Is Gluon in constant development?


#1

Hi everyone

I have some experience with Keras and Tensorflow as backend. Looking at Internet for ways to execute numpy code over GPUs I encountered the package NDArray and I read the documentation of MXnet, in specific Gluon API, I have to say that it is very good and motivating for me!!.

I’m thinking to move to use Mxnet with Gluon API for a rapid prototyping. However, I’m wonder about the support of Gluon API, I looked in https://github.com/gluon-api/gluon-api and I have seen that it only has 6 contributors and 13 commits (the last commit was made about 6 months ago).

My questions are:
Is the Gluon API in constant development?
Is better to use keras-apache-mxnet instead?

Best regards,


#2

Hi, using mxnet with the gluon API is the right call and there is a lot of support from the mxnet community for gluon, the code for gluon is now the mxnet repo and is being developed there, so you get gluon when you do a mxnet import. https://github.com/apache/incubator-mxnet/tree/master/python/mxnet/gluon


#3

Answering your questions more pointedly.

Yes, gluon is in constant development

No, you should stick with gluon unless you want to do something like comparing backends or you really really really prefer keras apis.


#4

Thanks @sad for your response. I will taste gluon!.