Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu! (when

device = torch.device("cuda") if torch.cuda.is_available() else torch.device("cpu")
model.to(device)
model.eval()
for i in df:
    content =  i.get('content')
    if len(content)>=30:
        test = tokenizer(content, padding="max_length", max_length=256, truncation=True,
                          return_tensors="pt")

        test.to(device)
        with torch.no_grad():
            outputs = model(test["input_ids"],
                            token_type_ids=None,
                            attention_mask=test["attention_mask"])
        pred_flat = np.argmax(outputs["logits"].cpu(), axis=1).numpy().squeeze()

gpu测试模型时，需要将模型和数据tokenlier后的张量都放入gpu模式允许，并且结果需要放入cpu输出。