Make deep_qa pip-installable

[codecov-io]

Codecov Report

Merging #210 into master will increase coverage by 0.18%. The diff coverage is n/a.

@@            Coverage Diff             @@
##           master     #210      +/-   ##
==========================================
+ Coverage    80.4%   80.59%   +0.18%     
==========================================
  Files         102      102              
  Lines        5283     5283              
==========================================
+ Hits         4248     4258      +10     
+ Misses       1035     1025      -10

Impacted Files	Coverage Δ
deep_qa/layers/backend/batch_dot.py	95.45% <ø> (+1.51%)	:white_check_mark:
deep_qa/tensors/similarity_functions/linear.py	95.83% <ø> (+2.08%)	:white_check_mark:
deep_qa/layers/overlap.py	100% <ø> (+3.33%)	:white_check_mark:
deep_qa/tensors/backend.py	94.11% <ø> (+3.92%)	:white_check_mark:
deep_qa/tensors/similarity_functions/bilinear.py	100% <ø> (+5.26%)	:white_check_mark:
deep_qa/layers/encoders/positional_encoder.py	91.17% <ø> (+5.88%)	:white_check_mark:
deep_qa/common/checks.py	81.48% <ø> (+7.4%)	:white_check_mark:

---

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update 5ad371a...2390bb6. Read the comment docs.

[DeNeutoy]

Ok I think this works now: if you run python setup.py sdist from the top level directory, it should create a compressed package which is pip installable. This is what we have to upload to pypi when we are ready. I tested it by copying the tar file to an aws instance and installing it - I was able to import encoders etc, so at least the basic functionality of the package works.

[nelson-liu]

hmm still looks like the rst is still not rendering properly (https://github.com/DeNeutoy/deep_qa/blob/4b56a4c99aaf9b3ef5fce0afe8eff2aa29c20e8e/README.rst)

[DeNeutoy]

@matt-gardner It would be good if you could look over this (given your name's on it haha). I don't see how we can really test this without just putting it on pypi and seeing if it works - I feel like there will be small things we'll have to figure out that are different from the local checks I have done - I can do this once you are basically happy with the metadata etc.

To test: run python setup.py sdist at the top level directory, which will create a tar file in a /dist directory. Copy that to an ec2 instance/ virtual env or something and see if you can install and use the package using pip (after decompressing it).

One other thing I think might be good is that you can set up python scripts to run as CLI arguments - it would be cool if we could have run_solver.py as a command like deep-qa-run which you could just call with a JSON argument file. I'll try adding that if you think it's a good idea.

If that works for you I can try getting it on Pypi this afternoon.

[matt-gardner]

We definitely want the setup.py - are you closing this so you can open a fresh PR?