sleepy/llama.cpp
1
0
Fork 0
Code Issues 8 Pull Requests Actions 170 Packages Projects Releases Wiki Activity

Add DIAG for CUDA (#17873)

Browse Source 
* Add DIAG for CUDA

* Refactor parameters
This commit is contained in:
Piotr Wilkin (ilintar)
2025-12-09 20:28:57 +01:00
committed by GitHub
parent 48f47565a7
commit b63509262a
4 changed files with 116 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tests/test-backend-ops.cpp
+29
View File
@@ -6253,6 +6253,31 @@ struct test_solve_tri : public test_case {
}
};
// GGML_OP_DIAG
struct test_diag : public test_case {
const ggml_type type;
const std::array<int64_t, 4> ne;
std::string vars() override { return VARS_TO_STR2(type, ne); }
test_diag(ggml_type type = GGML_TYPE_F32,
std::array<int64_t, 4> ne = { 10, 1, 4, 3 })
: type(type), ne(ne) {}
ggml_tensor * build_graph(ggml_context * ctx) override {
GGML_ASSERT(ne[1] == 1);
ggml_tensor * a = ggml_new_tensor_4d(ctx, type, ne[0], ne[1], ne[2], ne[3]);
ggml_set_param(a);
ggml_set_name(a, "a");
ggml_tensor * out = ggml_diag(ctx, a);
ggml_set_name(out, "out");
return out;
}
};
enum llm_norm_type {
LLM_NORM,
LLM_NORM_RMS,
@@ -7826,6 +7851,10 @@ static std::vector<std::unique_ptr<test_case>> make_test_cases_eval() {
test_cases.emplace_back(new test_fill(-152.0f, GGML_TYPE_F32, { 800, 600, 4, 4 }));
test_cases.emplace_back(new test_fill(3.5f, GGML_TYPE_F32, { 2048, 512, 2, 2 }));
test_cases.emplace_back(new test_diag());
test_cases.emplace_back(new test_diag(GGML_TYPE_F32, { 79, 1, 19, 13 }));
test_cases.emplace_back(new test_diag(GGML_TYPE_F32, { 256, 1, 8, 16 }));
test_cases.emplace_back(new test_solve_tri());
test_cases.emplace_back(new test_solve_tri(GGML_TYPE_F32, { 11, 11, 1, 1 }, { 5, 11, 1, 1 }));
test_cases.emplace_back(new test_solve_tri(GGML_TYPE_F32, { 17, 17, 2, 4 }, { 9, 17, 2, 4 }));