Igor Kotenkov
Igor Kotenkov
link folder with map into /map/ dir
You should change key to load model. Print checkpoint.keys() first to see options, and choice right one. BTW after that you got load weights problem, caused my DataParallel apply -...
https://github.com/skyhehe123/SA-SSD/issues/19#issuecomment-619886333 @mengxingshifen1218
What's pytorch version you use?
https://github.com/skyhehe123/SA-SSD/blob/e8d4db87fe7651dff3b3693c53b8fc2d4afb17cc/mmdet/core/anchor/anchor3d_generator.py#L3
It's all warnings, right? I got same but training and inference still normal
How much classes you train at same time? I occur same error in different framework (but still mmdet) when pass wrong class indexes (e.g index 15 when using only 14...
> @dongqiaqia The loss looks exactly different than mines. Let me check it first. Maybe a docker env. is needed. hello. I'm starting to work with your repo, and Docker...
sorry, yes, i'm paste error from second dockerfile, which build with 3.6. In first i see same error except py version. ` ImportError: /workspaces/SASSD/mmdet/ops/points_op/points_op_cpu.cpython-37m-x86_64-linux-gnu.so: undefined symbol: _ZN6caffe26detail36_typeMetaDataInstance_preallocated_7E`
by the way there's a funny moment: error gone when i change 2-3 lines order here: https://github.com/skyhehe123/SA-SSD/blob/master/mmdet/ops/points_op/__init__.py#L2 i.e. load torch first, then extensions. This fixes both docker images, lol. Any...