Public/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-09-18 16:55:05 +02:00
Code Issues Packages Projects Releases Wiki Activity

CUDA: Add top-k implementation

Browse Source
This commit is contained in:
Oliver Simons
2025-11-21 12:01:32 +01:00
parent ec047e12ee
commit f23b306cc5
8 changed files with 189 additions and 13 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tests/test-backend-ops.cpp
+5
View File
@@ -8035,6 +8035,11 @@ static std::vector<std::unique_ptr<test_case>> make_test_cases_perf() {
test_cases.emplace_back(new test_argsort(GGML_TYPE_F32, {200000, 1, 1, 1}));
test_cases.emplace_back(new test_argsort(GGML_TYPE_F32, {200000, 16, 1, 1}));
test_cases.emplace_back(new test_top_k(GGML_TYPE_F32, {65000, 16, 1, 1}, 40));
test_cases.emplace_back(new test_top_k(GGML_TYPE_F32, {65000, 1, 1, 1}, 40));
test_cases.emplace_back(new test_top_k(GGML_TYPE_F32, {65000, 1, 1, 1}, 1));
test_cases.emplace_back(new test_top_k(GGML_TYPE_F32, {200000, 1, 1, 1}, 400));
test_cases.emplace_back(new test_top_k(GGML_TYPE_F32, {200000, 1, 1, 1}, 40));
test_cases.emplace_back(new test_top_k(GGML_TYPE_F32, {200000, 1, 1, 1}, 1));
return test_cases;
}