I am trying to use a batched pointcloud with R3PointConv

[Satvik1701]

I have a batched point cloud input to create an equivariant point cloud encoder. However, I think the GeometricTensor does not accept a batch dimension if coords is not None. I have a desired input of x = [batch_size, #points, channels] e.g. [256, 4096, 6] for each of the 4096 points having a XYZRGB. I don't think there is a way to currently this type of input because of the batch_size as I was going to split the channels so for group = gspaces.rot2dOnR3(8) the input_geom = nn.GeometricTensor(x, nn.FieldType(group, 3 * [group.trivial_repr]), coords=x[:, :, :3]) then x would be x[:, :, 3:] for the RGB channels. I was wondering if there was a way to do this, any help would be much appreciated. Thanks!