Sebastian Raschka
Sebastian Raschka
I think that works! Do you have a unit test handy? Otherwise, I can help looking into this!
Wow this looks good! Currently traveling and will take one more close look in a calm moment, when I am not out and about. Thanks a lot!
Arg totally missed this and am currently traveling (on mobile). I will make a reminder to check this out as soon as I am back on my computer again!
Just did a little documentation update (sorry for the delay, I was traveling during the last couple of weeks, and there's tons of stuff to catch up on)
Can confirm, just bumped into this as well! To reproduce the bug, I'd add ```python trainer = L.Trainer( callbacks=[StochasticWeightAveraging(swa_lrs=0.01)], auto_lr_find=True, ) results = trainer.tune(model=lightning_model, datamodule=dm) ```
Thanks for offering your help once more @NimaSarajpoor! And please don't worry about the timeline! Here is some more context: I noticed that the performance becomes poor / gets stuck...
Hey @NimaSarajpoor no worries at all and thanks for the note! 😊
Hi @NimaSarajpoor, sorry about the confusion. That's actually intentional (but maybe confusing if people are too familiar with joblib, haha). So, I was thinking since joblib doesn't use `n_jobs=0/False` anyways,...
> If I understand correctly, you want to use the case n_jobs=0 to turn off joblib, which means the program will be run on 1 CPU, which is is similar...
Oh I see! Makes sense then -- I just started using Jupyter Lab (coming from Jupyter Notebook) and somehow had this assumption that multiple Jupyter notebooks could be opened in...