Is it expected that a ComputationGraph be .fit() on a single DataSetIterator, or can I train a model by invoking .fit() over multiple DataSetIterators on the same ComputationGraph instance ?
@dltrumpet sorry was catching up on questions and found this. You generally use a MultiDataSetIterator for that, not a single one. A MultidataSetIterator will be 1 DataSetIterator per input.
thank you. I guess what I’m really asking is can I invoke .fit() as many times as I want on the same ComputationGraph ?
@dltrumpet yes fit is basically 1 update call.