nn.AdaptiveMaxPool1d¶

class lucid.nn.AdaptiveMaxPool1d(output_size: int)¶

The AdaptiveMaxPool1d performs adaptive max pooling on a 1D input tensor, automatically computing kernel size, stride, and padding to produce a specified output size.

Class Signature¶

class AdaptiveMaxPool1d(nn.Module):
    def __init__(self, output_size: int)

Parameters¶

output_size (int): The desired output length \(L_{out}\).

Attributes¶

output_size (int): Stores the target output length for the pooling operation.

Forward Calculation¶

Applies adaptive max pooling on input of shape \((N, C, L)\), producing an output of shape \((N, C, L_{out})\). Pooling parameters are dynamically calculated based on input length.

Behavior¶

This module wraps the functional adaptive_max_pool1d API to provide a module interface, automatically adjusting pooling parameters to meet the target output size.

Examples¶

import lucid.nn as nn

input_ = Tensor([[[1, 4, 2, 3, 6, 5, 7, 0, 8, 1]]])
pool = nn.AdaptiveMaxPool1d(output_size=5)
output = pool(input_)
print(output)  # Shape: (1, 1, 5)