I am getting the following error during training on multiple GPUs.
The weird thing here is that the training runs for quite some time without any problems but after a