# Dynamically adjusting BatchSize during training of NetChain

Is it possible to dynamically adjust BatchSize during NetTrain training session? For instance, can I create a schedule to adjust BatchSize based on the training round?

-CJ

• That's no possible now.But you can create multi-NetTrain with different BatchSize – partida Nov 17 '17 at 10:05
• In theory, you might be able to write a generator to do this. A generator function replaces the data (ie the second argument) in NetTrain. This generator function gets two arguments - BatchSize and Round so something like the following may work: generator=Function[RandomSample[data, #BatchSize+(10*#Round)]]; NetTrain[net,generator,BatchSize->10]. – Carl Lange Aug 6 '18 at 13:07