CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-10-27 08:21:30 +00:00
Code Issues Packages Projects Releases Wiki Activity

ggml: fix cuda kernel launch configuration for k_compute_batched_ptrs to support large batch (#16744)

Browse Source 
* fix k_compute_batched_ptrs

* add backend ops test

* Update ggml/src/ggml-cuda/ggml-cuda.cu

Co-authored-by: Johannes Gäßler <johannesg@5d6.de>

* reduce the batch size

---------

Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
This commit is contained in:
leejet
2025-10-27 02:13:31 +08:00
committed by GitHub
parent 73a48c9790
commit bbac6a26b2
2 changed files with 12 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
tests/test-backend-ops.cpp
View File

@@ -6697,6 +6697,9 @@ static std::vector<std::unique_ptr<test_case>> make_test_cases_eval() {
test_cases.emplace_back(new test_mul_mat(type_a, type_b, 16, 1, 1024, {3, 2}, {1, 1}));
test_cases.emplace_back(new test_mul_mat(type_a, type_b, 16, 8, 1024, {3, 2}, {1, 1}));
test_cases.emplace_back(new test_mul_mat(type_a, type_b, 16, 16, 1024, {3, 2}, {1, 1}));
// test cases with large batch size
test_cases.emplace_back(new test_mul_mat(type_a, type_b, 16, 8, 256, {1536, 1}, {1, 1}));
}
}
for (ggml_type type_a : other_types) {