import torch from torchmetrics.image import ErrorRelativeGlobalDimensionlessSynthesis preds = torch.rand([16, 1, 16, 16], generator=torch.manual_seed(42)) target = preds * 0.75 metric = ErrorRelativeGlobalDimensionlessSynthesis() metric.update(preds, target) fig_, ax_ = metric.plot()