Saw non-null label

[kei6]

Running command: python multi_gputrain.py Error: Saw non-null label

I've fixed this error by change self.NUM_CLASSES into self.NUM_CLASSES+1 in recognizer.py, because CTC loss in tensorflow need number of class plus 1 to work with:

self.logits = slim.fully_connected(lstmout, self.NUM_CLASSES+1, activation_fn=None, scope='fc')

This action help to pass the bug. I'm on my work to check its right for training.