ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-07-12 02:05:50 +00:00
Code Issues Packages Projects Releases Wiki Activity

save outputs

Browse Source
This commit is contained in:
coderfeli
2025-02-05 08:19:41 +00:00
parent 9afc4a0b4d
commit e947d11ea0
1 changed files with 4 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
example/65_gemm_multiply_multiply/moe_gemm_fp16.cpp
View File

@@ -80,9 +80,10 @@ struct MultiplyMultiply
// }
};
void preShuffleBuffer(const F16* src, F16* dst, int N, int K, int NXdl)
{
int KPack = 8;
int KPack = 16 / sizeof(F16);
int NLane = NXdl;
int KLane = 64 / NLane;
@@ -226,7 +227,8 @@ int main(int argc, char* argv[])
else
sorted_token_ids.mData[i] = tokens;
}
expert_ids.savetxt("expert_ids.txt", "int");
sorted_token_ids.savetxt("sorted_token_ids.txt", "int");
Tensor<A0DataType> a0_t_k(HostTensorDescriptor({tokens, K}, {K, 1}));
Tensor<B0DataType> b0_e_n_k(HostTensorDescriptor({experts, N, K}, {N*K, K, 1}));
Tensor<B0DataType> b0_preshuffled(HostTensorDescriptor({experts, N, K}, {N*K, K, 1}));