Commit e947d11e authored by coderfeli's avatar coderfeli
Browse files

save outputs

parent 9afc4a0b
...@@ -80,9 +80,10 @@ struct MultiplyMultiply ...@@ -80,9 +80,10 @@ struct MultiplyMultiply
// } // }
}; };
void preShuffleBuffer(const F16* src, F16* dst, int N, int K, int NXdl) 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 NLane = NXdl;
int KLane = 64 / NLane; int KLane = 64 / NLane;
...@@ -226,7 +227,8 @@ int main(int argc, char* argv[]) ...@@ -226,7 +227,8 @@ int main(int argc, char* argv[])
else else
sorted_token_ids.mData[i] = tokens; 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<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_e_n_k(HostTensorDescriptor({experts, N, K}, {N*K, K, 1}));
Tensor<B0DataType> b0_preshuffled(HostTensorDescriptor({experts, N, K}, {N*K, K, 1})); Tensor<B0DataType> b0_preshuffled(HostTensorDescriptor({experts, N, K}, {N*K, K, 1}));
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment