You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Currently, GSplat supports single-scene, multi-camera rendering in parallel. However, in some cases, we need to render multiple scenes with multiple cameras in parallel, such as for deformable scenes. A simple approach is to use a for loop, but a better option is to use torch.vmap, which enables parallelism for unbatched operations.
However, to use vmap with a custom torch.autograd.Function, we need to make a small adjustment. As specified here, we simply need to override an additional static method, setup_context, in the custom Function class.
If you're interested, I can help create a PR.
The text was updated successfully, but these errors were encountered:
Currently, GSplat supports single-scene, multi-camera rendering in parallel. However, in some cases, we need to render multiple scenes with multiple cameras in parallel, such as for deformable scenes. A simple approach is to use a for loop, but a better option is to use torch.vmap, which enables parallelism for unbatched operations.
However, to use vmap with a custom torch.autograd.Function, we need to make a small adjustment. As specified here, we simply need to override an additional static method, setup_context, in the custom Function class.
If you're interested, I can help create a PR.
The text was updated successfully, but these errors were encountered: