Codecov Report

Merging #352 (2d045de) into master (1781472) will decrease coverage by 0.03%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master     #352      +/-   ##
==========================================
- Coverage   93.63%   93.60%   -0.04%     
==========================================
  Files          22       22              
  Lines        5591     5595       +4     
==========================================
+ Hits         5235     5237       +2     
- Misses        356      358       +2     

Thanks for the quick PR! One thing I wonder though is that how to check if the build on Circle CI does run the test with GPU? https://github.com/PythonOT/POT/runs/5389073677?check_suite_focus=true

Because if you do this on the test file POT/test/testgromov.py:

https://github.com/ncassereau-idris/POT/blob/c0b2eda0ccdd6da7e7f5296c9abf4e0e9720ec0e/test/test_gromov.py#L185-L186

if torch.cuda.is_available():
    devices.append(torch.device("cuda"))

This line might not be hit at all during running the test file

No you're right, tests performed here do not have access to a GPU. But I ran it on my machine which is equipped with V100s and the test passes. As always with GPU tests, they should be run locally to detect any issue.

No you're right, tests performed here do not have access to a GPU. But I ran it on my machine which is equipped with a V100 and the test passes. As always with GPU tests, they should be run locally to be detected.

The tests passed on my local run as well, thanks.

Thank you to both of you this was a quickly found and squashed bug, we need to do a release soon