add wordrank in dockerfile

[parulsethi]

Fixes #1457

Used mpich instead of openmpi to avoid multithreading errors in wordrank.

[menshikh-iv]

About wordrank

======================================================================
ERROR: testEnsemble (gensim.test.test_wordrank_wrapper.TestWordrank)
Test ensemble of two embeddings
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/gensim/gensim/test/test_wordrank_wrapper.py", line 37, in setUp
    self.test_model = wordrank.Wordrank.train(self.wr_path, self.corpus_file, self.out_name, iter=6, dump_period=5, period=5, cleanup_files=True)
  File "/gensim/gensim/models/wrappers/wordrank.py", line 161, in train
    output = utils.check_output(args=cmd)
  File "/gensim/gensim/utils.py", line 1177, in check_output
    raise error
CalledProcessError: Command '['mpirun', '-np', '1', '../wordrank', '--beta', '99', '--loss', 'hinge', '--epsilon', '0.75', '--period', '5', '--alpha', '100', '--lrate', '0.001', '--path', 'meta', '--dump_period', '5', '--dump_prefix', 'model', '--dim', '100', '--sgd_num', '100', '--nthread', '8', '--iter', '6', '--reg', '0']' returned non-zero exit status 139

root@18d3d950726b:/gensim/gensim_dependencies/wordrank# mpirun -np 1 ../wordrank --beta 99 --loss hinge --epsilon 0.75 --period 5 --alpha 100 --lrate 0.001 --path meta --dump_period 5 --dump_prefix model --dim 100 --sgd_num 100 --nthread 8 --iter 6 --reg 0
[proxy:0:0@18d3d950726b] HYDU_create_process (utils/launch/launch.c:75): execvp error on file ../wordrank (Permission denied)

Also, please update develop branch in your gensim fork & use updated version as base branch for your PRs in future (because I see old errors too, but this errors already fixed)

[parulsethi]

@menshikh-iv Updated my add_wordrank_in_docker branch with develop. And the Segmentation fault (signal 11) error can be reproduced by first running gensim's test_wordrank_wrapper.py (to create testmodel directory containing metadata, vocab.txt etc.) and then running the

mpirun -np 1 ../wordrank --beta 99 --loss hinge --epsilon 0.75 --period 5 --alpha 100 --lrate 0.001 --path meta --dump_period 5 --dump_prefix model --dim 100 --sgd_num 100 --nthread 8 --iter 6 --reg 0

from directory /gensim/gensim_dependencies/wordrank/testmodel

[menshikh-iv]

Please remove this script from PR. You already download it with the wordrank repo. After it, use awk or sed to replace #export CC=icc CXX=icc to export CC=gcc CXX=g++

[menshikh-iv]

Also, do several things

1. Add parameter for -np to WordRank __init__
2. Remove cd call from wrapper and use absolute path to binary instead of relative ../wordrank
3. Add comment & link in wrapper about ompi+docker problems (for example, this thread)