How to DataParallel when batch is a list of Tensors

Hi there,

I’m using the torchvision fasterrcnn pretrained network, which expects a batch input in the form of a list of 3d tensor images and a corresponding list of annotations for each image.

Then, for faster training, I’d like to use DataParallel. But this does not seem to be supported for list input.

Could anyone please suggest a workaround? Thanks in advance