For the complete documentation index, see llms.txt. Markdown versions of all pages are available by appending .md to any URL (e.g. /max/get-started.md).

Mojo function

pool_shape

def pool_shape[input_type: DType, filter_type: DType, strides_type: DType, dilations_type: DType, paddings_type: DType](input_buf: TileTensor[input_type, Storage=input_buf.Storage, address_space=input_buf.address_space, linear_idx_type=input_buf.linear_idx_type], filter_buf: TileTensor[filter_type, Storage=filter_buf.Storage, address_space=filter_buf.address_space, linear_idx_type=filter_buf.linear_idx_type], strides_buf: TileTensor[strides_type, Storage=strides_buf.Storage, address_space=strides_buf.address_space, linear_idx_type=strides_buf.linear_idx_type], dilations_buf: TileTensor[dilations_type, Storage=dilations_buf.Storage, address_space=dilations_buf.address_space, linear_idx_type=dilations_buf.linear_idx_type], paddings_buf: TileTensor[paddings_type, Storage=paddings_buf.Storage, address_space=paddings_buf.address_space, linear_idx_type=paddings_buf.linear_idx_type]) -> IndexList[input_buf.LayoutType.rank]

Computes the output shape of a pooling operation using floor rounding for the spatial dimensions.

Parameters:

input_type (DType): Data type of the input tensor.
filter_type (DType): Data type of the filter tensor.
strides_type (DType): Data type of the strides tensor.
dilations_type (DType): Data type of the dilations tensor.
paddings_type (DType): Data type of the paddings tensor.

Args:

input_buf (TileTensor[input_type, Storage=input_buf.Storage, address_space=input_buf.address_space, linear_idx_type=input_buf.linear_idx_type]): The input tensor.
filter_buf (TileTensor[filter_type, Storage=filter_buf.Storage, address_space=filter_buf.address_space, linear_idx_type=filter_buf.linear_idx_type]): The filter size buffer.
strides_buf (TileTensor[strides_type, Storage=strides_buf.Storage, address_space=strides_buf.address_space, linear_idx_type=strides_buf.linear_idx_type]): The strides size buffer.
dilations_buf (TileTensor[dilations_type, Storage=dilations_buf.Storage, address_space=dilations_buf.address_space, linear_idx_type=dilations_buf.linear_idx_type]): The dilations size buffer.
paddings_buf (TileTensor[paddings_type, Storage=paddings_buf.Storage, address_space=paddings_buf.address_space, linear_idx_type=paddings_buf.linear_idx_type]): The paddings size buffer.

Returns:

IndexList[input_buf.LayoutType.rank]: The output shape with floor-mode rounding applied.