38 KiB
38 KiB
import torch
x = [[1,2],[3,4],[5,6],[7,8]]
y = [[3],[7],[11],[15]]
X = torch.tensor(x).float()
Y = torch.tensor(y).float()
device = 'cuda' if torch.cuda.is_available() else 'cpu'
X = X.to(device)
Y = Y.to(device)
import torch.nn as nn
class MyNeuralNet(nn.Module):
def __init__(self):
super().__init__()
self.input_to_hidden_layer = nn.Linear(2,8)
self.hidden_layer_activation = nn.ReLU()
self.hidden_to_output_layer = nn.Linear(8,1)
def forward(self, x):
x = self.input_to_hidden_layer(x)
x = self.hidden_layer_activation(x)
x = self.hidden_to_output_layer(x)
return x
torch.random.manual_seed(10)
mynet = MyNeuralNet().to(device)
loss_func = nn.MSELoss()
_Y = mynet(X)
loss_value = loss_func(_Y,Y)
print(loss_value)
tensor(102.1545, grad_fn=<MseLossBackward>)
from torch.optim import SGD
opt = SGD(mynet.parameters(), lr = 0.001)
loss_history = []
for _ in range(50):
opt.zero_grad()
loss_value = loss_func(mynet(X),Y)
loss_value.backward()
opt.step()
loss_history.append(loss_value)
import matplotlib.pyplot as plt
%matplotlib inline
plt.plot(loss_history)
plt.title('Loss variation over increasing epochs')
plt.xlabel('epochs')
plt.ylabel('loss value')
Text(0, 0.5, 'loss value')
1. Fetching intermediate values by directly calling the intermediate layer
mynet.input_to_hidden_layer(X)
tensor([[-4.0139e-01, 7.2155e-03, -4.9131e-01, 1.4615e+00, -3.8093e-01, -7.1646e-01, 4.6765e-01, 2.0814e+00], [-5.6844e-01, -2.2575e-01, -1.5498e+00, 3.1695e+00, -5.2755e-01, -7.3935e-01, 1.9716e+00, 5.3073e+00], [-7.3548e-01, -4.5871e-01, -2.6083e+00, 4.8776e+00, -6.7418e-01, -7.6225e-01, 3.4756e+00, 8.5332e+00], [-9.0252e-01, -6.9167e-01, -3.6667e+00, 6.5856e+00, -8.2080e-01, -7.8514e-01, 4.9795e+00, 1.1759e+01]], grad_fn=<AddmmBackward>)
2. Fetching intermediate values by returning them in nn.Module
definition
torch.random.manual_seed(10)
class MyNeuralNet(nn.Module):
def __init__(self):
super().__init__()
self.input_to_hidden_layer = nn.Linear(2,8)
self.hidden_layer_activation = nn.ReLU()
self.hidden_to_output_layer = nn.Linear(8,1)
def forward(self, x):
hidden1 = self.input_to_hidden_layer(x)
hidden2 = self.hidden_layer_activation(hidden1)
x = self.hidden_to_output_layer(hidden2)
return x, hidden1
mynet = MyNeuralNet().to(device)
loss_func = nn.MSELoss()
_Y, _Y_hidden = mynet(X)
loss_value = loss_func(_Y,Y)
opt = SGD(mynet.parameters(), lr = 0.001)
loss_history = []
for _ in range(50):
opt.zero_grad()
loss_value = loss_func(mynet(X)[0],Y)
loss_value.backward()
opt.step()
loss_history.append(loss_value)
import matplotlib.pyplot as plt
%matplotlib inline
plt.plot(loss_history)
plt.title('Loss variation over increasing epochs')
plt.xlabel('epochs')
plt.ylabel('loss value')
Text(0, 0.5, 'loss value')
mynet(X)[1]
tensor([[-4.0139e-01, 7.2155e-03, -4.9131e-01, 1.4615e+00, -3.8093e-01, -7.1646e-01, 4.6765e-01, 2.0814e+00], [-5.6844e-01, -2.2575e-01, -1.5498e+00, 3.1695e+00, -5.2755e-01, -7.3935e-01, 1.9716e+00, 5.3073e+00], [-7.3548e-01, -4.5871e-01, -2.6083e+00, 4.8776e+00, -6.7418e-01, -7.6225e-01, 3.4756e+00, 8.5332e+00], [-9.0252e-01, -6.9167e-01, -3.6667e+00, 6.5856e+00, -8.2080e-01, -7.8514e-01, 4.9795e+00, 1.1759e+01]], grad_fn=<AddmmBackward>)