Bugfix: relax the requirements

[denproc]

Closes #143
The problem was related to the upgrade of the environment with torch incompatible with CUDA.

Proposed Changes

• made requirements on pytorch and torchvision less strict
• added warning to the BRISQUE score that backprop is not available using `torch==1.5.0'
• added notes to the docstring and readme

[codecov]

Codecov Report

Merging #154 into master will decrease coverage by 0.05%.
The diff coverage is 66.66%.

@@            Coverage Diff             @@
##           master     #154      +/-   ##
==========================================
- Coverage   94.64%   94.59%   -0.06%     
==========================================
  Files          24       24              
  Lines        1624     1627       +3     
==========================================
+ Hits         1537     1539       +2     
- Misses         87       88       +1     

Flag	Coverage Δ
#unittests	94.59% <66.66%> (-0.06%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
piq/brisque.py	99.04% <66.66%> (-0.96%)	⬇️

[denproc]

Ready to merge after #153

[zakajd]

LGTM 👍

[snk4tr]

I am afraid this fix will not help with the problem of different CUDA versions, which was discussed offline

[sonarqubecloud]

Kudos, SonarCloud Quality Gate passed!

0 Bugs
0 Vulnerabilities (and 0 Security Hotspots to review)
0 Code Smells

No Coverage information
0.0% Duplication

[denproc]

Let me look into it a bit more after @snk4tr comment

[denproc]

@snk4tr @zakajd the proposed approach works, ready for review.

As I suggested, requirement torch>=1.2 covers the tags used to specify the cuda driver version.

Additional context just in case:

[In]: pip list
[Out]: torch                  1.5.0+cu101 
[Out]: torchvision            0.6.0+cu101 

[In]: python setup.py install
[In]: pip list
[Out]: piq                    0.5.1    
[Out]: torch                  1.5.0+cu101 
[Out]: torchvision            0.6.0+cu101 

[snk4tr]

Great!
Merging 👍