Fix pylint

[nikita-klsh]

This PR fixes pylint output in the latest docker image analysiscenter1/ds-py3.

A lot of missing-function-docstring and abstract-method warnings happened in batchflow/models/torch folder we cant fix, because they occur on the torch side. They are fixing this and this.

So we can:

1. Close eyes on this warnings.
2. Temporary set additional restrictions for pylint for torch folder only like its currently done.

@roman-kh

[codecov]

Codecov Report

Merging #522 into master will decrease coverage by 0.08%.
The diff coverage is 25.35%.

@@            Coverage Diff             @@
##           master     #522      +/-   ##
==========================================
- Coverage   41.46%   41.38%   -0.09%     
==========================================
  Files         147      148       +1     
  Lines       14587    14843     +256     
==========================================
+ Hits         6049     6143      +94     
- Misses       8538     8700     +162     

Impacted Files	Coverage Δ
batchflow/batch.py	55.53% <0.00%> (ø)
batchflow/dataset.py	88.54% <ø> (ø)
batchflow/decorators.py	27.73% <0.00%> (ø)
batchflow/dsindex.py	87.55% <ø> (ø)
batchflow/models/tf/fcn.py	27.45% <0.00%> (ø)
batchflow/models/tf/inception_base.py	15.51% <0.00%> (ø)
batchflow/models/tf/inception_resnet_v2.py	18.01% <0.00%> (ø)
batchflow/models/tf/inception_v1.py	35.71% <0.00%> (ø)
batchflow/models/tf/inception_v3.py	22.22% <0.00%> (ø)
batchflow/models/tf/inception_v4.py	19.27% <0.00%> (ø)
... and 38 more

---

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 e8957da...da0a17a. Read the comment docs.

[roman-kh]

@nikita-klsh It seems pytorch issues are solved already.

[nikita-klsh]

@nikita-klsh It seems pytorch issues are solved already.

@roman-kh There are no releases with updates yet. Or we want to build it from github?

[nikita-klsh]

Running pylint with torch 1.7.0 still raises warnings about forward methods missing docs.
Torch solved the issue and now the method formally obtains the doc, i.e. nn.Module.forward.__doc__ is not None, even tho docs are not explicitly speсified in the method's body.
I suppose this might be pylint false warning.

[roman-kh]

Why do you think this is a better approach?

[nikita-klsh]

Looks weird to reinitialize attributes, that have already been initialized in the parent class Baseset

[nikita-klsh]

I return it as it was initially and plug pylint: disable=attribute-defined-outside-init stub.
In the code we assign values to self.train / self.test / self.validation attributes that WERE defined in the init.
It looks like pylint false positive