Model checkpoint size mismatch for backbone #2
Description
Thank you for your great work!
I downloaded aligned_sstnet_scannet.pth for training on ScanNet but I ran into the following error:
[rank0]: Traceback (most recent call last):
[rank0]: File "/root/autodl-tmp/code/SegDINO3D/train_3d.py", line 178, in <module>
[rank0]: load_model(
[rank0]: File "/root/autodl-tmp/code/SegDINO3D/utils/ckpt_utils.py", line 107, in load_model
[rank0]: loadinfo = model_without_ddp.load_state_dict(clean_state_dict(checkpoint['model']), strict=False)
[rank0]: File "/root/autodl-tmp/conda/envs/dino3d/lib/python3.9/site-packages/torch/nn/modules/module.py", line 2624, in load_state_dict
[rank0]: raise RuntimeError(
[rank0]: RuntimeError: Error(s) in loading state_dict for Baseline3D:
[rank0]: size mismatch for backbone.blocks.block0.conv_branch.2.weight: copying a param with shape ('backbone.blocks.block0.conv_branch.2.weight', torch.Size([32, 32, 3, 3, 3])) from checkpoint,the shape in current model is torch.Size([32, 3, 3, 3, 32]).
[rank0]: size mismatch for backbone.blocks.block0.conv_branch.5.weight: copying a param with shape ('backbone.blocks.block0.conv_branch.5.weight', torch.Size([32, 32, 3, 3, 3])) from checkpoint,the shape in current model is torch.Size([32, 3, 3, 3, 32]).
[rank0]: size mismatch for backbone.blocks.block1.conv_branch.2.weight: copying a param with shape ('backbone.blocks.block1.conv_branch.2.weight', torch.Size([32, 32, 3, 3, 3])) from checkpoint,the shape in current model is torch.Size([32, 3, 3, 3, 32]).
[rank0]: size mismatch for backbone.blocks.block1.conv_branch.5.weight: copying a param with shape ('backbone.blocks.block1.conv_branch.5.weight', torch.Size([32, 32, 3, 3, 3])) from checkpoint,the shape in current model is torch.Size([32, 3, 3, 3, 32]).
[rank0]: size mismatch for backbone.conv.2.weight: copying a param with shape ('backbone.conv.2.weight', torch.Size([32, 64, 2, 2, 2])) from checkpoint,the shape in current model is torch.Size([64, 2, 2, 2, 32]).
[rank0]: size mismatch for backbone.u.blocks.block0.conv_branch.2.weight: copying a param with shape ('backbone.u.blocks.block0.conv_branch.2.weight', torch.Size([64, 64, 3, 3, 3])) from checkpoint,the shape in current model is torch.Size([64, 3, 3, 3, 64]).
[rank0]: size mismatch for backbone.u.blocks.block0.conv_branch.5.weight: copying a param with shape ('backbone.u.blocks.block0.conv_branch.5.weight', torch.Size([64, 64, 3, 3, 3])) from checkpoint,the shape in current model is torch.Size([64, 3, 3, 3, 64]).
[rank0]: size mismatch for backbone.u.blocks.block1.conv_branch.2.weight: copying a param with shape ('backbone.u.blocks.block1.conv_branch.2.weight', torch.Size([64, 64, 3, 3, 3])) from checkpoint,the shape in current model is torch.Size([64, 3, 3, 3, 64]).
[rank0]: size mismatch for backbone.u.blocks.block1.conv_branch.5.weight: copying a param with shape ('backbone.u.blocks.block1.conv_branch.5.weight', torch.Size([64, 64, 3, 3, 3])) from checkpoint,the shape in current model is torch.Size([64, 3, 3, 3, 64]).
[rank0]: size mismatch for backbone.u.conv.2.weight: copying a param with shape ('backbone.u.conv.2.weight', torch.Size([64, 96, 2, 2, 2])) from checkpoint,the shape in current model is torch.Size([96, 2, 2, 2, 64]).
...
Could you please fix it? Thank you for your attention and assistance.