TypeError: can't convert cuda:0 device type tensor to numpy. Use Tensor.cpu() to copy the tensor to host memory first.I tried answers but did not work

Question

enter image description here

test_loss = test_loss / len(test_loader) # average over number of mini-batch
            acc = 100. * correct / len(test_loader.dataset)
            logger('\tEpoch {} Test set: Average loss: {:.4f}, Accuracy: {}/{} ({:.0f}%)'.format(
                epoch, test_loss, correct, len(test_loader.dataset), acc))
            accuracy = acc.cpu().data.numpy()
            np.savetxt(out, [[epoch, test_loss, accuracy]], delimiter=",",fmt='%f')
            
            if acc > best_acc:
                new_file = os.path.join(args.logdir, 'best-{}.pth'.format(epoch))
                misc.model_save(model, new_file, old_file=old_file, verbose=True)
                best_acc = acc
                old_file = new_file
            call(["/bin/bash", "./layer_record/trace_command.sh"])