rml_op_pool_2d_params

Parameters for the RML_OP_POOL_2D_AVG and RML_OP_POOL_2D_MAX operations.

RML_OP_POOL_2D* operations apply pooling across the tensor according to kernel size, strides and dilations.

API

Graph Manipulation

Items

Name	Type	Description
input	rml_op	Operation that outputs tensor with input data.
padding_type	rml_padding_type	Padding calculation algorithm.
kernel_size	rml_size_2d	Spatial sizes of pooling kernel.
strides	rml_size_2d	Stride along each spatial axis. If it holds zeros, the stride is set to 1 along each spatial axis.
dilations	rml_size_2d	Dilation value along each spatial axis of the kernel. If it holds zeros, the dilation is set to 1 along each spatial axis.
start_paddings	rml_size_2d	Number of pixels added to the beginning of the each spatial axis. Only make sense if padding type is RML_PADDING_EXPLICIT.
end_paddings	rml_size_2d	Number of pixels added to the end of the each spatial axis. Only make sense if padding type is RML_PADDING_EXPLICIT.
ceil_mode	rml_bool	Whether to use ceil (RML_TRUE) or floor (RML_FALSE) to compute the output shape.