Cuda runtime error 30, resnet not loading

You can try to increase batch size. Often the model may not be “deep” enough to fully utilize the available GPU. Anyway there are a few threads on this topic on the forums like this one.