Support for other Deep Learning Libraries

[NathanNZ]

First of all it's awesome seeing an initiative to help bring a bit of DevOps to ML, something I feel is well overdue - however I have a question just to clarify my own understanding of what KubeFlow actually is.

The Kubeflow project is dedicated to making Machine Learning on Kubernetes easy, portable and scalable. Our goal is not to recreate other services, but to provide a straightforward way for spinning up best of breed OSS solutions.

Regarding this statement in the readme it seems to imply that other OSS solutions will be supported however I'm not sure to what extent, especially regarding TensorFlow. My uncertainty rises from the fact that this is a Google run endeavour and the project name itself (implying the mash up between K8 and TF). As such would it be correct in thinking that other data flow programming libraries are out of this projects scope? (Caffe, CNTK etc)?

[elmiko]

@NathanNZ i think issue #33 covered some of this in the comments. imo, it will be great to see Kubeflow grow beyond the perception that it is just a TensorFlow related project.

[aronchick]

Correct! Though we do include TensorFlow to start, we're not TensorFlow exclusive - at all! We only mention "flow" because it's such a common term in ML. We'd ABSOLUTELY love some help including other frameworks!

…

On Tue, Jan 2, 2018 at 1:26 PM michael mccune ***@***.***> wrote: @NathanNZ <https://github.com/nathannz> i think issue #33 <#33> covered some of this in the comments. imo, it will be great to see Kubeflow grow beyond the perception that it is just a TensorFlow related project. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#74 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AADIde83Ioos3VfH215bGxcW8Xk2NDsPks5tGp72gaJpZM4RPfxR> .

[jlewi]

I came across a CRD for mxnet in deepinsight/mxnet-operator. We'd be very receptive to including this in Kubeflow if someone is interested in taking this up.

Forking the ksonnet prototype for TfJob to create a prototype for mxnet should be fairly straightforward.

is anyone interested in taking this on?

/cc @loadwiki @NathanNZ

[YujiOshima]

@jlewi Hi! It is great. I'm so interested in a CRD for MxNet.
Will you have any plan to add mxnet-operator repository to kubeflow?
That repository does not look active now. But if it would be supported by kubeflow officially, I'm willing to contribute it!

[jlewi]

pytorch work has started
https://github.com/kubeflow/pytorch-operator

caffe2 has started
https://github.com/kubeflow/caffe2-operator