Open vpuri3 opened 4 months ago
Hm. Not sure how to fix this on my end because the broadcasted copyto!
method for Diagonal
s does scalar indexing in a for
loop. I don't want to special-case this for each matrix type. Might be worth figuring out how
1f0 .* Diagonal(CUDA.ones(4))
works. Maybe that's being special-cased by CUDA
?