csarofeen/pytorch

Update nvFuser number math to be in float

Closed this issue · 0 comments

mruberry commented

Currently tensor x float operations result in float64 results