[WIP] Adding docker containers in Gensim. Fix #497

[danielbdias]

I am adding a docker image to run gensim tutorials.

In this PR I created a Dockerfile with Ubuntu 16.04, Python 3 and gensim dependencies and a docker-compose.yml to run the jupyter notebooks.

[tmylk]

Thanks for the contribution.
Remaining steps:

• docker hub upload
• transfer credentials for docker hub container

[danielbdias]

Hi guys !

I updated the PR with the check in the FAST_VERSION and a README.md to pull this image in docker hub.

Now I need to fix some problems when we are pushing this image directly from docker hub. (Running the image gives a container_linux.go:247: starting container process caused "exec: \"-p\": executable file not found in $PATH" error)

After that, we need to create an official image we will need to create a PR to https://github.com/docker-library/official-images and https://github.com/docker-library/docs

[danielbdias]

Another update: I fixed the problem with docker hub, but now we need to refer to a fixed version of gensim.

For now I created an image called danielbdias/gensim in docker hub to test the environment.

Can you check if all the setup in the Dockerfile is ok? With docker file ok, I will start to apply the missing guidelines to start the PR in docker-library/official-images and build the official container.

[canktech]

consider including binaries for wordrank and fasttext

[tmylk]

And varembed, mallet, dtm

[danielbdias]

Hi guys,

I've found the installation commands for mallet and fasttext (through pip). How can I install the other packages (wordrank, varembed and dtm)?

[tmylk]

https://bitbucket.org/shihaoji/wordrank

https://github.com/rguthrie3/MorphologicalPriorsForWordEmbeddings

https://github.com/fedorn/dtm

[tmylk]

Mallet is actually a java package so it can't be installed through pip
http://mallet.cs.umass.edu/

[danielbdias]

Hi guys !

Giving some updates on this issue after a while: I've updated the dockerfile to support WordRank, MorphologicalPriorsForWordEmbeddings, DTM and Mallet.

Is there a issue yet with the FAST_VERSION... some dependency broke the version.

Fixing this problem, I'll be able to upload a version to docker hub and create a gensim user there.

Thanks !

[tmylk]

Thanks @danielbdias What is the exact error with FAST_VERSION? Does the new version of the C compiler not work?

[tmylk]

This docker will be really useful for testing the wrappers during releases.

[danielbdias]

Hi @tmylk ! I'm receiving FAST_VERSION = -5 when I'm building the docker. I think that the reason is the dependencies that I installed for WordRank, MorphologicalPriorsForWordEmbeddings.

I will check this weekend on a way to solve this.

[menshikh-iv]

It's a nice PR, I would like use this for full test run (with all wrappers)

[tmylk]

This container will be part of our release infrastructure as we need to run all the wrapper tests every release and they are not part of the usual Github commit CI run.

[menshikh-iv]

Continues in #1368