dvlab-research / Stratified-Transformer

Stratified Transformer for 3D Point Cloud Segmentation (CVPR 2022)
MIT License
362 stars 40 forks source link

Does anyone come across the problems of gcc version #10

Closed yuchenlichuck closed 2 years ago

yuchenlichuck commented 2 years ago

_OSError: /lib64/libm.so.6: version `GLIBC_2.27' not found (required by /home/liy0r/anaconda3/envs/stformer/lib/python3.7/site-packages/torch_spline_conv/_basiscuda.so)

X-Lai commented 2 years ago

May I ask the version of gcc that you use?

Eddieeee-Liu commented 2 years ago

@yuchenlichuck please change your gcc version to 7.5.0

yuchenlichuck commented 2 years ago

gcc 8.2.0 gcc 10.2.0 gcc 11.1.0, I try this three, all don't works

yuchenlichuck commented 2 years ago

I can run the code somehow, but I also come across another issue

(stformer) [liy0r@gpu202-09-l Stratified-Transformer]$ module load gcc/8.2.0
GNU 8.2.0 is now loaded
(stformer) [liy0r@gpu202-09-l Stratified-Transformer]$ python3 train.py --config config/s3dis/s3dis_stratified_transformer.yaml
terminate called after throwing an instance of 'std::bad_alloc'
  what():  std::bad_alloc
Aborted
(stformer) [liy0r@gpu202-09-l Stratified-Transformer]$ module load gcc/10.2.0
GNU 10.2.0 is now loaded
(stformer) [liy0r@gpu202-09-l Stratified-Transformer]$ python3 train.py --config config/s3dis/s3dis_stratified_transformer.yaml
Segmentation fault
(stformer) [liy0r@gpu202-09-l Stratified-Transformer]$ module load gcc/11.1.0
GNU 11.1.0 is now loaded
(stformer) [liy0r@gpu202-09-l Stratified-Transformer]$ python3 train.py --config config/s3dis/s3dis_stratified_transformer.yaml
terminate called after throwing an instance of 'std::bad_alloc'
  what():  std::bad_alloc
Aborted
yuchenlichuck commented 2 years ago

I get that my torch-geometric install is incorrect, but I was followed by the instructions, maybe it because the pytorch version or gcc or other libs is not matching

X-Lai commented 2 years ago

FYI, I have pushed the requirements.txt.

X-Lai commented 2 years ago

gcc==7.5.0 can work.

yuchenlichuck commented 2 years ago

and please remind to commend or delete sharedarray

yuchenlichuck commented 2 years ago

pip install h5py