zhenyuw16/UniDetector

environment error

Opened this issue · 5 comments

Hello, thanks for the great work!
But when I create an environment following https://github.com/open-mmlab/mmdetection/blob/v2.18.0/docs/get_started.md for training, the import error has came.
1
OS: Linux
CUDA: 11.7
pytorch: 1.13

Perhaps this is because the error of mmcv. You should follow the instructions of mmdetection to install the correct version of mmcv. The version of mmcv should also match your pytorch version.

Kegard commented

I hava meet the same problem too, and I have tride changed the version but failed, because it will raise other error, can you tell me the version of mmcv, mmcv-full, and mmdet?

Kegard commented

I have solved the problem. the environment as follow:
mmcv-full 1.3.14
mmdet 2.18.0
numpy 1.23.5
Don't install the mmcv >=2.0.0 and the numpy <1.24.0 is better.

I have solved the problem. the environment as follow: mmcv-full 1.3.14 mmdet 2.18.0 numpy 1.23.5 Don't install the mmcv >=2.0.0 and the numpy <1.24.0 is better.
May I ask if you encountered 'FasterRCNN: 'CLIPResNet is not in the models registry' while running 'bash tools/dist_train.sh configs/singledataset/clip_end2end_faster_rcnn_r50_c4_1x_coco.py 8 --cfg-options load_from=regionclip_pretrained-cc_rn50_mmdet.pth' in this environment?"

Kegard commented

you can pip unistall mmdet and try python setup.py develop