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

get_padding_output_shape

def get_padding_output_shape[rank: Int](input_shape: IndexList[rank], paddings: TileTensor[DType.int, Storage=paddings.Storage, address_space=paddings.address_space, linear_idx_type=paddings.linear_idx_type]) -> IndexList[rank]

Computes the output shape produced by padding input_shape with the before/after amounts given in paddings.

paddings is a flat one-dimensional tensor of length 2 * rank ordered as (before_axis0, after_axis0, before_axis1, after_axis1, ...), and the returned shape has each axis i set to before[i] + input_shape[i] + after[i].

Args:

input_shape (IndexList[rank]): Shape of the tensor before padding.
paddings (TileTensor[DType.int, Storage=paddings.Storage, address_space=paddings.address_space, linear_idx_type=paddings.linear_idx_type]): Flat (before, after) padding sizes for each axis.

Returns:

IndexList[rank]: The padded output shape with the same rank as input_shape.