Specify your own dataset for a model?

Say if I have:

model = ConvnetBuilder(f = resnet34, c = 10, ps = 0.5, xtra_fc = [1024,1024], is_multi = False, is_reg = False)

And I want it to train it on cifar10 that I load in from keras:

from keras.datasets import cifar10
(x_train, y_train), (x_test, y_test) = cifar10.load_data()

how do I train on that data? Also how can I enable transformations on the training/testing data?