Add Kaggle notebook Dockerfile

[pdmack]

Related to #1057 - pytorch notebook image
Related to #258 - Kaggle image
PTAL
/cc @jlewi
/cc @ankushagarwal
/cc @johnugeorge

Launched for me but obviously not exhaustively tested. Created a contrib dir to give it that YMMV vibe.

Notes (IMHO):

• py3 only
• we should just manually build/push to gcr.io once per release, no other support
• definitely should not be in any argo or release workflows
• waaaay too big an image, there may be some slimming that could be done with my mods but almost certainly it would remain north of 20 Gb

---

This change is 

[pdmack]

/retest

[jlewi]

This is great.

Do you have a prebuilt image that I can push to our public registry?

You should be able to push to gcr.io/kubeflow-dev

/lgtm
/approve

[pdmack]

@jlewi I'll try to push something later today.

[pdmack]

It's up there at kubeflow-dev but getting a SIGABRT running MNIST hello world in d.k.o. Stay tuned.

[pdmack]

Heh, guess the only place I can test this is on my i7-4870HQ:

>>> import tensorflow as tf
2018-07-02 19:36:50.944006: F tensorflow/core/platform/cpu_feature_guard.cc:36] The TensorFlow library was compiled to useAVX2 instructions, but these aren't available on your machine.
Aborted (core dumped)

[pdmack]

/retest

@jlewi, MNIST hello world works on my 2015 MBP (AVX2). Weighs in at a svelte 27 Gb!

gcr.io/kubeflow-dev/kubeflow-kaggle-notebook:latest

[pdmack]

/test all

[jlewi]

GPU test failures look unrelated.
#1126

[jlewi]

/test all

[pdmack]

/test all

[jlewi]

#1128 for the latest test flake.

/test all

[jlewi]

Latest failure looks like a timeout waiting for the minikube workflow; but that ultimately ran successfully.

/test all

[pdmack]

This is fun

[jlewi]

Another timeout but looks like workflows succeeded. Let me bump the timeout.

[pdmack]

minikube is the bane of our e2e existence

[pdmack]

@jlewi I think we now have an MVP at gcr.io/kubeflow-dev/kubeflow-kaggle-notebook:latest

PTAL and cancel hold if you feel it's ready

Name: tensorflow
Version: 1.9.0rc0

Built for AVX2

[pdmack]

/retest

[jlewi]

Retagged it gcr.io/kubeflow-images-public/kaggle-notebook:v20180711
Testing it now.

[pdmack]

Be gentle

[jlewi]

When I tried gcr.io/kubeflow-images-public/kaggle-notebook:v20180711 I got the error
gcr.io/kubeflow-images-public/kaggle-notebook@sha256:30061303799b3f7927e3b6b684b228360d7a8370e2c797972519be407c4d1a75

checking if /home/jovyan volume needs init...
...creating /home/jovyan/work
...load initial content into /home/jovyan...
...done
Execute the command
/usr/local/bin/start.sh: line 62: exec: jupyterhub-singleuser: not found

[pdmack]

I took out too much in that last cut. Stay tuned.

[pdmack]

Update: Just trying to wrangle the addition of the jupyterhub package. It is surprisingly more disruptive than expected.

[pdmack]

@jlewi please retry gcr.io/kubeflow-dev/kubeflow-kaggle-notebook:latest
feel a little more confident in this one
but adding jupyterhub 0.8.1 triggers some incompats

tensorflow 1.9.0rc0 has requirement setuptools<=39.1.0, but you'll have setuptools 40.0.0 which is incompatible.
kmodes 0.9 has requirement scikit-learn<0.20.0,>=0.19.0, but you'll have scikit-learn 0.20.dev0 which is incompatible.
kmeans-smote 0.1.0 has requirement imbalanced-learn<0.4,>=0.3.1, but you'll have imbalanced-learn 0.4.0.dev0 which is incompatible.
kmeans-smote 0.1.0 has requirement scikit-learn<0.20,>=0.19.0, but you'll have scikit-learn 0.20.dev0 which is incompatible.
hep-ml 0.5.0 has requirement theano==0.8.2, but you'll have theano 1.0.2 which is incompatible.
fastai 0.7.0 has requirement torch<0.4, but you'll have torch 0.4.0 which is incompatible.

Actually I'm assuming that, but maybe this happens in the original also.

[jlewi]

Thanks I'm on the road today so might not get to it until this weekend

…

On Fri, Jul 13, 2018, 9:51 AM Pete MacKinnon ***@***.***> wrote: @jlewi <https://github.com/jlewi> please retry gcr.io/kubeflow-dev/kubeflow-kaggle-notebook:latest feel a little more confident in this one but adding jupyterhub 0.8.1 triggers some incompats tensorflow 1.9.0rc0 has requirement setuptools<=39.1.0, but you'll have setuptools 40.0.0 which is incompatible. kmodes 0.9 has requirement scikit-learn<0.20.0,>=0.19.0, but you'll have scikit-learn 0.20.dev0 which is incompatible. kmeans-smote 0.1.0 has requirement imbalanced-learn<0.4,>=0.3.1, but you'll have imbalanced-learn 0.4.0.dev0 which is incompatible. kmeans-smote 0.1.0 has requirement scikit-learn<0.20,>=0.19.0, but you'll have scikit-learn 0.20.dev0 which is incompatible. hep-ml 0.5.0 has requirement theano==0.8.2, but you'll have theano 1.0.2 which is incompatible. fastai 0.7.0 has requirement torch<0.4, but you'll have torch 0.4.0 which is incompatible. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#1109 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAvcAwah8dUtMOohXkp7DVoldPVjh2EBks5uGNAvgaJpZM4U9zoq> .

[jlewi]

Pushed to:
gcr.io/kubeflow-images-public/kaggle-notebook:v20180713

Trying it now.

[jlewi]

Was able to successfully launch a notebook after changing the home directory

ks param set kubeflow-core jupyterNotebookPVCMount /home/jovyan/work

I loaded up this notebook. The imports worked. I was missing the data for the notebook so it didn't work.

/lgtm
/approve

[k8s-ci-robot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: jlewi

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Needs approval from an approver in each of these files:

• OWNERS [jlewi]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[jlewi]

Thank you so much @pdmack for al the hard working making this work.

[pdmack]

PTL

[pdmack]

/hold cancel