Thunder Compute logo

Forward Pass

Computing the output of a neural network from input to prediction

A forward pass is the process of feeding input data through a neural network layer by layer to produce an output (prediction). Each layer applies its weights and activation functions in sequence.

Example

import torch
import torch.nn as nn

model = nn.Sequential(
    nn.Linear(784, 256),
    nn.ReLU(),
    nn.Linear(256, 10),
)
model = model.cuda()

x = torch.randn(32, 784, device="cuda")  # batch of 32
output = model(x)                          # forward pass → shape [32, 10]

Key Points

  • Computes predictions from input data
  • Each layer's intermediate outputs are called activations
  • Activations are stored in memory during training (needed for backpropagation)

See Also