Layer Node

Conv2dNode

layer_node.Conv2dNode(in_channels, out_channels, kernel_size, 
                    stride, padding, paddingmode, dilation, bias, 
                    groups, task_list)

Conv2d embedded with task.

argument

• arguments except task_list are the same as Conv2d
• task_list should be a list of tasks that you are interested in trainning

BN2dNode

layer_node.BN2dNode(num_features: int,
                    eps: float, 
                    momentum: float, 
                    affine: bool, 
                    track_running_stats: bool, 
                    task_list: list)

BN2d embedded with task

argument

• arguments except task_list are the same as BatchNorm2d
• task_list should be a list of tasks that you are interested in trainning

Sequential

layer_node.Sequential(seq)

A useful wrapper class for applying special forwarding in nn.Sequential

Argument

• seq: should be nn.Sequential type of module