关于根据训练好的模型来预测的问题

[zhengguanyu]

您好，向您请教一个问题

[zhengguanyu]

将status设置成test之后，test_file文件的格式有规定吗？
我将evaluate函数里涉及到的gold_result都删掉了，在带有真实标签的测试集上是可行的。格式如下：

测 O
测 B-XXX
测 E-XXX

相当于对原有的evaluate函数微调一下，也可以预测。
但是我单独做的测试集就是空的。
test_file设置成如下格式:

测
测
测

但是输出的pred_results是空的。
又模仿训练集格式做成如下格式:

测 O
测 O
测 O

输出的pred_results依旧是空的。

请问是不是代码没有提供这部分功能呢？
如果有的话，请指正，感谢

[Llin1785361283]

@zhengguanyu 您好，我在调用模型进行测试的时候总是显示如下错误，请问您有遇到过类似的错误吗？
RuntimeError: Error(s) in loading state_dict for GazLSTM:
size mismatch for NERmodel.lstm.weight_ih_l0: copying a param of torch.Size([1200, 250]) from checkpoint, where the shape is torch.Size([800, 250]) in current model.
size mismatch for NERmodel.lstm.weight_hh_l0: copying a param of torch.Size([1200, 300]) from checkpoint, where the shape is torch.Size([800, 200]) in current model.
size mismatch for NERmodel.lstm.bias_ih_l0: copying a param of torch.Size([1200]) from checkpoint, where the shape is torch.Size([800]) in current model.
size mismatch for NERmodel.lstm.bias_hh_l0: copying a param of torch.Size([1200]) from checkpoint, where the shape is torch.Size([800]) in current model.
size mismatch for NERmodel.lstm.weight_ih_l0_reverse: copying a param of torch.Size([1200, 250]) from checkpoint, where the shape is torch.Size([800, 250]) in current model.
size mismatch for NERmodel.lstm.weight_hh_l0_reverse: copying a param of torch.Size([1200, 300]) from checkpoint, where the shape is torch.Size([800, 200]) in current model.
size mismatch for NERmodel.lstm.bias_ih_l0_reverse: copying a param of torch.Size([1200]) from checkpoint, where the shape is torch.Size([800]) in current model.
size mismatch for NERmodel.lstm.bias_hh_l0_reverse: copying a param of torch.Size([1200]) from checkpoint, where the shape is torch.Size([800]) in current model.
size mismatch for hidden2tag.weight: copying a param of torch.Size([7, 600]) from checkpoint, where the shape is torch.Size([7, 400]) in current model.