add package info

[Borda]

reaction to #358 and re-opening #359

It would be fine to make a homepage for example from the main README file, as Github offers
see: https://guides.github.com/features/pages/
it is about %min and couple clicks... :)

[Borda]

@williamFalcon could you please add WIP app which would block merging unfinished PR... Thx
see: https://github.com/marketplace/wip

[Borda]

the package is still deeply linked on test-tube see:

from test_tube import HyperOptArgumentParser
from test_tube import Experiment

[williamFalcon]

awesome. could u rebase master for tests?

did you try pip install -e .?

[Borda]

it still does not work :( I check the scikit-learn and scikit-image and they are not using buildins package which somehow is part of future so it is a but strange...

[Borda]

@williamFalcon any idea why the tests are failing?

'gpus_param_tests' not found in `markers` configuration option

[williamFalcon]

@neggert @Ir1d @ jeffling @shubham anyone want to PR review this?

[Ir1d]

Hi, I ran pip install git+https://github.com/Borda/pytorch-lightning@pkg-info.
With pytorch 1.3 installed in a fresh conda env, I got No module named 'test_tube' when importing with import pytorch_lightning as pl.
Have you met this before?

[Borda]

@Ir1d it is a consequence of one of the past decision (#36) to cut examples from the package and keep the torch_ligthning independent on workflows like mlflow/test-tube/...
but examples use test-tube and we still somehow keep pl_examples to be installed

[Borda]

@Ir1d anyway missing test_tube is not related to this PR since it was not touched in requirements... :)

[Borda]

@williamFalcon \ping

[neggert]

Looks good. Any thoughts on using the extras functionality in setuptools to handle optional testtube/mlflow/cometml dependencies?