You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hi ,
Was doing some testing and found that log/geometrically spaced quantization of weights resulted in better performing models. So have implemented a 4bit log /geometric quant and wrapped it structurally with NF4 semantics. PR is #1943
Caveats: I dont have CUDA access/experience so the kernels and anything to do with that needs both review and testing.
THanks
reacted with thumbs up emoji reacted with thumbs down emoji reacted with laugh emoji reacted with hooray emoji reacted with confused emoji reacted with heart emoji reacted with rocket emoji reacted with eyes emoji
Uh oh!
There was an error while loading. Please reload this page.
-
Hi ,
Was doing some testing and found that log/geometrically spaced quantization of weights resulted in better performing models. So have implemented a 4bit log /geometric quant and wrapped it structurally with NF4 semantics. PR is #1943
Caveats: I dont have CUDA access/experience so the kernels and anything to do with that needs both review and testing.
THanks
Beta Was this translation helpful? Give feedback.
All reactions