ScaledGemmTunableOp Class — pytorch Architecture
Architecture documentation for the ScaledGemmTunableOp class in TunableGemm.h from the pytorch codebase.
Entity Profile
Source Code
aten/src/ATen/cuda/tunable/TunableGemm.h lines 306–327
class ScaledGemmTunableOp : public TunableOp<ScaledGemmParams<CT>> {
public:
ScaledGemmTunableOp() {
this->RegisterOp(std::string("Default"), std::make_unique<DefaultScaledGemmOp<CT>>());
#ifdef USE_ROCM
for (auto&& [name, op] : GetHipBlasLtScaledGemmTypeStringAndOps<AT, BT, CT, ALayout, BLayout>()) {
this->RegisterOp(std::move(name), std::move(op));
}
#endif
this->RegisterOp(std::string("Default"), std::make_unique<DefaultScaledGemmOp<CT>>());
}
std::string Signature() override {
return fmt::sprintf("ScaledGemmTunableOp_%s_%s_%s_%c%c",
TypeName<AT>(AT{}),
TypeName<BT>(BT{}),
TypeName<CT>(CT{}),
BlasOpToString(ALayout), BlasOpToString(BLayout));
}
};
Source
Analyze Your Own Codebase
Get architecture documentation, dependency graphs, and domain analysis for your codebase in minutes.
Try Supermodel Free