Various fix and new stuff

[nicolas-chaulet]

Various small things:

• Support for class weights in the losses of votenet
• Refactor the class weight computation code + tests
• Fix tiny bug in trainer

[codecov]

Codecov Report

Merging #430 into master will increase coverage by 0.15%.
The diff coverage is 40.19%.

@@            Coverage Diff             @@
##           master     #430      +/-   ##
==========================================
+ Coverage   64.09%   64.25%   +0.15%     
==========================================
  Files         168      168              
  Lines       13155    13161       +6     
  Branches     1785     1781       -4     
==========================================
+ Hits         8432     8456      +24     
+ Misses       4249     4227      -22     
- Partials      474      478       +4     

Flag	Coverage Δ
#unittests	64.25% <40.19%> (+0.15%)	⬆️

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

Impacted Files	Coverage Δ
...rch_points3d/core/data_transform/grid_transform.py	60.16% <0.00%> (+4.70%)	⬆️
torch_points3d/datasets/panoptic/s3dis.py	0.00% <0.00%> (ø)
torch_points3d/trainer.py	71.89% <0.00%> (ø)
torch_points3d/datasets/segmentation/s3dis.py	22.51% <3.44%> (+0.84%)	⬆️
torch_points3d/datasets/segmentation/scannet.py	38.20% <50.00%> (+0.19%)	⬆️
torch_points3d/models/object_detection/votenet.py	78.57% <50.00%> (-1.68%)	⬇️
torch_points3d/models/object_detection/votenet2.py	73.68% <50.00%> (-0.94%)	⬇️
torch_points3d/metrics/confusion_matrix.py	65.90% <66.66%> (-0.76%)	⬇️
torch_points3d/core/data_transform/features.py	83.22% <71.42%> (-0.11%)	⬇️
torch_points3d/datasets/base_dataset.py	72.66% <88.88%> (+1.03%)	⬆️
... and 2 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 dfa4e26...6f4549e. Read the comment docs.