Tensorflow frontend - Softmax


Currently, a TF softmax op gets mapped to NNVM “Flatten + Softmax” symbol.


I see that softmax now supports axis. So, is flatten symbol still required? If not, I can work on it and send a PR.


I think ‘take’ op is similar here


@janimesh yes, the extra fixes not needed now.
Please send a PR and also add a test case for the same. Thanks.