zlacker

>>fheins+(OP)
There's a graveyard of 100s of papers with "approximate near linear time attention."

They always hope the speed increase makes up for the lower quality, but it never does. The quadratic time seems inherent to the problem.

Indeed, there are lower bounds showing that sub n^2 algorithms can't work: https://arxiv.org/pdf/2302.13214

>>thomas+Yc
Dumb question: is the quadratic time complexity for training, inference, or both?

>>cobole+hk
Attention is calculated during the forward pass of the model, which happens in both inference (forward only) and training (forward & backward).