Training with multiple data sets

I'm trying to train a neural network that recognizes shapes. I'm using the algorithms built into the Deep Learning Package and I was wondering if it's possible to train on multiple data sets. Currently it feels like it trains and then starts a new neural net when it starts training on the second data set. I also tried combining the data into one big array, but then I can't put in an individual data set to check if it's giving decent results. Is there another way to train a neural network with multiple data sets?