Closed TianhaoFu closed 2 years ago
hi, compiling dcn is optional (we don't use dcn in the most recent version). And it seems to only work with old torch version (e.g. 1.1).
I updated the install guide, you only need to compile the nms
https://github.com/tianweiy/CenterPoint/blob/master/docs/INSTALL.md#cuda-extensions
hi, compiling dcn is optional (we don't use dcn in the most recent version). And it seems to only work with old torch version (e.g. 1.1).
I updated the install guide, you only need to compile the nms
https://github.com/tianweiy/CenterPoint/blob/master/docs/INSTALL.md#cuda-extensions
ok, thanks!
Hi, thx for sharing your works. If I skip installing DeformConv, the above error ocurrs:
NameError: name 'DeformConv' is not defined
what should I do to skip installing DeformConv?
at which line of code and which config are you using?
Thanks for your prompt reply when I run commands:
cd det3d/ops/dcn python setup.py build_ext --inplace
It returns:
ninja: build stopped: subcommand failed. Traceback (most recent call last): File "/root/miniconda3/envs/centerpoint/lib/python3.8/site-packages/torch/utils/cpp_extension.py", line 1717, in _run_ninja_build subprocess.run( File "/root/miniconda3/envs/centerpoint/lib/python3.8/subprocess.py", line 516, in run raise CalledProcessError(retcode, process.args, subprocess.CalledProcessError: Command '['ninja', '-v']' returned non-zero exit status 1.
The above exception was the direct cause of the following exception:
Traceback (most recent call last):
File "setup.py", line 4, in
When I run the following commands according to the https://github.com/tianweiy/CenterPoint/blob/master/docs/GETTING_START.md:
python tools/dist_test.py configs/nusc/voxelnet/nusc_centerpoint_voxelnet_0075voxel_dcn_flip.py --work_dir work_dirs/nusc_centerpoint_voxelnet_dcn_0075voxel_flip_testset --checkpoint work_dirs/nusc_0075_dcn_flip_track/voxelnet_converted.pth --testset --speed_test
It returns:
Use Deformable Convolution in the CenterHead!
Traceback (most recent call last):
File "tools/dist_test.py", line 211, in
hi, please see my reply here https://github.com/tianweiy/CenterPoint/issues/362#issuecomment-1284026185
basically, dcn doesn't work with latest torch version so I suggest you to switch to non-dcn configs (which actually gets better performance). If you just need the detection file, you can refer to https://github.com/tianweiy/CenterPoint/issues/249
otherwise, just change the config name and checkpoint path to https://github.com/tianweiy/CenterPoint/tree/master/configs/nusc#voxelnet
Thanks
Good suggestions! Now I can run the eval code. Thx.
Hi, Thanks for your code. When I was building environment following INSTALL.md[Cuda Extensions], I came across such error:
I don't know how to deal with it. Could you give me some ideas? Thanks!