Skip to content
Merged
Changes from all commits
Commits
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
7 changes: 4 additions & 3 deletions ggml/src/ggml-opencl/kernels/flash_attn_f32.cl
Original file line number Diff line number Diff line change
Expand Up @@ -4,6 +4,7 @@
#define ACC_TYPE4 float4
#define DATA_TYPE float
#define DATA_TYPE4 float4
#define MASK_DATA_TYPE half
#define CONVERT_ACC4(x) (x)
#define CONVERT_DATA4(x) (x)

Expand Down Expand Up @@ -148,7 +149,7 @@ __kernel void flash_attn_f32(
if (k_row1 >= n_kv) score1 = -INFINITY;

if (mask_base != NULL) {
const global DATA_TYPE* mask_ptr = (const global DATA_TYPE*)(mask_base + my_query_row * mask_nb1);
const global MASK_DATA_TYPE* mask_ptr = (const global MASK_DATA_TYPE*)(mask_base + my_query_row * mask_nb1);
if (k_row0 < n_kv) score0 += slope * (ACC_TYPE)mask_ptr[k_row0];
if (k_row1 < n_kv) score1 += slope * (ACC_TYPE)mask_ptr[k_row1];
}
Expand Down Expand Up @@ -281,7 +282,7 @@ __kernel void flash_attn_f32_q1(
}
ACC_TYPE score = (dot_acc.s0 + dot_acc.s1 + dot_acc.s2 + dot_acc.s3) * scale;
if (mask_base != NULL) {
const global DATA_TYPE* mask_ptr = (const global DATA_TYPE*)(mask_base);
const global MASK_DATA_TYPE* mask_ptr = (const global MASK_DATA_TYPE*)(mask_base);
score += slope * (ACC_TYPE)mask_ptr[k_idx];
}
if (logit_softcap > 0.0f) {
Expand Down Expand Up @@ -317,7 +318,7 @@ __kernel void flash_attn_f32_q1(
}
ACC_TYPE score = (dot_acc.s0 + dot_acc.s1 + dot_acc.s2 + dot_acc.s3) * scale;
if (mask_base != NULL) {
const global DATA_TYPE* mask_ptr = (const global DATA_TYPE*)(mask_base);
const global MASK_DATA_TYPE* mask_ptr = (const global MASK_DATA_TYPE*)(mask_base);
score += slope * (ACC_TYPE)mask_ptr[k_idx];
}
if (logit_softcap > 0.0f) {
Expand Down
Loading