Question about pagedattention

[SherrySwift]

Hi, thanks for your great work! I have a small question.
In kernels/csrc/fused_attention/applyBiasRopeUpdateKVCache.h, I saw that you set store_contiguous_qkv = true. Does it mean that pagedattention is not used when quantize KV Cache into int8/int4?
Looking forward to your reply, thank you!

[ys-2020]

Hi,

Thanks for your interests in QServe. We do use page attention when KV Cache is quantized into int8/int4. Please ignore that flag.