llama.cpp
0eca4d49 - cuda : prevent integer truncation and overflow errors when using KQ mask strides in flash_attn_mask_to_KV_max kernel (#24945)

Commit

131 days ago

cuda : prevent integer truncation and overflow errors when using KQ mask strides in flash_attn_mask_to_KV_max kernel (#24945) Co-authored-by: Stanisław Szymczyk <sszymczy@gmail.com>

References

#24945 - cuda : prevent integer truncation and overflow errors when using KQ mask strides in flash_attn_mask_to_KV_max kernel

Author

fairydreaming

Parents

4f31eedb

llama.cpp 0eca4d49 - cuda : prevent integer truncation and overflow errors when using KQ mask strides in flash_attn_mask_to_KV_max kernel (#24945)

llama.cpp
0eca4d49 - cuda : prevent integer truncation and overflow errors when using KQ mask strides in flash_attn_mask_to_KV_max kernel (#24945)