Merge pull request #8 from jeffin07/alexnet

Alexnet example

Author: imdeepmind

alexnet/model.py

@@ -0,0 +1,66 @@
+from neuralpy.models import Sequential
+from neuralpy.layers.linear import Dense
+from neuralpy.layers.convolutional import Conv2D
+from neuralpy.layers.activation_functions import ReLU,Softmax
+from neuralpy.layers.pooling import MaxPool2D
+from neuralpy.layers.other import Flatten
+from neuralpy.loss_functions import CrossEntropyLoss
+from neuralpy.optimizer import SGD,Adam
+
+import torch
+import torchvision.datasets as datasets
+import torchvision.transforms as transforms
+
+# Create a Sequential model Instance
+model = Sequential()
+
+model.add(Conv2D(input_shape=(1,224,224), filters=96, kernel_size=11, stride=4))
+model.add(ReLU())
+model.add(MaxPool2D(kernel_size=3, stride=2))
+model.add(ReLU())
+model.add(Conv2D(filters=256, kernel_size=5, stride=1, padding=2))
+model.add(ReLU())
+model.add(MaxPool2D(kernel_size=3, stride=2))
+model.add(ReLU())
+model.add(Conv2D(filters=384, kernel_size=3, stride=1, padding=1))
+model.add(ReLU())
+model.add(Conv2D(filters=384, kernel_size=3, stride=1, padding=1))
+model.add(ReLU())
+model.add(Conv2D(filters=256, kernel_size=3, stride=1, padding=1))
+model.add(ReLU())
+model.add(MaxPool2D(kernel_size=3, stride=2))
+model.add(ReLU())
+model.add(Flatten())
+model.add(Dense(n_nodes=4096))
+model.add(ReLU())
+model.add(Dense(n_nodes=4096))
+model.add(ReLU())
+model.add(Dense(n_nodes=10))
+model.add(Softmax())
+
+model.build()
+model.compile(optimizer=Adam(), loss_function=CrossEntropyLoss(), metrics=["accuracy"])
+print(model.summary())
+
+# Get the training Data
+train_set = datasets.MNIST(
+    root='./data'
+    ,train=True
+    ,download=True
+    ,transform=transforms.Compose([
+        transforms.CenterCrop(224),
+        transforms.ToTensor()
+    ])
+)
+
+# Load the dataset from pytorch's Dataloader function
+train_loader  = torch.utils.data.DataLoader(train_set, batch_size=1000)
+#Get the data
+mnist_data = next(iter(train_loader))
+#Split into train and test set
+train_imgs = mnist_data[0][:500]
+train_labels = mnist_data[1][:500]
+test_imgs = mnist_data[0][500:]
+test_labels = mnist_data[1][500:]
+# Train Model
+model.fit(train_data=(train_imgs,train_labels), epochs=5, validation_data=(test_imgs,test_labels),batch_size=1)

alexnet/requirements.txt

@@ -0,0 +1,3 @@
+torch==1.7.1+cpu
+numpy==1.17.4
+neuralpy-torch==0.2.1