HackerTrans
TopNewTrendsCommentsPastAskShowJobs

scv119

no profile record

Submissions

What we talk about when we talk about system design

maheshba.bitbucket.io
259 points·by scv119·vor 3 Jahren·35 comments

comments

scv119
·vor 3 Jahren·discuss
The tweet is referring to a paper that fine tunes Chinese dataset on english base model. I'm not surprised with LoRA's poor result in this setup.
scv119
·vor 3 Jahren·discuss
I believe you can slightly change the flash attention kernel to implement the same kernel of this page attention, since both of them work on the key/value cache at block level.
scv119
·vor 3 Jahren·discuss
Pretty cool stuff and the results are amazing. Hoping we will see virtual memory get standardized in pytorch or cuda.