[TF-TRT] Add LogSoftmax Support for TF-TRT
This PR adds TF-TRT support to tf.nn.log_softmax operation. This is performed using the formula logsoftmax = logits - log(reduce_sum(exp(logits), axis=-1)) . The implemented TRT layers are fused into a single op.
@DEKHTIARJonathan @tfeher : Please review the changes.
@pavanimajety Please add a PR description.
Looks good to me now. Would you please squash the commits?
Done. Thanks for reviewing
Sorry, rebased for further work by mistake.
This is merged, wonder why it doesn't show as merged here.
This is merged, wonder why it doesn't show as merged here.
Would you mind checking again? I don't see in tensorflow:master : https://github.com/tensorflow/tensorflow/tree/master/tensorflow/compiler/tf2tensorrt/convert/ops
It did not show as merged here because copybara sync was broken. Fixed in big squash commit 49f97f135a2e1d5d22e60d2a80ec668d53f9708a