vrr/prima.cpp
2
0
Fork 0
mirror of https://github.com/Lizonghang/prima.cpp.git synced 2025-09-10 12:04:35 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

use small tensors for test on cpu

Browse source
This commit is contained in:
Zonghang Li 2024-12-12 15:52:27 +04:00
parent b642d70188
commit 6ea692d974
1 changed files with 1 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

2
common/profiler.cpp
View file

@ -101,7 +101,7 @@ uint32_t device_cpu_cores() {
static float device_flops(struct llama_model * model, enum ggml_type src0t, enum ggml_type src1t, enum profiler_backend_type btype, int n_threads) { static float device_flops(struct llama_model * model, enum ggml_type src0t, enum ggml_type src1t, enum profiler_backend_type btype, int n_threads) {
int n_repeat = 1; int n_repeat = 1;
int n_embd = std::min(llama_n_embd(model), 4096); int n_embd = std::min(llama_n_embd(model), 4096);
// if (btype == PROFILER_BACKEND_TYPE_CPU) n_embd /= 8; // simulate small tensor calculation on cpu if (btype == PROFILER_BACKEND_TYPE_CPU) n_embd /= 8; // simulate small tensor calculation on cpu
std::vector<float> matrix_A(n_embd * n_embd, 1.0f); std::vector<float> matrix_A(n_embd * n_embd, 1.0f);
std::vector<float> matrix_B(n_embd * n_embd, 1.0f / n_embd); std::vector<float> matrix_B(n_embd * n_embd, 1.0f / n_embd);