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_partition

def get_partition(task_id: Int, num_partitions: IndexList[Int(4)], conv_shape: ConvShape, micro_kernel_height: Int, micro_kernel_f_size: Int) -> ConvPartition

Computes the ConvPartition work range for a given task id.

Args:

task_id (Int): Linear task identifier.
num_partitions (IndexList[Int(4)]): Per-dimension partition counts as (batch_group, channel, filter, spatial).
conv_shape (ConvShape): Convolution shape describing the workload.
micro_kernel_height (Int): Micro-kernel height, used as the spatial work unit when output loops are merged.
micro_kernel_f_size (Int): Micro-kernel size in the filter dimension.

Returns:

ConvPartition: The ConvPartition describing the offset and size of this task's work across batch/group, channel, filter, and spatial dimensions.