Skip to content

vulkan: Handle FA with all -inf mask values #16447

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

Open
wants to merge 1 commit into
base: master
Choose a base branch
from
+8 −4
Open

vulkan: Handle FA with all -inf mask values #16447

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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-vulkan/vulkan-shaders/flash_attn.comp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -345,7 +345,7 @@ void main() {

float Lfrcp[Br];
[[unroll]] for (uint32_t r = 0; r < Br; ++r) {
Lfrcp[r] = 1.0 / Lf[r];
Lfrcp[r] = (Lf[r] == 0.0) ? 0.0 : (1.0 / Lf[r]);
}

[[unroll]] for (uint32_t d = 0; d < HSV_per_thread / 4; ++d) {
Expand Down
2 changes: 1 addition & 1 deletion ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_cm1.comp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -380,7 +380,7 @@ void main() {

float Lfrcp[rows_per_thread];
[[unroll]] for (uint32_t r = 0; r < rows_per_thread; ++r) {
Lfrcp[r] = 1.0 / Lf[r];
Lfrcp[r] = (Lf[r] == 0.0) ? 0.0 : (1.0 / Lf[r]);
}

[[unroll]] for (uint32_t d = 0; d < HSV_per_thread / 4; ++d) {
Expand Down
6 changes: 5 additions & 1 deletion ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_cm2.comp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -121,7 +121,11 @@ void main() {
const float NEG_FLT_MAX_OVER_2 = uintBitsToFloat(0xFEFFFFFF);

L = coopmat<ACC_TYPE, gl_ScopeWorkgroup, Br, Bc, gl_MatrixUseAccumulator>(0);
#if defined(ACC_TYPE_MAX)
M = coopmat<ACC_TYPE, gl_ScopeWorkgroup, Br, Bc, gl_MatrixUseAccumulator>(-ACC_TYPE_MAX / ACC_TYPE(2));
#else
M = coopmat<ACC_TYPE, gl_ScopeWorkgroup, Br, Bc, gl_MatrixUseAccumulator>(NEG_FLT_MAX_OVER_2);
#endif

coopmat<ACC_TYPE, gl_ScopeWorkgroup, Br, Bc, gl_MatrixUseAccumulator> slopeMat = coopmat<ACC_TYPE, gl_ScopeWorkgroup, Br, Bc, gl_MatrixUseAccumulator>(1.0);

Expand Down Expand Up @@ -294,7 +298,7 @@ void main() {

[[unroll]]
for (int k = 0; k < Ldiag.length(); ++k) {
Ldiag[k] = ACC_TYPE(1.0) / Ldiag[k];
Ldiag[k] = (Ldiag[k] == 0.0) ? ACC_TYPE(0.0) : (ACC_TYPE(1.0) / Ldiag[k]);
}

O = Ldiag*O;
Expand Down
2 changes: 1 addition & 1 deletion ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_split_k_reduce.comp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -91,7 +91,7 @@ void main() {
L = L*ms + vs;
}

L = 1.0 / L;
L = (L == 0.0) ? 0.0 : 1.0 / L;

// D dimension is split across workgroups in the y dimension
uint d = tid + gl_WorkGroupID.y * BLOCK_SIZE;
Expand Down
Loading