Skip to content

CUDA: Prefer vector flash decoding kernel for Gemma models #12738

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
Apr 3, 2025
Merged

CUDA: Prefer vector flash decoding kernel for Gemma models #12738

Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
f7d07dd
Prefer vector flash decoding kernel for Gemma models
gaugarg-nv Apr 3, 2025
ce71aba
Update ggml/src/ggml-cuda/fattn.cu
gaugarg-nv Apr 3, 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/fattn.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -299,7 +299,7 @@ void ggml_cuda_flash_attn_ext(ggml_backend_cuda_context & ctx, ggml_tensor * dst
const bool gqa_opt_applies = ((Q->ne[2] / K->ne[2]) % 2 == 0) && mask; // The mma-based kernels have GQA-specific optimizations
const bool mma_needs_data_conversion = K->type != GGML_TYPE_F16 || V->type != GGML_TYPE_F16;
const bool mma_faster_for_bs1 = new_mma_available(cc) && gqa_opt_applies && cc < GGML_CUDA_CC_ADA_LOVELACE && !mma_needs_data_conversion;
const bool can_use_vector_kernel = (Q->ne[0] % (2*warp_size) == 0) && (prec == GGML_PREC_DEFAULT || Q->ne[0] <= 128);
const bool can_use_vector_kernel = (Q->ne[0] % (2*warp_size) == 0) && (Q->ne[0] <= 256);
if (Q->ne[1] == 1 && can_use_vector_kernel && !mma_faster_for_bs1) {
if (prec == GGML_PREC_DEFAULT) {
ggml_cuda_flash_attn_ext_vec_f16(ctx, dst);
Expand Down
Loading