Skip to main content

Mojo function

split_k_reduce

split_k_reduce[c_type: DType, work_space_type: DType, c_layout: Layout, work_space_layout: Layout, elementwise_lambda_fn: OptionalReg[elementwise_epilogue_type] = None](c: LayoutTensor[c_type, c_layout, origin], work_space: LayoutTensor[work_space_type, work_space_layout, origin], ctx: DeviceContext)

Was this page helpful?