why not use flash attention

[chenqy4933]

I found there is no flash attention in the inference kernel，what is the reason，is the flash attention slow in this case?

[Tomorin523]

Actually we are considering Flash Attention for future optimization, it may take some time.