Multi-Head Attention module Modified from github.com/jadore801120/attention-is-all-you-need-pytorch

Scaled Dot-Product Attention Modified from github.com/jadore801120/attention-is-all-you-need-pytorch

The Losses meta-object that can take a mix of losses.

Computes the mean intersection-over-union (miou), the binary change score (bc), the semantic change score (sc) and the semantic change segmen

Helper module for convolutional encoding blocks that are shared across a sequence. This module adds the self.smart_forward() method the the b

Linear Warmup learning rate scheduler. After warmup, learning rate is constant. After warmup, learning rate follows a cosine decay.

Linear Warmup learning rate scheduler. After warmup, learning rate is constant. Args: optimizer (torch.optim.Optimizer): optimiz