Day #226 - Pytorch Sessions (17-19)

Key Lessons
Class Basic Example

	#Class name
	class Lizard:
	#Constructor
	#Called for New Instance
	def __init__(self,name):
	self.name = name
	#custom method
	def set_name(self,name):
	self.name = name
	lizard = Lizard('deep')
	print(lizard.name)
	#object.function
	lizard.set_name('lizard')
	#Object.attribute name
	print(lizard.name)

view raw pytorch_class.py hosted with ❤ by GitHub

NN code

	#parameters - Inside Funtion Definition
	#Arguments - Acutal Value passed
	#Build CNN
	#Kernel Channel - Set Size of filter, Convolutional Kernel/Filter
	#Out Channel - Set the number of filters / Feature Maps
	#Out Features - Last Linear Layer
	class Network:
	def __init__(self):
	self.layer = None
	def forward(self,t):
	t = self.layer(t)
	return t
	import torch.nn as nn
	class Network(nn.Module):
	def __init__(self):
	super(Network,self).__init__()
	#Convolutional Layers
	#in_channels Number of Color Channels
	self.conv1 = nn.Conv2d(in_channels=1,out_channels=6,kernel_size=5)
	self.conv2 = nn.Conv2d(in_channels=6,out_channels=12,kernel_size=5)
	#fully connected
	#Linear Layers
	#12 Output Channels
	#4 * 4 -
	self.fc1 = nn.Linear(in_features=12*4*4,out_features=120)
	#Expand features
	self.fc2 = nn.Linear(in_features=120,out_features=60)
	#Output Layer
	#Shrink Layers
	#out_features Number of Classes in Training Set
	self.out = nn.Linear(in_features=60,out_features=10)
	def forward(self,t):
	t = self.layer(t)
	return t
	#Override
	def __repr__(self):
	return "lizardnet"
	network = Network()
	#Check weights
	#Values are Learnable Parameters
	#Trained to minimise loss function
	print(network.conv1)
	print(network.conv1.weight)
	print(network.conv1.weight.shape)
	print(network.conv2)
	print(network.conv2.weight)
	print(network.conv2.weight.shape)
	print(network.fc1)
	print(network.fc1.weight)
	print(network.fc1.weight.shape)
	#Height - Output Feature
	#Width - Input Feature
	print(network)
	for param in network.parameters():
	print(param.shape)
	for name,param in network.named_parameters():
	print(name,'\t\t',param.shape)

view raw PytorchNNExample.py hosted with ❤ by GitHub

Multiplication

	import torch
	import torchvision
	import torchvision.transforms as transforms
	in_features = torch.tensor([1,2,3,4],dtype=torch.float32)
	weight_matrix = torch.tensor([[1,2,3,4],[2,3,4,5],[3,4,5,6]],dtype=torch.float32)
	#matrix multiplication
	print(weight_matrix.matmul(in_features))

view raw Pytorch_matmul.py hosted with ❤ by GitHub

Happy Mastering DL!!!!