Skip to content

Fix quantized K cache without FA #680

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Aug 8, 2025
Merged

Fix quantized K cache without FA #680

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
0dce7f9
Prevent assert with quantized K cache and no FA
Aug 8, 2025
f747cbc
Fix MMQ when running with quantized K cache without FA
Aug 8, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion ggml/src/ggml-cuda.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1647,7 +1647,7 @@ static void ggml_cuda_op_mul_mat(
}

const int64_t src1_col_stride = split && used_devices > 1 ? MUL_MAT_SRC1_COL_STRIDE : ne11;
if (!(split && used_devices > 1) && quantization_done && ne11 == 1 && ne12 > 1 && ne13 == 1) {
if (!(split && used_devices > 1) && quantization_done && ne11 == 1 && ne12 > 1 && ne13 == 1 && ne02 == ne12 && ne02 == dst->ne[2]) {
//printf("invoking fast path for %s x %s\n", src0->name, src1->name);
int id = ctx.device;
char * src0_dd_i = dev[id].src0_dd;
Expand Down
2 changes: 1 addition & 1 deletion ggml/src/ggml-cuda/mmq.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -14,7 +14,7 @@ void ggml_cuda_op_mul_mat_q(
const int64_t src1_padded_row_size, cudaStream_t stream) {

const int64_t ne00 = src0->ne[0];
const int64_t nb01 = src0->nb[1];
const int64_t nb01 = ggml_row_size(src0->type, ne00);

const int64_t ne10 = src1->ne[0];
const int64_t ne11 = src1->ne[1];
Expand Down