 0.1568\n",
|
|
"Epoch: 004/010 | Batch 800/938 | Cost: 0.0702\n",
|
|
"Epoch: 004/010 | Batch 850/938 | Cost: 0.0533\n",
|
|
"Epoch: 004/010 | Batch 900/938 | Cost: 0.1500\n",
|
|
"Epoch: 004/010 training accuracy: 98.01%\n",
|
|
"Time elapsed: 0.89 min\n",
|
|
"Epoch: 005/010 | Batch 000/938 | Cost: 0.0225\n",
|
|
"Epoch: 005/010 | Batch 050/938 | Cost: 0.0168\n",
|
|
"Epoch: 005/010 | Batch 100/938 | Cost: 0.1133\n",
|
|
"Epoch: 005/010 | Batch 150/938 | Cost: 0.0691\n",
|
|
"Epoch: 005/010 | Batch 200/938 | Cost: 0.0413\n",
|
|
"Epoch: 005/010 | Batch 250/938 | Cost: 0.0840\n",
|
|
"Epoch: 005/010 | Batch 300/938 | Cost: 0.0697\n",
|
|
"Epoch: 005/010 | Batch 350/938 | Cost: 0.0901\n",
|
|
"Epoch: 005/010 | Batch 400/938 | Cost: 0.0370\n",
|
|
"Epoch: 005/010 | Batch 450/938 | Cost: 0.0514\n",
|
|
"Epoch: 005/010 | Batch 500/938 | Cost: 0.1403\n",
|
|
"Epoch: 005/010 | Batch 550/938 | Cost: 0.1164\n",
|
|
"Epoch: 005/010 | Batch 600/938 | Cost: 0.0624\n",
|
|
"Epoch: 005/010 | Batch 650/938 | Cost: 0.0280\n",
|
|
"Epoch: 005/010 | Batch 700/938 | Cost: 0.0555\n",
|
|
"Epoch: 005/010 | Batch 750/938 | Cost: 0.0432\n",
|
|
"Epoch: 005/010 | Batch 800/938 | Cost: 0.0434\n",
|
|
"Epoch: 005/010 | Batch 850/938 | Cost: 0.1074\n",
|
|
"Epoch: 005/010 | Batch 900/938 | Cost: 0.0353\n",
|
|
"Epoch: 005/010 training accuracy: 98.56%\n",
|
|
"Time elapsed: 1.12 min\n",
|
|
"Epoch: 006/010 | Batch 000/938 | Cost: 0.1073\n",
|
|
"Epoch: 006/010 | Batch 050/938 | Cost: 0.0901\n",
|
|
"Epoch: 006/010 | Batch 100/938 | Cost: 0.0736\n",
|
|
"Epoch: 006/010 | Batch 150/938 | Cost: 0.0290\n",
|
|
"Epoch: 006/010 | Batch 200/938 | Cost: 0.0399\n",
|
|
"Epoch: 006/010 | Batch 250/938 | Cost: 0.0720\n",
|
|
"Epoch: 006/010 | Batch 300/938 | Cost: 0.0863\n",
|
|
"Epoch: 006/010 | Batch 350/938 | Cost: 0.0629\n",
|
|
"Epoch: 006/010 | Batch 400/938 | Cost: 0.1095\n",
|
|
"Epoch: 006/010 | Batch 450/938 | Cost: 0.0531\n",
|
|
"Epoch: 006/010 | Batch 500/938 | Cost: 0.0680\n",
|
|
"Epoch: 006/010 | Batch 550/938 | Cost: 0.1777\n",
|
|
"Epoch: 006/010 | Batch 600/938 | Cost: 0.0525\n",
|
|
"Epoch: 006/010 | Batch 650/938 | Cost: 0.0364\n",
|
|
"Epoch: 006/010 | Batch 700/938 | Cost: 0.0836\n",
|
|
"Epoch: 006/010 | Batch 750/938 | Cost: 0.1251\n",
|
|
"Epoch: 006/010 | Batch 800/938 | Cost: 0.0638\n",
|
|
"Epoch: 006/010 | Batch 850/938 | Cost: 0.0725\n",
|
|
"Epoch: 006/010 | Batch 900/938 | Cost: 0.1785\n",
|
|
"Epoch: 006/010 training accuracy: 98.88%\n",
|
|
"Time elapsed: 1.34 min\n",
|
|
"Epoch: 007/010 | Batch 000/938 | Cost: 0.0043\n",
|
|
"Epoch: 007/010 | Batch 050/938 | Cost: 0.0928\n",
|
|
"Epoch: 007/010 | Batch 100/938 | Cost: 0.0375\n",
|
|
"Epoch: 007/010 | Batch 150/938 | Cost: 0.1094\n",
|
|
"Epoch: 007/010 | Batch 200/938 | Cost: 0.0237\n",
|
|
"Epoch: 007/010 | Batch 250/938 | Cost: 0.0398\n",
|
|
"Epoch: 007/010 | Batch 300/938 | Cost: 0.0592\n",
|
|
"Epoch: 007/010 | Batch 350/938 | Cost: 0.0340\n",
|
|
"Epoch: 007/010 | Batch 400/938 | Cost: 0.0255\n",
|
|
"Epoch: 007/010 | Batch 450/938 | Cost: 0.0472\n",
|
|
"Epoch: 007/010 | Batch 500/938 | Cost: 0.0172\n",
|
|
"Epoch: 007/010 | Batch 550/938 | Cost: 0.0860\n",
|
|
"Epoch: 007/010 | Batch 600/938 | Cost: 0.1124\n",
|
|
"Epoch: 007/010 | Batch 650/938 | Cost: 0.0783\n",
|
|
"Epoch: 007/010 | Batch 700/938 | Cost: 0.0307\n",
|
|
"Epoch: 007/010 | Batch 750/938 | Cost: 0.0686\n",
|
|
"Epoch: 007/010 | Batch 800/938 | Cost: 0.0128\n",
|
|
"Epoch: 007/010 | Batch 850/938 | Cost: 0.0684\n",
|
|
"Epoch: 007/010 | Batch 900/938 | Cost: 0.0541\n",
|
|
"Epoch: 007/010 training accuracy: 98.98%\n",
|
|
"Time elapsed: 1.56 min\n",
|
|
"Epoch: 008/010 | Batch 000/938 | Cost: 0.0237\n",
|
|
"Epoch: 008/010 | Batch 050/938 | Cost: 0.0167\n",
|
|
"Epoch: 008/010 | Batch 100/938 | Cost: 0.0446\n",
|
|
"Epoch: 008/010 | Batch 150/938 | Cost: 0.0324\n",
|
|
"Epoch: 008/010 | Batch 200/938 | Cost: 0.0068\n",
|
|
"Epoch: 008/010 | Batch 250/938 | Cost: 0.0529\n",
|
|
"Epoch: 008/010 | Batch 300/938 | Cost: 0.0278\n",
|
|
"Epoch: 008/010 | Batch 350/938 | Cost: 0.0289\n",
|
|
"Epoch: 008/010 | Batch 400/938 | Cost: 0.0099\n",
|
|
"Epoch: 008/010 | Batch 450/938 | Cost: 0.0052\n",
|
|
"Epoch: 008/010 | Batch 500/938 | Cost: 0.0116\n",
|
|
"Epoch: 008/010 | Batch 550/938 | Cost: 0.0049\n",
|
|
"Epoch: 008/010 | Batch 600/938 | Cost: 0.0214\n",
|
|
"Epoch: 008/010 | Batch 650/938 | Cost: 0.0397\n",
|
|
"Epoch: 008/010 | Batch 700/938 | Cost: 0.0494\n",
|
|
"Epoch: 008/010 | Batch 750/938 | Cost: 0.0166\n",
|
|
"Epoch: 008/010 | Batch 800/938 | Cost: 0.0482\n",
|
|
"Epoch: 008/010 | Batch 850/938 | Cost: 0.0049\n",
|
|
"Epoch: 008/010 | Batch 900/938 | Cost: 0.0482\n",
|
|
"Epoch: 008/010 training accuracy: 99.25%\n",
|
|
"Time elapsed: 1.78 min\n",
|
|
"Epoch: 009/010 | Batch 000/938 | Cost: 0.0367\n",
|
|
"Epoch: 009/010 | Batch 050/938 | Cost: 0.0624\n",
|
|
"Epoch: 009/010 | Batch 100/938 | Cost: 0.0095\n",
|
|
"Epoch: 009/010 | Batch 150/938 | Cost: 0.0166\n",
|
|
"Epoch: 009/010 | Batch 200/938 | Cost: 0.0764\n",
|
|
"Epoch: 009/010 | Batch 250/938 | Cost: 0.0056\n",
|
|
"Epoch: 009/010 | Batch 300/938 | Cost: 0.0066\n",
|
|
"Epoch: 009/010 | Batch 350/938 | Cost: 0.0060\n",
|
|
"Epoch: 009/010 | Batch 400/938 | Cost: 0.0583\n",
|
|
"Epoch: 009/010 | Batch 450/938 | Cost: 0.0170\n",
|
|
"Epoch: 009/010 | Batch 500/938 | Cost: 0.0252\n",
|
|
"Epoch: 009/010 | Batch 550/938 | Cost: 0.0140\n",
|
|
"Epoch: 009/010 | Batch 600/938 | Cost: 0.0911\n",
|
|
"Epoch: 009/010 | Batch 650/938 | Cost: 0.0469\n",
|
|
"Epoch: 009/010 | Batch 700/938 | Cost: 0.0146\n",
|
|
"Epoch: 009/010 | Batch 750/938 | Cost: 0.0573\n",
|
|
"Epoch: 009/010 | Batch 800/938 | Cost: 0.0067\n",
|
|
"Epoch: 009/010 | Batch 850/938 | Cost: 0.1513\n",
|
|
"Epoch: 009/010 | Batch 900/938 | Cost: 0.0029\n",
|
|
"Epoch: 009/010 training accuracy: 99.30%\n",
|
|
"Time elapsed: 2.01 min\n",
|
|
"Epoch: 010/010 | Batch 000/938 | Cost: 0.0207\n",
|
|
"Epoch: 010/010 | Batch 050/938 | Cost: 0.0154\n",
|
|
"Epoch: 010/010 | Batch 100/938 | Cost: 0.0135\n",
|
|
"Epoch: 010/010 | Batch 150/938 | Cost: 0.0122\n",
|
|
"Epoch: 010/010 | Batch 200/938 | Cost: 0.0354\n",
|
|
"Epoch: 010/010 | Batch 250/938 | Cost: 0.0217\n",
|
|
"Epoch: 010/010 | Batch 300/938 | Cost: 0.0102\n",
|
|
"Epoch: 010/010 | Batch 350/938 | Cost: 0.0189\n",
|
|
"Epoch: 010/010 | Batch 400/938 | Cost: 0.0015\n",
|
|
"Epoch: 010/010 | Batch 450/938 | Cost: 0.0101\n",
|
|
"Epoch: 010/010 | Batch 500/938 | Cost: 0.0950\n",
|
|
"Epoch: 010/010 | Batch 550/938 | Cost: 0.0130\n",
|
|
"Epoch: 010/010 | Batch 600/938 | Cost: 0.0328\n",
|
|
"Epoch: 010/010 | Batch 650/938 | Cost: 0.0506\n",
|
|
"Epoch: 010/010 | Batch 700/938 | Cost: 0.0182\n",
|
|
"Epoch: 010/010 | Batch 750/938 | Cost: 0.0091\n",
|
|
"Epoch: 010/010 | Batch 800/938 | Cost: 0.0075\n",
|
|
"Epoch: 010/010 | Batch 850/938 | Cost: 0.0092\n",
|
|
"Epoch: 010/010 | Batch 900/938 | Cost: 0.0413\n",
|
|
"Epoch: 010/010 training accuracy: 99.00%\n",
|
|
"Time elapsed: 2.23 min\n",
|
|
"Total Training Time: 2.23 min\n"
|
|
]
|
|
}
|
|
],
|
|
"source": [
|
|
"def compute_accuracy(net, data_loader):\n",
|
|
" net.eval()\n",
|
|
" correct_pred, num_examples = 0, 0\n",
|
|
" with torch.no_grad():\n",
|
|
" for features, targets in data_loader:\n",
|
|
" features = features.view(-1, 28*28).to(device)\n",
|
|
" targets = targets.to(device)\n",
|
|
" logits, probas = net(features)\n",
|
|
" _, predicted_labels = torch.max(probas, 1)\n",
|
|
" num_examples += targets.size(0)\n",
|
|
" correct_pred += (predicted_labels == targets).sum()\n",
|
|
" return correct_pred.float()/num_examples * 100\n",
|
|
" \n",
|
|
"\n",
|
|
"start_time = time.time()\n",
|
|
"for epoch in range(num_epochs):\n",
|
|
" model.train()\n",
|
|
" for batch_idx, (features, targets) in enumerate(train_loader):\n",
|
|
" \n",
|
|
" features = features.view(-1, 28*28).to(device)\n",
|
|
" targets = targets.to(device)\n",
|
|
" \n",
|
|
" ### FORWARD AND BACK PROP\n",
|
|
" logits, probas = model(features)\n",
|
|
" cost = F.cross_entropy(logits, targets)\n",
|
|
" optimizer.zero_grad()\n",
|
|
" \n",
|
|
" cost.backward()\n",
|
|
" \n",
|
|
" ### UPDATE MODEL PARAMETERS\n",
|
|
" optimizer.step()\n",
|
|
" \n",
|
|
" ### LOGGING\n",
|
|
" if not batch_idx % 50:\n",
|
|
" print ('Epoch: %03d/%03d | Batch %03d/%03d | Cost: %.4f' \n",
|
|
" %(epoch+1, num_epochs, batch_idx, \n",
|
|
" len(train_loader), cost))\n",
|
|
"\n",
|
|
" with torch.set_grad_enabled(False):\n",
|
|
" print('Epoch: %03d/%03d training accuracy: %.2f%%' % (\n",
|
|
" epoch+1, num_epochs, \n",
|
|
" compute_accuracy(model, train_loader)))\n",
|
|
" \n",
|
|
" print('Time elapsed: %.2f min' % ((time.time() - start_time)/60))\n",
|
|
" \n",
|
|
"print('Total Training Time: %.2f min' % ((time.time() - start_time)/60))"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": 6,
|
|
"metadata": {},
|
|
"outputs": [
|
|
{
|
|
"name": "stdout",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"Test accuracy: 97.54%\n"
|
|
]
|
|
}
|
|
],
|
|
"source": [
|
|
"print('Test accuracy: %.2f%%' % (compute_accuracy(model, test_loader)))"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": 7,
|
|
"metadata": {},
|
|
"outputs": [
|
|
{
|
|
"name": "stdout",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"numpy 1.15.4\n",
|
|
"torch 1.0.0\n",
|
|
"\n"
|
|
]
|
|
}
|
|
],
|
|
"source": [
|
|
"%watermark -iv"
|
|
]
|
|
}
|
|
],
|
|
"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": "3.7.1"
|
|
},
|
|
"toc": {
|
|
"nav_menu": {},
|
|
"number_sections": true,
|
|
"sideBar": true,
|
|
"skip_h1_title": false,
|
|
"title_cell": "Table of Contents",
|
|
"title_sidebar": "Contents",
|
|
"toc_cell": false,
|
|
"toc_position": {},
|
|
"toc_section_display": true,
|
|
"toc_window_display": false
|
|
}
|
|
},
|
|
"nbformat": 4,
|
|
"nbformat_minor": 2
|
|
}
|