PyTorch torch.nn.Dropout2d Function

PyTorch torch.nn 参考手册PyTorch torch.nn Reference Manual

torch.nn.Dropout2dis a two-dimensional Dropout module in PyTorch.

It randomly zeros entire channels, suitable for feature maps of convolutional layers.

Function Definition

torch.nn.Dropout2d(p=0.5, inplace=False)

Features

  • Randomly zero out along the channel dimension
  • All channels of each sample maintain a consistent dropout pattern

Usage Examples

Example 1: Basic Usage

Example

import torch
import torch.nn as nn

dropout2d = nn.Dropout2d(p=0.5)
dropout2d.train()

# Input: batch=4, channels=8, height=16, width=16
x = torch.ones(4, 8, 16, 16)

output = dropout2d(x)

# Count the number of dropped channels
non_zero_channels = (output.sum(dim=(2, 3)) != 0).float()
print("Non-zero channel ratio:", non_zero_channels.mean().item())
print("Expected about 0.5 of channels to be retained")

Example 2: Comparing Dropout

Example

import torch
import torch.nn as nn

dropout1d = nn.Dropout(0.5)
dropout2d = nn.Dropout2d(0.5)

# Input
x = torch.randn(2, 4, 8, 8)

# Dropout: randomly zeros individual elements
out1 = dropout1d(x)
# Dropout2d: zeros by channel
out2 = dropout2d(x)

print("Dropout shape:", out1.shape)
print("Dropout2d shape:", out2.shape)

Example 3: Using in CNN

Example

import torch
import torch.nn as nn

# CNN with Dropout2d
model = nn.Sequential(
    nn.Conv2d(3, 64, 3, padding=1),
    nn.BatchNorm2d(64),
    nn.ReLU(),
    nn.Dropout2d(0.3),  # Dropout at feature map level
    nn.Conv2d(64, 128, 3, padding=1),
    nn.BatchNorm2d(128),
    nn.ReLU(),
    nn.AdaptiveAvgPool2d(1),
    nn.Flatten(),
    nn.Linear(128, 10)
)

x = torch.randn(4, 3, 32, 32)
output = model(x)

print("Input:", x.shape, "-> Output:", output.shape)

Use Cases

  • Convolutional networks: Feature map level regularization
  • Reduce channel dependency

Note: Dropout2d drops channels during training and has no effect during evaluation.

PyTorch torch.nn 参考手册PyTorch torch.nn Reference Manual

Other Extensions