AttributeError: 'Trainer' object has no attribute 'running_sanity_check"

[annalauralerede]

Hi,

I am trying to optimise the hyperparameters of my network using raytune. My implementation is pretty much based on this:
https://docs.ray.io/en/master/tune/tutorials/tune-pytorch-lightning.html#selecting-a-scheduler

When I train my network using pre-set hyperparameters, it works smoothly. The problems come from the callback, so when I add the following line:
TuneReportCallback({"loss":"val_loss"}, on="validation_end")

I get the following error:

Anyone knows how to solve this??
I don't think the problem is with my code as I haven't done anything different compared to the tutorial!

My code can be found here:
https://github.com/annalauralerede/anomaly-detection/blob/main/lstmae_pl_opt.py

[pvmilk]

I think it dues to this : ray-project/ray#20741

As of ray[tune]==1.10.0, either downgrade your pytorch-lightning to 1.4.
Or upgrade your raytune to be compatible with pytorch-lightning 1.5+ ( the fix has been merged in this commit ray-project/ray#20562).
$ sudo pip install ray==1.11.0rc0