preds = torch.tensor(preds).to('cuda') if not isinstance(preds,torch.Tensor) else preds
targets = torch.tensor(targets).to('cuda') if not isinstance(targets, torch.Tensor) else targets
probs = torch.tensor(probs).to('cuda') if not isinstance(probs, torch.Tensor) else probs
accuracy = torchmetrics.Accuracy(num_classes=2, average='micro').to('cuda')
lesion_acc = accuracy(preds, targets)
cm = confusion_matrix(targets.squeeze().cpu().numpy(), preds.squeeze().cpu().numpy())
在计算acc的时候,pred和target需要是torch.Tensor类型,且在cuda上,在计算混淆矩阵的时候,则需要转成numpy,且在cpu上