My error makes test unable to support theano and it is fixed now, but theano bug problem in train_model still can't be solved, maybe a new loss function can solve this problem.
At least this page keras-users/EhWwuq6R0lQ says that the problem may be related to that sparse_categorical_crossentropy function.
I can only do these.
Very good repo!
My error makes test unable to support theano and it is fixed now, but theano bug problem in train_model still can't be solved, maybe a new loss function can solve this problem. At least this page keras-users/EhWwuq6R0lQ says that the problem may be related to that sparse_categorical_crossentropy function. I can only do these. Very good repo!