Tensorboard Images infer rows/num_images based on dimensions

If you pass a 4D tensor (rows, columns, height, width) to TensorBoardImages then it would be good if it could detect this and infer the make_grid parameters: num_images, nrow.

Would mean you don't have to set these things both where you create the display tensor and in tensorboard. 