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
For large matrices (e.g. encoders with millions of neurons), the geometry planning (the for n in items loop creating all the indices, etc., in block_impl) is quite slow.
This may also be the case for other GEMV implementations; I haven't checked.
The text was updated successfully, but these errors were encountered:
For large matrices (e.g. encoders with millions of neurons), the geometry planning (the
for n in items
loop creating all the indices, etc., inblock_impl
) is quite slow.This may also be the case for other GEMV implementations; I haven't checked.
The text was updated successfully, but these errors were encountered: