deterministic torch.random.random_split

[mshooter]

I am using

torch.utils.data.random_split()

to create train, val,test split, however, I am testing the data on two different models which were trained pytroch lightning, but it is not splitting deterministically while I am setting seed_everything:42

How can I address this?

[adosar]

What do you mean it does "not splitting deterministically"? This function is part of Pytorch and has nothing to do with Lightning.

From the Pytorch docs:

generator = torch.Generator().manual_seed(42)
random_split(range(10), [3, 7], generator=generator)