Handling OOMs nicely when enabling batching in general

From the overiew page of Pipelines, we get some general recommendations on when we should enable batching. It mentions handloing OOMs nicely when using batching. Any concrete example of what it means to handle OOMs nicely? I understand it’s highly dependent on the applications but still was hoping to get a concrete example.

1 Like

Has there been any update on this? The closest thing I can find is this thread on Pytorch.