What is your experiment environment?

[susanbao]

When I tried to train for SECOND. I meet this problem:
Traceback (most recent call last):
File "opencood/tools/train.py", line 138, in
main()
File "opencood/tools/train.py", line 100, in main
ouput_dict = model(batch_data['ego'])
File "/home/sas20048/anaconda3/envs/opencood/lib/python3.7/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "/home/sas20048/code/OpenCOOD/opencood/models/second_intermediate.py", line 45, in forward
batch_dict = self.backbone_3d(batch_dict)
File "/home/sas20048/anaconda3/envs/opencood/lib/python3.7/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "/home/sas20048/code/OpenCOOD/opencood/models/sub_modules/sparse_backbone_3d.py", line 106, in forward
x = self.conv_input(input_sp_tensor)
File "/home/sas20048/anaconda3/envs/opencood/lib/python3.7/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "/home/sas20048/anaconda3/envs/opencood/lib/python3.7/site-packages/spconv/modules.py", line 134, in forward
input = module(input)
File "/home/sas20048/anaconda3/envs/opencood/lib/python3.7/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "/home/sas20048/anaconda3/envs/opencood/lib/python3.7/site-packages/spconv/conv.py", line 181, in forward
use_hash=self.use_hash)
File "/home/sas20048/anaconda3/envs/opencood/lib/python3.7/site-packages/spconv/ops.py", line 95, in get_indice_pairs
int(use_hash))
ValueError: /home/sas20048/code/spconv/src/spconv/spconv_ops.cc 87
unknown device type

I found this issue traveller59/spconv#58 and tried their method. But failed.

So I think maybe my cuda version or pytorch version mismatches spconv1.2.1. So could you tell me your experiment environment information, especially for cuda and pytorch? Thanks.

I used cuda 10.1, pytorch 1.10.0.

[DerrickXuNu]

For sure! I know spconv can have some issues and give you a hard time....
OS: Ubuntu 1804
Pytorch: 1.10.2
CUDA version: Cuda compilation tools, release 11.1, V11.1.105

Can you run pointpillar or voxelnet? Is it just SECOND?

[susanbao]

I just met this problem on SECOND. the other three models work fine.

[DerrickXuNu]

Ok,as this is mainly an issue related to spconv, I will close it.