thek0tyara/llama-cpp-turboquant
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions 4

Added template parameters to opmult class

Browse source
This commit is contained in:
Alejandro Saucedo 2020-08-22 13:33:21 +01:00
parent c92425dd87
commit d74a999e12
4 changed files with 43 additions and 10 deletions
Download patch file Download diff file Expand all files Collapse all files

5
src/OpMult.hpp
View file

@ -17,6 +17,7 @@
namespace kp {
template<uint32_t tX = 0, uint32_t tY = 0, uint32_t tZ = 0>
class OpMult : public OpBase
{
public:
@ -40,6 +41,10 @@ class OpMult : public OpBase
std::shared_ptr<Tensor> mTensorRHS;
std::shared_ptr<Tensor> mTensorOutput;
std::shared_ptr<Tensor> mTensorOutputStaging;
uint32_t mX;
uint32_t mY;
uint32_t mZ;
};
} // End namespace kp