bhack
bhack
We have now a PR at https://github.com/tensorflow/tensorflow/pull/57630 if you want to support/review/imporve this baseline.
I've added a few comment. More in general I think that we need to distribute an official Docker devel image that could be used also in Github Action to be...
Have you tried to set the `average` arg: https://www.tensorflow.org/addons/api_docs/python/tfa/metrics/F1Score
Can you share a small google Colab to reproduce this?
Thanks for sharing the roadmap. It could be a little bit hard to understand steps mentioning "internal/our" requirements but I think it is expected. Taking a look at the new...
Was this done for `shufflenet`?
What I meant is that shufflenet has not this randomizzation requirements so probably we need to do something different. I don't think that we care about gradients in KLPs cause...
Yes like: https://github.com/tensorpack/tensorpack/blob/master/examples/ImageNetModels/shufflenet.py#L40-L49
Let's see if we could unify the two use cases. But It isn't clear to me if we want to reuse the KLP namespace in differentiable layers...
I think that we could separate Shuffle in Shufflenet and KLP layers as they have two very different behaviors. Can we maintain the original impl for KLP?