Open ZJLi2013 opened 2 months ago
to debug 02_gemm_add_add_fastgelu with client api, I tried to enable arg.Print() under Invoker:;Run() as following:
02_gemm_add_add_fastgelu
// Invoker struct Invoker : public BaseInvoker { using Argument = DeviceOp::Argument; float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{}) { // Fix ME arg.Print(); ...
and the original Print()
while during building, these TensorDescriptor , a_grid_desc_mk, b_grid_desc_nk, ds_grid_desc_mn, e_grid_desc_mn doesn't support << operation, give building errrors.
<<
as TensorDescriptor class has its own Print() , looks the following way works:
void Print() const { std::cout << "{ \n" ; a_grid_desc_m_k_.Print() ; b_grid_desc_n_k_.Print(); static_for<0, NumDTensor, 1>{}( [&](auto i) { ds_grid_desc_m_n_[i].Print(); }); e_grid_desc_m_n_.Print(); std::cout >> "} \n" ; }
Thanks for clarifying this
David
ubuntu 20.04
Ryzen
AMD Instinct MI300
rocm 6.1.3
ROCm 6.0.0
No response
related to https://github.com/ROCm/composable_kernel/issues/419 ?
Problem Description
to debug
02_gemm_add_add_fastgelu
with client api, I tried to enable arg.Print() under Invoker:;Run() as following:and the original Print()
while during building, these TensorDescriptor , a_grid_desc_mk, b_grid_desc_nk, ds_grid_desc_mn, e_grid_desc_mn doesn't support
<<
operation, give building errrors.as TensorDescriptor class has its own Print() , looks the following way works:
Thanks for clarifying this
David
Operating System
ubuntu 20.04
CPU
Ryzen
GPU
AMD Instinct MI300
Other
rocm 6.1.3
ROCm Version
ROCm 6.0.0
ROCm Component
No response
Steps to Reproduce
No response
(Optional for Linux users) Output of /opt/rocm/bin/rocminfo --support
No response
Additional Information
No response