the PillarFeatureNet(nn.Module) in export_onnx.py is different from item in pillar_encoder.py, which results an error when export onnx files.

[HULoaf]

In export.onnx.py, it's this:
( if with_voxel_center:
in_channels += 3)
but in pillar_encoder.py:
(if with_voxel_center:
in_channels += 2)

when i export onnx files, the error is :
size mismatch for pfn_layers.0.linear.weight: copying a param with shape torch.Size([64, 9]) from checkpoint, the shape in current model is torch.Size([64, 10]).

so ,should I change in_channels += 3 to += 2 in ecport_onnx.py?
I tried it , but when i inference in C++ ,the results were terrible, should i change the preprocess.cu?