Training Object Detection with FasterRCNN

Hello,

i am currently trying to train object detection with TLT with FasterRCNN and vgg16 as backbone model. When i run the training it gives me the following error:


Any ideas on that error ?
Thanks in advance

Did you have 3 gpus?

Yes.

How about running with gpus 1 or gpus 2?

More, what is the batch size in your spec? Try smaller bs please.