Skip to content

Commit 67639a0

Browse files
tinhb92tinhb92
committed
Created using Colaboratory
1 parent 67b3215 commit 67639a0

File tree

1 file changed

+244
-0
lines changed

1 file changed

+244
-0
lines changed

‎deep_learning/lstm.ipynb

+244
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,244 @@
1+
{
2+
"nbformat": 4,
3+
"nbformat_minor": 0,
4+
"metadata": {
5+
"colab": {
6+
"name": "lstm",
7+
"provenance": [],
8+
"include_colab_link": true
9+
},
10+
"kernelspec": {
11+
"name": "python3",
12+
"display_name": "Python 3"
13+
}
14+
},
15+
"cells": [
16+
{
17+
"cell_type": "markdown",
18+
"metadata": {
19+
"id": "view-in-github",
20+
"colab_type": "text"
21+
},
22+
"source": [
23+
"<a href=\"https://colab.research.google.com/github/tinhb92/relax_ml/blob/master/deep_learning/lstm.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
24+
]
25+
},
26+
{
27+
"cell_type": "code",
28+
"metadata": {
29+
"id": "414DjyHwHBH7",
30+
"colab_type": "code",
31+
"colab": {}
32+
},
33+
"source": [
34+
"import torch \n",
35+
"import torch.nn as nn\n",
36+
"import torchvision\n",
37+
"import torchvision.transforms as transforms\n",
38+
"\n",
39+
"# Device configuration\n",
40+
"device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')\n",
41+
"\n",
42+
"# Hyper-parameters\n",
43+
"sequence_length = 28\n",
44+
"input_size = 28\n",
45+
"hidden_size = 128\n",
46+
"num_layers = 2\n",
47+
"num_classes = 10\n",
48+
"batch_size = 100\n",
49+
"num_epochs = 2\n",
50+
"learning_rate = 0.01"
51+
],
52+
"execution_count": 0,
53+
"outputs": []
54+
},
55+
{
56+
"cell_type": "code",
57+
"metadata": {
58+
"id": "e2NCj8qbGI2r",
59+
"colab_type": "code",
60+
"colab": {}
61+
},
62+
"source": [
63+
"# MNIST dataset\n",
64+
"train_dataset = torchvision.datasets.MNIST(root='../../data/',\n",
65+
" train=True, \n",
66+
" transform=transforms.ToTensor(),\n",
67+
" download=True)\n",
68+
"\n",
69+
"test_dataset = torchvision.datasets.MNIST(root='../../data/',\n",
70+
" train=False, \n",
71+
" transform=transforms.ToTensor())\n",
72+
"\n",
73+
"# Data loader\n",
74+
"train_loader = torch.utils.data.DataLoader(dataset=train_dataset,\n",
75+
" batch_size=batch_size, \n",
76+
" shuffle=True)\n",
77+
"\n",
78+
"test_loader = torch.utils.data.DataLoader(dataset=test_dataset,\n",
79+
" batch_size=batch_size, \n",
80+
" shuffle=False)"
81+
],
82+
"execution_count": 0,
83+
"outputs": []
84+
},
85+
{
86+
"cell_type": "code",
87+
"metadata": {
88+
"id": "o55sVKwpGMLx",
89+
"colab_type": "code",
90+
"colab": {}
91+
},
92+
"source": [
93+
"# Recurrent neural network (many-to-one)\n",
94+
"class RNN(nn.Module):\n",
95+
" def __init__(self, input_size, hidden_size, num_layers, num_classes):\n",
96+
" super(RNN, self).__init__()\n",
97+
" self.hidden_size = hidden_size\n",
98+
" self.num_layers = num_layers\n",
99+
" self.lstm = nn.LSTM(input_size, hidden_size, num_layers, batch_first=True)\n",
100+
" self.fc = nn.Linear(hidden_size, num_classes)\n",
101+
" \n",
102+
" def forward(self, x):\n",
103+
" # Set initial hidden and cell states \n",
104+
" h0 = torch.zeros(self.num_layers, x.size(0), self.hidden_size).to(device) \n",
105+
" c0 = torch.zeros(self.num_layers, x.size(0), self.hidden_size).to(device)\n",
106+
" \n",
107+
" # Forward propagate LSTM\n",
108+
" out, _ = self.lstm(x, (h0, c0)) # out: tensor of shape (batch_size, seq_length, hidden_size)\n",
109+
" \n",
110+
" # Decode the hidden state of the last time step\n",
111+
" out = self.fc(out[:, -1, :])\n",
112+
" return out\n",
113+
"\n",
114+
"model = RNN(input_size, hidden_size, num_layers, num_classes).to(device)"
115+
],
116+
"execution_count": 0,
117+
"outputs": []
118+
},
119+
{
120+
"cell_type": "code",
121+
"metadata": {
122+
"id": "EeQKqwxAGOod",
123+
"colab_type": "code",
124+
"colab": {
125+
"base_uri": "https://localhost:8080/",
126+
"height": 218
127+
},
128+
"outputId": "5e6bf8ff-34ff-4d39-8e93-c2741f1cf6f8"
129+
},
130+
"source": [
131+
"# Loss and optimizer\n",
132+
"criterion = nn.CrossEntropyLoss()\n",
133+
"optimizer = torch.optim.Adam(model.parameters(), lr=learning_rate)\n",
134+
"\n",
135+
"# Train the model\n",
136+
"total_step = len(train_loader)\n",
137+
"for epoch in range(num_epochs):\n",
138+
" for i, (images, labels) in enumerate(train_loader):\n",
139+
" images = images.reshape(-1, sequence_length, input_size).to(device)\n",
140+
" labels = labels.to(device)\n",
141+
" \n",
142+
" # Forward pass\n",
143+
" outputs = model(images)\n",
144+
" loss = criterion(outputs, labels)\n",
145+
" \n",
146+
" # Backward and optimize\n",
147+
" optimizer.zero_grad()\n",
148+
" loss.backward()\n",
149+
" optimizer.step()\n",
150+
" \n",
151+
" if (i+1) % 100 == 0:\n",
152+
" print ('Epoch [{}/{}], Step [{}/{}], Loss: {:.4f}' \n",
153+
" .format(epoch+1, num_epochs, i+1, total_step, loss.item()))"
154+
],
155+
"execution_count": 22,
156+
"outputs": [
157+
{
158+
"output_type": "stream",
159+
"text": [
160+
"Epoch [1/2], Step [100/600], Loss: 0.4692\n",
161+
"Epoch [1/2], Step [200/600], Loss: 0.2797\n",
162+
"Epoch [1/2], Step [300/600], Loss: 0.1271\n",
163+
"Epoch [1/2], Step [400/600], Loss: 0.2750\n",
164+
"Epoch [1/2], Step [500/600], Loss: 0.1792\n",
165+
"Epoch [1/2], Step [600/600], Loss: 0.0991\n",
166+
"Epoch [2/2], Step [100/600], Loss: 0.0826\n",
167+
"Epoch [2/2], Step [200/600], Loss: 0.1674\n",
168+
"Epoch [2/2], Step [300/600], Loss: 0.1562\n",
169+
"Epoch [2/2], Step [400/600], Loss: 0.1447\n",
170+
"Epoch [2/2], Step [500/600], Loss: 0.0842\n",
171+
"Epoch [2/2], Step [600/600], Loss: 0.0283\n"
172+
],
173+
"name": "stdout"
174+
}
175+
]
176+
},
177+
{
178+
"cell_type": "code",
179+
"metadata": {
180+
"id": "j-5yxyHfEa7z",
181+
"colab_type": "code",
182+
"colab": {
183+
"base_uri": "https://localhost:8080/",
184+
"height": 34
185+
},
186+
"outputId": "84f5134c-0b9e-42e3-fcaf-42f528a3f363"
187+
},
188+
"source": [
189+
"# Test the model\n",
190+
"with torch.no_grad():\n",
191+
" correct = 0\n",
192+
" total = 0\n",
193+
" for images, labels in test_loader:\n",
194+
" images = images.reshape(-1, sequence_length, input_size).to(device)\n",
195+
" labels = labels.to(device)\n",
196+
" outputs = model(images)\n",
197+
" _, predicted = torch.max(outputs.data, 1)\n",
198+
" total += labels.size(0)\n",
199+
" correct += (predicted == labels).sum().item()\n",
200+
"\n",
201+
" print('Test Accuracy of the model on the 10000 test images: {} %'.format(100 * correct / total)) \n",
202+
"\n",
203+
"# Save the model checkpoint\n",
204+
"# torch.save(model.state_dict(), 'model.ckpt')"
205+
],
206+
"execution_count": 23,
207+
"outputs": [
208+
{
209+
"output_type": "stream",
210+
"text": [
211+
"Test Accuracy of the model on the 10000 test images: 97.55 %\n"
212+
],
213+
"name": "stdout"
214+
}
215+
]
216+
},
217+
{
218+
"cell_type": "code",
219+
"metadata": {
220+
"id": "cs2NL18qEaz2",
221+
"colab_type": "code",
222+
"colab": {}
223+
},
224+
"source": [
225+
""
226+
],
227+
"execution_count": 0,
228+
"outputs": []
229+
},
230+
{
231+
"cell_type": "code",
232+
"metadata": {
233+
"id": "1BY0qUqTCrMq",
234+
"colab_type": "code",
235+
"colab": {}
236+
},
237+
"source": [
238+
""
239+
],
240+
"execution_count": 0,
241+
"outputs": []
242+
}
243+
]
244+
}

0 commit comments

Comments
 (0)