Fix .llm import trigger of .distributed depr warning

[puririshi98]

No description provided.

[rusty1s]

I don't think that will prevent the warning since you still depend on torch_geometric.distributed.

[puririshi98]

I don't think that will prevent the warning since you still depend on torch_geometric.distributed.

im confused, where do you see the warning on your end. I tested importing pyg and dont see any warnings

Python 3.12.3 (main, Aug 14 2025, 17:47:21) [GCC 13.3.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import torch_geometric
>>> 

[rusty1s]

you don't see them since DeprecationWarnings are not shown by default. It is visible in pytest though.

[puririshi98]

you don't see them since DeprecationWarnings are not shown by default. It is visible in pytest though.

is there a quick way to turn that on for interactive debugging?

[rusty1s]

just call the warning as warn("foo"), without the stacklevel and category.

[puririshi98]

just call the warning as warn("foo"), without the stacklevel and category.

okay since we still use local graph/feature store i put the warnings where most would see them for initializing loaders because if your doing that you would probably benefit from using cugraph instead. hopefully this works for you. lmk if you recommend another way

[rusty1s]

Looks good. Do we need the warning in all three places?