[updated] improve softmax implementation (#11189)
* adjust softmax implementation * theano only handle 2-dim case for softmax [theano softmax](http://deeplearning.net/software/theano/library/tensor/nnet/nnet.html#theano.tensor.nnet.nnet.softmax) * clean softmax activation and add 3d softmax test * fix axis * fix 1d case and test case * fix 1d case and test case * change to standard test * fix test case * Update activations_test.py * got the value error back. * Moved correctness test to the backend tests. * Added a correctness test to activations_test.py.
Showing
想要评论请 注册 或 登录