run forward/backward passes for each of the loss graphs created by loss-module on the batch.

Each entry in batch as treated as arguments to module/graph to build an expression for example loss

Trains a model using the following parameters:

• model a flare.model/PModel instance
• get-loss-node a function which takes each data input and returns a node representing the loss in the input
• data-batch-fn assumed that calling this function yields a lazy sequence of data batches, looping over each batch and calls get-loss-node
• train-opts, see spec above