Adjust batch size in Intel Xeon Phi Clusters?

[Priyadhars_K_Intel]

Hi ,

We have an architecture that lets us to deploy models in a distributed way over single servers with our own balancing policies but would be good to use clusters for training time to take advantage of Intel Xeon distributed deep learning capabilites.  How we can adjust the training batch size (images processed / iteration in training phase) in Intel Xeon Phi clusters ? 

Please help 

[SergeyKostrov]

>>...How we can adjust the training batch size ( images processed / iteration in training phase )... My questions are: Do you have training data sets in a database? Or, how did you store all your data sets? I think you're asking about data sets partitioning for different jobs executed on different computers.