AvgPool2d
AdaptiveAvgPool2d
MaxPool2d
LastLevelMaxPool
This model does not include MaxPool2d
calculating the average for each patch of the feature map
calculating the correct kernel size to use to average each patch of the feature map
This model does not include LastLevelMaxPool
BatchNorm2d
LayerNorm
FrozenBatchNorm2d
LayerNorm2d
applying a 2D Batch Normalization layer over a 4D input
This model does not include LayerNorm
This model does not include FrozenBatchNorm2d
This model does not include LayerNorm2d
Conv2d
Sep_conv2d
Dil_conv2d
creating a convolution kernel and convolve it with the layer input
This model does not include Sep_conv2d
This model does not include Dil_conv2d
ReLU
GELU
Hardswish
ReLU6
SiLU
Sigmoid
Hardsigmoid
This model does not include GELU
calculating Rectified Linear Unit (relu) which is a piecewise function
This model does not include Hardswish
This model does not include ReLU6
This model does not include SiLU
This model does not include Sigmoid
This model does not include Hardsigmoid
3*3
7*7
1*1
Identity
Linear
Dropout
calculating the maximum value for each patch of the feature map
dividing single convolutions into two or more and reducing number of parameters while producing the same output
5*5
Zero
creating a wider kernel by inserting spaces between the kernel elements
