Lumpiasty/llama.cpp
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

metal : add TQ2_0 support (#26980)

Browse Source 
* metal: add TQ2_0 support

Add support for the GGML_TYPE_TQ2_0 (ternary, 2 bits per element) type in
the Metal backend.

Assisted-by: llama.cpp:DeepSeek-v4-Flash-0731

* cont : optimize mul_mv kernel

- float ops over integer ops
- precalculate sums
- hoist coef out of the inner loop
- contiguous y loads

llama.cpp:DeepSeek-v4-Flash-0731
This commit is contained in:
Georgi Gerganov
2026-08-13 14:33:53 +03:00
committed by GitHub
parent f65e568fd8
commit 4a84b0ad10
5 changed files with 232 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tests/test-backend-ops.cpp
+7 -1
View File
@@ -8995,7 +8995,13 @@ static std::vector<std::unique_ptr<test_case>> make_test_cases_eval() {
for (ggml_type type_a : all_types) {
for (int i = 1; i < 10; ++i) {
test_cases.emplace_back(new test_mul_mat(type_a, GGML_TYPE_F32, 16, i, 256, { 1, 1}, {1, 1}));
test_cases.emplace_back(new test_mul_mat(type_a, GGML_TYPE_F32, 16, i, 1*256, { 1, 1}, {1, 1}));
//test_cases.emplace_back(new test_mul_mat(type_a, GGML_TYPE_F32, 12, i, 2*256, { 2, 1}, {1, 1}));
//test_cases.emplace_back(new test_mul_mat(type_a, GGML_TYPE_F32, 11, i, 3*256, { 1, 3}, {5, 1}));
//test_cases.emplace_back(new test_mul_mat(type_a, GGML_TYPE_F32, 13, i, 4*256, { 2, 3}, {1, 1}));
//test_cases.emplace_back(new test_mul_mat(type_a, GGML_TYPE_F32, 17, i, 31*256, { 4, 1}, {1, 1}));
//test_cases.emplace_back(new test_mul_mat(type_a, GGML_TYPE_F32, 18, i, 32*256, { 1, 1}, {8, 1}));
//test_cases.emplace_back(new test_mul_mat(type_a, GGML_TYPE_F32, 19, i, 33*256, { 1, 1}, {1, 1}));
}
}