PyTorch torch.quantized_max_pool2d Function
Pytorch torch Reference Manual
torch.quantized_max_pool2dIs a function in PyTorch used to perform two-dimensional max pooling operations on quantized tensors. This function is commonly used for spatial downsampling in quantized convolutional neural networks.
Function Definition
torch.quantized_max_pool2d(input, kernel_size, stride, padding, dilation)
Parameter Description
input: Input quantized tensor (4D: batch x channel x height x width)kernel_size: Pooling window sizestride: Stride (optional)padding: Padding (optional)dilation: Dilation (optional)
Usage Example
Example
import torch
# Create quantized input tensor (batch=1, channel=1, height=4, width=4)
input = torch.quantize_per_tensor(torch.randn(1, 1, 4, 4), scale=0.1, zero_point=0, dtype=torch.quint8)
# Perform quantized max pooling (2x2 pooling window)
output = torch.quantized_max_pool2d(input, kernel_size=2, stride=2)
print("Input shape:", input.shape)
print("Output shape:", output.shape)
# Create quantized input tensor (batch=1, channel=1, height=4, width=4)
input = torch.quantize_per_tensor(torch.randn(1, 1, 4, 4), scale=0.1, zero_point=0, dtype=torch.quint8)
# Perform quantized max pooling (2x2 pooling window)
output = torch.quantized_max_pool2d(input, kernel_size=2, stride=2)
print("Input shape:", input.shape)
print("Output shape:", output.shape)
The output result is:
输入形状: torch.Size([1, 1, 4, 4]) 输出形状: torch.Size([1, 1, 2, 2])
Other Extensions