ArchTag Class — pytorch Architecture
Architecture documentation for the ArchTag class in gemm_kernel_utils.h from the pytorch codebase.
Entity Profile
Source Code
aten/src/ATen/native/transformers/cuda/mem_eff_attention/gemm_kernel_utils.h lines 102–110
template <typename ArchTag, typename scalar_t_, typename Enable = void>
struct DefaultGemmType {
static constexpr int ThreadK = 8;
static constexpr int WarpK = 8;
static constexpr int kMinimumAlignment = 1;
using InstructionShape = cutlass::gemm::GemmShape<1, 1, 1>;
using OpClass = cutlass::arch::OpClassSimt;
using Operator = cutlass::arch::OpMultiplyAdd;
};
Source
Analyze Your Own Codebase
Get architecture documentation, dependency graphs, and domain analysis for your codebase in minutes.
Try Supermodel Free