What is nn.Module?
- Machine Learning course
- Advanced
- Lesson 29/78
- 5 min
nn.Module is the base class for all neural network models.
What is nn.Module?
nn.Module is the base class every model and layer in PyTorch derives from. A module keeps its parameters, its submodules and the forward computation together:
class Model(nn.Module):
def __init__(self):
super().__init__()
self.katman = nn.Linear(4, 3)
def forward(self, x):
return self.katman(x)
m = Model()
print(m(torch.zeros(2, 4)).shape) # torch.Size([2, 3])A module is called as m(x); behind the scenes this runs forward.
Ready-made modules
Inside torch.nn, layers (Linear, Conv2d, LSTM), activations (ReLU, Sigmoid), losses (MSELoss, CrossEntropyLoss) and regularisers (Dropout, BatchNorm1d) are all modules. For simple chains, nn.Sequential builds a model without writing a class:
model = nn.Sequential(
nn.Linear(4, 16),
nn.ReLU(),
nn.Linear(16, 3),
)
print(model(torch.zeros(5, 4)).shape) # torch.Size([5, 3])Practice First module
Build the model nn.Sequential(nn.Linear(4, 16), nn.ReLU(), nn.Linear(16, 3)). Print the output shape for the input torch.zeros(5, 4) and the model's length (len(model)) on one line.
Starter code
import torch
from torch import nn
# buraya yazSample solution
import torch
from torch import nn
model = nn.Sequential(nn.Linear(4, 16), nn.ReLU(), nn.Linear(16, 3))
print(model(torch.zeros(5, 4)).shape, len(model))Expected output
torch.Size([5, 3]) 3
Questions and answers
Review the lesson questions together with their answers. To test yourself, solve the lesson interactively.
1 Why is <code>m(x)</code> written instead of <code>m.forward(x)</code> to run the model?
- m(x) also runs hooks and internal machinery Correct answer
- Because it is shorter
- There is no forward
- There is no difference
Explanation: __call__ wraps forward; calling forward directly skips these steps.
2 Chain the layers in order.
model = nn.____(nn.Linear(4, 8), nn.ReLU(), nn.Linear(8, 2))model = nn.Sequential(nn.Linear(4, 8), nn.ReLU(), nn.Linear(8, 2))Explanation: nn.Sequential passes the input through the modules in order.
3 Match the modules.
| Term | Meaning |
|---|---|
nn.Linear |
Linear layer |
nn.ReLU |
Activation |
nn.CrossEntropyLoss |
Loss |
nn.Sequential |
Sequential container |