llama.cpp
cuda : prevent integer truncation and overflow errors when using KQ mask strides in flash_attn_mask_to_KV_max kernel
#24945
Merged
Merged
cuda : prevent integer truncation and overflow errors when using KQ mask strides in flash_attn_mask_to_KV_max kernel #24945
JohannesGaessler
merged
1 commit into
ggml-org:master
from
fairydreaming:stride-narrow-conv-fix
am17an
approved these changes
on 2026-06-23
JohannesGaessler
approved these changes
on 2026-06-30
cuda : prevent integer truncation and overflow errors when using KQ m…
4615a025
JohannesGaessler
approved these changes
on 2026-06-30
Login to write a write a comment.
Login via GitHub