Exception has occurred: TypeError only integer tensors of a single element can be converted to an index

[chronbird]

Hi @chrischoy ,

Thanks for the great work! When I am trying training semantic segmentation on ScanNet, I encounter the following exception:

Traceback (most recent call last): File "/home/chronbird/anaconda3/envs/pointcloud/lib/python3.8/runpy.py", line 194, in _run_module_as_main return _run_code(code, main_globals, None, File "/home/chronbird/anaconda3/envs/pointcloud/lib/python3.8/runpy.py", line 87, in _run_code exec(code, run_globals) File "/home/chronbird/SpatioTemporalSegmentation/main.py", line 156, in <module> main() File "/home/chronbird/SpatioTemporalSegmentation/main.py", line 149, in main train(model, train_data_loader, val_data_loader, config) File "/home/chronbird/SpatioTemporalSegmentation/lib/train.py", line 91, in train soutput = model(*inputs) File "/home/chronbird/anaconda3/envs/pointcloud/lib/python3.8/site-packages/torch/nn/modules/module.py", line 550, in __call__ result = self.forward(*input, **kwargs) File "/home/chronbird/SpatioTemporalSegmentation/models/res16unet.py", line 204, in forward out_b1p2 = self.block1(out) File "/home/chronbird/anaconda3/envs/pointcloud/lib/python3.8/site-packages/torch/nn/modules/module.py", line 550, in __call__ result = self.forward(*input, **kwargs) File "/home/chronbird/anaconda3/envs/pointcloud/lib/python3.8/site-packages/torch/nn/modules/container.py", line 100, in forward input = module(input) File "/home/chronbird/anaconda3/envs/pointcloud/lib/python3.8/site-packages/torch/nn/modules/module.py", line 550, in __call__ result = self.forward(*input, **kwargs) File "/home/chronbird/SpatioTemporalSegmentation/models/modules/resnet_block.py", line 42, in forward out = self.conv1(x) File "/home/chronbird/anaconda3/envs/pointcloud/lib/python3.8/site-packages/torch/nn/modules/module.py", line 550, in __call__ result = self.forward(*input, **kwargs) File "/home/chronbird/anaconda3/envs/pointcloud/lib/python3.8/site-packages/MinkowskiEngine/MinkowskiConvolution.py", line 263, in forward self.kernel_generator.get_kernel(input.tensor_stride, self.is_transpose) File "/home/chronbird/anaconda3/envs/pointcloud/lib/python3.8/site-packages/MinkowskiEngine/Common.py", line 347, in get_kernel self.cache[tuple(tensor_stride)] = convert_region_type( File "/home/chronbird/anaconda3/envs/pointcloud/lib/python3.8/site-packages/MinkowskiEngine/Common.py", line 272, in convert_region_type region_offset = torch.IntTensor(region_offset) TypeError: only integer tensors of a single element can be converted to an index

Then I check about the input of self.block1, it is
SparseTensor( Coords=tensor([[ 1, 74, 8, 150], [ 1, 70, 0, 192], [ 8, -190, 350, 82], ..., [ 1, 240, 6, 192], [ 5, 380, -82, 136], [ 6, 70, -68, 110]], dtype=torch.int32) Feats=tensor([[2.4673, 1.2257, 2.8942, ..., 0.0000, 2.3988, 0.0000], [0.5576, 0.0000, 0.0000, ..., 0.9546, 0.0000, 0.1357], [0.2138, 0.1608, 0.0000, ..., 0.0000, 0.3920, 0.0000], ..., [0.0000, 0.0000, 0.0000, ..., 0.0000, 0.0000, 0.4807], [0.0000, 0.0366, 0.0000, ..., 0.5666, 0.0000, 1.0899], [0.0000, 1.6132, 0.0000, ..., 0.1976, 0.0000, 1.4563]], device='cuda:0', grad_fn=<ReluBackward1>) coords_key=< CoordsKey, key: 16908437251554604741, tensor_stride: [2, 2, 2, ��] in dimension: 3 > tensor_stride=[2, 2, 2] coords_man=< CoordsManager Number of Coordinate Maps: 2 Coordinate Map Key: 16908437251554604741, Size: 323293 Coordinate Map Key: 15034981587763204738, Size: 821840 Number of Kernel Maps: 2 Kernel In-Out Map Key: 1762388836632698508, Size: 821840 Kernel In-Out Map Key: 13453830109827794797, Size: 7072854 > spatial dimension=3)

and self.block1
Sequential( (0): BasicBlock( (conv1): MinkowskiConvolution(in=32, out=32, region_type=RegionType.HYBRID, kernel_volume=27, stride=[1, 1, 1], dilation=[1, 1, 1]) (norm1): MinkowskiBatchNorm(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) (conv2): MinkowskiConvolution(in=32, out=32, region_type=RegionType.HYBRID, kernel_volume=27, stride=[1, 1, 1], dilation=[1, 1, 1]) (norm2): MinkowskiBatchNorm(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) (relu): MinkowskiReLU() ) (1): BasicBlock( (conv1): MinkowskiConvolution(in=32, out=32, region_type=RegionType.HYBRID, kernel_volume=27, stride=[1, 1, 1], dilation=[1, 1, 1]) (norm1): MinkowskiBatchNorm(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) (conv2): MinkowskiConvolution(in=32, out=32, region_type=RegionType.HYBRID, kernel_volume=27, stride=[1, 1, 1], dilation=[1, 1, 1]) (norm2): MinkowskiBatchNorm(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True) (relu): MinkowskiReLU() ) )

I am using Pytorch 1.5.1 and the latest MinkowskiEngine built from source. Could you please help me check what might possibly cause the error? Thanks in advance.

[chronbird]

found solution at /home/chronbird/anaconda3/envs/pointcloud/lib/python3.8/site-packages/MinkowskiEngine/Common.py, where each
(tensor_stride[d] / up_stride[d])
can be modified as
(tensor_stride[d] / up_stride[d]).int()
then the exception will disappear.

[chrischoy]

Thanks for reporting the issue. I'll fix this on the next version.