ClimberXIE
ClimberXIE
> Thanks so much!!! I have solved it now!
> Hi, the env looks ok, can you check that the CPU is utilised when the dataloader starts? Usually it is a problem with `OMP_NUM_THREADS` or other external libraries which...
> The training pipeline looks something like this > > CPU process 1: load batch -> augment batch ---> | CPU process 2: load batch -> augment batch ---> |...
I also want to ask this, as far as I know, the ROI Align gpu kernel used by Mask RCNN requires Torch 0.4 version, but nnDetetcion requires pytorch >1.10+, so...