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
When I printed out the MLIR, I found that triton uses float32 operations and then converts to float16.
Is triton considering direct float16 computation, or can it accept a PR? Can we enable it as a compilation option?
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
-
When I try to solve this issue #2680 and simplify it to the following test:
Let's run this test:
Surprisingly, triton has higher precision.
When I printed out the MLIR, I found that triton uses float32 operations and then converts to float16.
Is triton considering direct float16 computation, or can it accept a PR? Can we enable it as a compilation option?
Beta Was this translation helpful? Give feedback.
All reactions