Mojo function
grouped_matmul_sm100
grouped_matmul_sm100[c_type: DType, c_shape: DimList, a_type: DType, a_shape: DimList, b_type: DType, b_shape: DimList, //, *, transpose_b: Bool = True, mma_shape: IndexList[3] = Index(64, 128, 16), block_tile_shape: IndexList[3] = Index(64, 128, 64), elementwise_lambda_fn: OptionalReg[elementwise_epilogue_type] = None](c: NDBuffer[c_type, 2, MutAnyOrigin, c_shape], a: NDBuffer[a_type, 2, MutAnyOrigin, a_shape], a_offsets: NDBuffer[DType.uint32, 1, MutAnyOrigin], max_num_tokens_per_expert: Int, b: NDBuffer[b_type, 3, MutAnyOrigin, b_shape], expert_ids: NDBuffer[DType.int32, 1, MutAnyOrigin], num_active_experts: Int, ctx: DeviceContext)
Was this page helpful?
Thank you! We'll create more content like this.
Thank you for helping us improve!