--optimized-turbo not working anymore?
pinilpypinilpy opened this issue · comments
Name says it all, on colab free tier the system ram runs out and the process gets killed.
A fix seems to be changing
if not opt.optimized: model.cuda()
to
model.cuda()
on line 360 and 361
Yes that was supposed to be for OOM optimized users but apparently it doesn't work, I can only go off what users report I dont have a 4gb vram card, sorry. The commit with that change has been reverted.