vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

ggml: update unit tests for the new vec_dot interface

Browse source
This commit is contained in:
Sunita Nadampalli 2024-02-02 21:30:35 +00:00
parent 9cd5b8de6e
commit bca726f06a
4 changed files with 6 additions and 6 deletions
Download patch file Download diff file Expand all files Collapse all files

4
pocs/vdot/q8dot.cpp
View file

@ -156,8 +156,8 @@ int main(int argc, char** argv) {
t1 = std::chrono::high_resolution_clock::now();
float fs;
if (type == 0) funcs.vec_dot(kVecSize * QK4_1, &fs, x40.data(), y.data());
else funcs.vec_dot(kVecSize * QK4_1, &fs, x41.data(), y.data());
if (type == 0) funcs.vec_dot(kVecSize * QK4_1, &fs, 0, x40.data(), 0, y.data(), 0, 1);
else funcs.vec_dot(kVecSize * QK4_1, &fs, 0, x41.data(), 0, y.data(), 0, 1);
t2 = std::chrono::high_resolution_clock::now();
t = 1e-3*std::chrono::duration_cast<std::chrono::nanoseconds>(t2-t1).count();
if (iloop > 3) ggml.addResult(fs, t);

4
pocs/vdot/vdot.cpp
View file

@ -284,8 +284,8 @@ int main(int argc, char** argv) {
else {
auto vdot = ggml_internal_get_type_traits(funcs.vec_dot_type);
vdot.from_float(y1.data(), q8.data(), kVecSize);
if (useQ4_1) funcs.vec_dot(kVecSize, &result, q41.data(), q8.data());
else funcs.vec_dot(kVecSize, &result, q40.data(), q8.data());
if (useQ4_1) funcs.vec_dot(kVecSize, &result, 0, q41.data(), 0, q8.data(), 0, 1);
else funcs.vec_dot(kVecSize, &result, 0, q40.data(), 0, q8.data(), 0, 1);
}
sumq += result;
t2 = std::chrono::high_resolution_clock::now();

2
tests/test-quantize-fns.cpp
View file

@ -87,7 +87,7 @@ static float dot_product_error(
vdot.from_float(test_data2, tmp_q2.data(), test_size);
float result = INFINITY;
qfns.vec_dot(test_size, &result, tmp_q1.data(), tmp_q2.data());
qfns.vec_dot(test_size, &result, 0, tmp_q1.data(), 0, tmp_q2.data(), 0, 1);
const float dot_ref = dot_product(test_data1, test_data2, test_size);

2
tests/test-quantize-perf.cpp
View file

@ -346,7 +346,7 @@ int main(int argc, char * argv[]) {
printf(" %zu values (%.2f MB)\n", size, 4*size/(float)(1024*1024));
auto quantize_fn = [&](void) -> float {
float result;
qfns.vec_dot(size, &result, test_q1, test_q2);
qfns.vec_dot(size, &result, 0, test_q1, 0, test_q2, 0, 1);
return result;
};
size_t quantized_size = ggml_row_size(type, size);