If your GPU runs out of memory while training a CNN$,$ what is NOT something you could do to try to solve the problem?

Question $2$ options:

Increase the mini$-$batch size.

Reduce dimensionality using a larger stride in one or more layers.

Remove one or more layers.

Distribute the CNN across multiple devices.