AttributeError: 'int' object has no attribute 'type'
Glad to able to ask you some questions. I have only two gpus and use "NUM_GPUS=2" "CONFIG=configs/maskrcnn_overlock/mask_rcnn_overlock_t_in1k_fpn_1x_coco.py" "bash scripts/dist_train.sh $CONFIG $NUM_GPUS" to do the detection, then it has the error: "AttributeError: 'int' object has no attribute 'type'", what should I do next to train?
Hi, could you please provide more information since I'm not seeing a similar error on my end? Additionally, please ensure that you have installed the packages with the recommended version.
I've installed the paceages with the recommended and solved the problem. This is because the interface of torch.device is incompatible with MMCV, but I have modified the code and it now works.
from .active_rotated_filter import active_rotated_filter
File "/home/yangjie1/anaconda3/envs/overlock/lib/python3.10/site-packages/mmcv/ops/active_rotated_filter.py", line 10, in
@zhangshuo0723 Hi, 看起来可能是mmcv编译的问题,卸载然后重新安装试试看
pip uninstall mmcv mmcv-full; pip install mmcv-full==1.7.2 --no-cache-dir
Closed as no activities remain.