fix documentation to match pytorch-lightning

[gdevos010]

the gpus arg in PyTorch Lightning has been deprecated in favor of devices

fixed broken PyTorch Lightning link

https://pytorch-lightning.readthedocs.io/en/stable/common/trainer.html#gpus

[codecov-commenter]

Codecov Report

Base: 94.02% // Head: 93.97% // Decreases project coverage by -0.04% ⚠️

Coverage data is based on head (46ff8c6) compared to base (d97932a).
Patch has no changes to coverable lines.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1244      +/-   ##
==========================================
- Coverage   94.02%   93.97%   -0.05%     
==========================================
  Files          73       73              
  Lines        8215     8182      -33     
==========================================
- Hits         7724     7689      -35     
- Misses        491      493       +2     

Impacted Files	Coverage Δ
darts/models/forecasting/block_rnn_model.py	98.24% <ø> (-0.04%)	⬇️
darts/models/forecasting/nbeats.py	98.11% <ø> (ø)
darts/models/forecasting/nhits.py	99.27% <ø> (-0.01%)	⬇️
darts/models/forecasting/rnn_model.py	97.64% <ø> (ø)
darts/models/forecasting/tcn_model.py	96.96% <ø> (ø)
darts/models/forecasting/tft_model.py	97.53% <ø> (ø)
...arts/models/forecasting/torch_forecasting_model.py	87.08% <ø> (-0.40%)	⬇️
darts/models/forecasting/transformer_model.py	100.00% <ø> (ø)
darts/logging.py	96.29% <0.00%> (-1.86%)	⬇️
darts/utils/statistics.py	88.65% <0.00%> (-1.04%)	⬇️
... and 4 more

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

☔ View full report at Codecov.
📢 Do you have feedback about the report comment? Let us know in this issue.

[dennisbader]

Thanks again for another PR :)
We would have to adapt the parts of the code where we generate params dicts for PyTorch Lightning Trainer with "gpus". Otherwise the deprecation warning will not disappear.

I added a suggestion.

[hrzn]

LGTM, however we could consider moving these comments to actual docstring, and remove the support for torch_device_str, WDYT?

[dennisbader]

LGTM, however we could consider moving these comments to actual docstring, and remove the support for torch_device_str, WDYT?

Agreed 👍

[hrzn]

LGTM, thanks @gdevos010 !
Just have a small question about the pytorch lightning version (>=1.7.0) check in torch_forecasting_model.py: is this needed?

[dennisbader]

Nice :) Only thing is that we should set the 3 parameters by default "accelerator", "gpus" (or "devices"), "auto_select_gpus", as we decided to always use CPU by default.

After that we can merge :)

[gdevos010]

@dennisbader we should be good to merge now

[dennisbader]

Yes, looks perferct now 👍 Thanks a lot for all the effort @gdevos010!