Closed xcharxlie closed 3 years ago
I fixed this problem in a dumb way. Uninstall the pytorch and torchvision twice and reinstall it. Fortunately, it works for no reason. I'll close the problem but welcome any comment if anyone knows the reason behind.
I'm now trying to train my own model in a conda environment, using ABCNet with Ubuntu 16.04, CUDA 10.2. Got a complaint: "libc10_cuda.so: cannot open shared object file: No such file or directory". For more details, please see the transcripts below.
OMP_NUM_THREADS=1 python tools/train_net.py --config-file configs/BAText/Pretrain/attn_R_50.yaml OUTPUT_DIR text_pretraining/attn_R_50
Traceback (most recent call last): File "tools/train_net.py", line 40, in
from adet.data.dataset_mapper import DatasetMapperWithBasis
File "/home/zzr/AdelaiDet/adet/init.py", line 1, in
from adet import modeling
File "/home/zzr/AdelaiDet/adet/modeling/init.py", line 2, in
from .fcos import FCOS
File "/home/zzr/AdelaiDet/adet/modeling/fcos/init.py", line 1, in
from .fcos import FCOS
File "/home/zzr/AdelaiDet/adet/modeling/fcos/fcos.py", line 10, in
from adet.layers import DFConv2d, NaiveGroupNorm
File "/home/zzr/AdelaiDet/adet/layers/init.py", line 5, in
from .bezier_align import BezierAlign
File "/home/zzr/AdelaiDet/adet/layers/bezier_align.py", line 7, in
from adet import _C
ImportError: libc10_cuda.so: cannot open shared object file: No such file or directory
Thanks for any help in advance.