Removed sageconv dropout due to not existing anymore(?)

lucidrains / meshgpt-pytorch

Implementation of MeshGPT, SOTA Mesh generation using Attention, in Pytorch

MIT License

700 stars 57 forks source link

Removed sageconv dropout due to not existing anymore(?) #69

Closed MarcusLoppe closed 5 months ago

MarcusLoppe commented 5 months ago

Seems like the torch_geometric dropped or change the args for SAGEConv so sageconv_dropout is no longer a valid arg.

@lucidrains Please approve this pull request since you are not able to create the autoencoder without removing this arg. I've seen no improvements in accuracy when using the dropout so it probably isn't needed or very useful. Thank you 😄

lucidrains commented 5 months ago

oh thanks Marcus!

edit: want to bump the patch version while you are at it?