support pytorch2.0

[RangiLyu]

No description provided.

[codecov]

Codecov Report

Merging #500 (4d85d0c) into main (d3fb34f) will increase coverage by 0.79%.
The diff coverage is 73.21%.

@@            Coverage Diff             @@
##             main     #500      +/-   ##
==========================================
+ Coverage   74.84%   75.63%   +0.79%     
==========================================
  Files          71       71              
  Lines        4691     4663      -28     
  Branches      728      726       -2     
==========================================
+ Hits         3511     3527      +16     
+ Misses        983      940      -43     
+ Partials      197      196       -1     

Flag	Coverage Δ
unittests	75.63% <73.21%> (+0.79%)	⬆️

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

Impacted Files	Coverage Δ
nanodet/data/collate.py	100.00% <ø> (+46.51%)	⬆️
nanodet/model/arch/one_stage_detector.py	55.81% <0.00%> (-4.19%)	⬇️
nanodet/trainer/task.py	71.18% <82.00%> (+15.95%)	⬆️

... and 4 files with indirect coverage changes

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[cansik]

@RangiLyu Thanks for the port to pytorch2.0. I just tested it on MacOS and it seems that MPS backend is currently not working. Also, the CPU inference seems to be broken, because the devices argument of the pl.Trainer has to be set to the amount of CPUs that can be used (instead of None).

Do you already know when you are going to merge this into main?