When I train PASCAL VOC → Clipart1K， there seems to be some problems.
File "/detectron2-0.6/detectron2/modeling/proposal_generator/proposal_utils.py", line 103, in find_top_rpn_proposals
raise FloatingPointError(
FloatingPointError: Predicted boxes or scores contain Inf/NaN. Training has diverged.
When I train PASCAL VOC → Clipart1K， there seems to be some problems.
File "/detectron2-0.6/detectron2/modeling/proposal_generator/proposal_utils.py", line 103, in find_top_rpn_proposals raise FloatingPointError( FloatingPointError: Predicted boxes or scores contain Inf/NaN. Training has diverged.
So, could you share the config about voc2clipark?