Increase batch_size during network training?

Question

Neural networks

I started training the CNN network with one batch_size, I realized that the GPU RAM allows you to double this number.
I want to stop training and start from the last checkpoint, but with a larger number of batch_size. Will I lose the learning progress I had before?

Answer 1

Too little information.
If you turn on the iron logic, then if less than 1/3 of the training has passed, then when resetting progress, it makes sense to change, otherwise not.