Keras: whether does "fit_generator" accept multi input generator like "fit"?

[dongzhuoyao]

from http://keras.io/getting-started/functional-api-guide/#multi-input-and-multi-output-models,

we can see the we can use multi input and multi output by "fit".

but when the data size is huge,it's not practical to load all data into memory and use "fit" method to start training.

so I think up a idea about "fit_generator",but from keras's document,"fit_generator" seems don't accept the generator that generates multi input(in this case,it may return a dict).the "fit_generator" only accepts the generator that returns a tuple which can only descibe single one input.

Does there exist any solution to solve the problem,or I have to hack the core to complete my demand？

[ChristianThomae]

You may return a tuple of lists for multi input/output, like:
yield [X1, X2], [y1, y2]

[hassaan90]

did you able to manege this thing ? I am having similar problem ...