I am encountering a size mismatch of both PN2 checkpoint and VI-Net checkpoint when trying to load the checkpoints of RGB-D-P in testing. The specific error message of VI-Net checkpoint is:
RuntimeError: The model and loaded state dict do not match exactly
size mismatch for spherical_fpn.encoder2.conv1.conv.weight: copying a param with shape torch.Size([64, 5, 3, 3]) from checkpoint, the shape in current model is torch.Size([64, 3, 3, 3]).
But when I try to test with the checkpoint from RGB-D there is no such issue. I guess this is because test_housecat.py is for RGB-D checkpoints not for RGB-D-P. Are you using separate testing script when you testing your RGB-D-P checkpoints?
Any help or suggestions would be greatly appreciated!
Hi,
I am encountering a size mismatch of both PN2 checkpoint and VI-Net checkpoint when trying to load the checkpoints of RGB-D-P in testing. The specific error message of VI-Net checkpoint is:
RuntimeError: The model and loaded state dict do not match exactly
size mismatch for spherical_fpn.encoder2.conv1.conv.weight: copying a param with shape torch.Size([64, 5, 3, 3]) from checkpoint, the shape in current model is torch.Size([64, 3, 3, 3]).
But when I try to test with the checkpoint from RGB-D there is no such issue. I guess this is because test_housecat.py is for RGB-D checkpoints not for RGB-D-P. Are you using separate testing script when you testing your RGB-D-P checkpoints?
Any help or suggestions would be greatly appreciated!
Thank you!