Compared to the L-BFGS optimizer in the original Lua Torch, this is unexpected behavior. Is there anyway I can prevent this increase in memory usage every iteration? This issue doesn’t seem to occur with Adam, but I’d like to use L-BFGS.
1 Like