furyhawk/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-07-21 10:15:53 +00:00
Code Issues Packages Projects Releases Wiki Activity

vulkan: support all backend tests for SQR/SQRT/SIN/COS/CLAMP/LEAKY_RELU/NORM (#24582)

Browse Source 
* vulkan: make SQR/SQRT/SIN/COS/CLAMP/LEAKY_RELU use unary.comp

* vulkan: make NORM support noncontig

* add noncontiguous row test cases for norm/l2_norm, handle this in the CPU backend and l2_norm.comp

* fix supports_op for cuda and webgpu
This commit is contained in:
Jeff Bolz
2026-06-23 09:48:24 -05:00
committed by GitHub
parent 92e854ab83
commit 72a9269172
15 changed files with 171 additions and 201 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tests/test-backend-ops.cpp
+26 -8
View File
@@ -3298,21 +3298,29 @@ struct test_norm : public test_case {
const std::array<int64_t, 4> ne;
const bool v; // whether a is a non-contiguous view
const float eps;
const bool noncontig_rows;
std::string vars() override {
return VARS_TO_STR4(type, ne, v, eps);
return VARS_TO_STR5(type, ne, v, eps, noncontig_rows);
}
test_norm(ggml_type type = GGML_TYPE_F32,
std::array<int64_t, 4> ne = {64, 5, 4, 3},
bool v = false,
float eps = 1e-6f)
: type(type), ne(ne), v(v), eps(eps) {}
float eps = 1e-6f,
bool noncontig_rows = false)
: type(type), ne(ne), v(v), eps(eps), noncontig_rows(noncontig_rows) {}
ggml_tensor * build_graph(ggml_context * ctx) override {
ggml_tensor * a = ggml_new_tensor(ctx, type, 4, ne.data());
const std::array<int64_t, 4> ne_a = noncontig_rows ?
std::array<int64_t, 4>{ ne[1], ne[0], ne[2], ne[3] } : ne;
ggml_tensor * a = ggml_new_tensor(ctx, type, 4, ne_a.data());
ggml_set_name(a, "a");
if (noncontig_rows) {
a = ggml_permute(ctx, a, 1, 0, 2, 3);
ggml_set_name(a, "permuted a");
}
if (v) {
a = ggml_view_4d(ctx, a, a->ne[0]/2, a->ne[1]/2, a->ne[2]/2, a->ne[3]/2, a->nb[1], a->nb[2], a->nb[3], 0);
ggml_set_name(a, "view of a");
@@ -6193,21 +6201,29 @@ struct test_l2_norm : public test_case {
const std::array<int64_t, 4> ne;
const float eps;
bool v;
bool noncontig_rows;
std::string vars() override {
return VARS_TO_STR4(type, ne, eps, v);
return VARS_TO_STR5(type, ne, eps, v, noncontig_rows);
}
test_l2_norm(ggml_type type = GGML_TYPE_F32,
std::array<int64_t, 4> ne = {64, 64, 320, 1},
float eps = 1e-12f,
bool v = false)
: type(type), ne(ne), eps(eps), v(v) {}
bool v = false,
bool noncontig_rows = false)
: type(type), ne(ne), eps(eps), v(v), noncontig_rows(noncontig_rows) {}
ggml_tensor * build_graph(ggml_context * ctx) override {
ggml_tensor * a = ggml_new_tensor(ctx, type, 4, ne.data());
const std::array<int64_t, 4> ne_a = noncontig_rows ?
std::array<int64_t, 4>{ ne[1], ne[0], ne[2], ne[3] } : ne;
ggml_tensor * a = ggml_new_tensor(ctx, type, 4, ne_a.data());
ggml_set_name(a, "a");
if (noncontig_rows) {
a = ggml_permute(ctx, a, 1, 0, 2, 3);
ggml_set_name(a, "permuted a");
}
if (v) {
a = ggml_view_4d(ctx, a, a->ne[0]/2, a->ne[1]/2, a->ne[2]/2, a->ne[3]/2, a->nb[1], a->nb[2], a->nb[3], 0);
ggml_set_name(a, "view of a");
@@ -8282,9 +8298,11 @@ static std::vector<std::unique_ptr<test_case>> make_test_cases_eval() {
test_cases.emplace_back(new test_norm(GGML_TYPE_F32, { n, 5, 4, 3 }, v, eps));
test_cases.emplace_back(new test_rms_norm(GGML_TYPE_F32, { n, 5, 4, 3 }, v, eps));
}
test_cases.emplace_back(new test_norm(GGML_TYPE_F32, { n, 5, 4, 3 }, false, eps, true));
test_cases.emplace_back(new test_rms_norm_back(GGML_TYPE_F32, { n, 5, 4, 3 }, eps));
test_cases.emplace_back(new test_l2_norm(GGML_TYPE_F32, { n, 5, 4, 3 }, eps, false));
test_cases.emplace_back(new test_l2_norm(GGML_TYPE_F32, { n, 5, 4, 3 }, eps, true));
test_cases.emplace_back(new test_l2_norm(GGML_TYPE_F32, { n, 5, 4, 3 }, eps, false, true));
}
}