求助，复现TNet模型时遇到了这个报错。

[Gaoshuang77]

您好，请问如何解决呢？谢谢。 epoch: 0 Traceback (most recent call last): File "train.py", line 294, in main() File "train.py", line 290, in main ins.run() File "train.py", line 170, in run best_model_path = self._train(criterion, optimizer, train_data_loader, val_data_loader) File "train.py", line 105, in _train outputs = self.model(inputs) File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py", line 550, in call result = self.forward(*input, **kwargs) File "/content/gdrive/Shared drives/ABSA-PyTorch-master/models/tnet_lf.py", line 67, in forward a = torch.bmm(e.transpose(1, 2), v) RuntimeError: expected scalar type Float but found Double

[Gaoshuang77]

我尝试在报错这行“a = torch.bmm(e.transpose(1, 2), v)”之后加入“v = torch.tensor(v, dtype=torch.float32)”将v转换成float向量，但是遇到了另一个报错： Traceback (most recent call last): File "train.py", line 294, in main() File "train.py", line 290, in main ins.run() File "train.py", line 170, in run best_model_path = self._train(criterion, optimizer, train_data_loader, val_data_loader) File "train.py", line 105, in _train outputs = self.model(inputs) File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py", line 550, in call result = self.forward(*input, kwargs) File "/content/gdrive/Shared drives/ABSA-PyTorch-master/models/tnet_lf.py", line 76, in forward z = F.relu(self.convs3(v)) # [(N,Co,L), ...]len(Ks) File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/module.py", line 550, in call result = self.forward(input, kwargs) File "/usr/local/lib/python3.6/dist-packages/torch/nn/modules/conv.py", line 208, in forward self.padding, self.dilation, self.groups) RuntimeError: Expected object of scalar type Double but got scalar type Float for argument #3 'mat1' in call to _thaddmm 请问到底应该从哪里改呢？

[Gaoshuang77]

@songyouwei 目前还是没有解决，请问您知道如何修改吗？

[songyouwei]

可能和版本有关，有没有安装requirements.txt呢？

[Gaoshuang77]

嗯嗯，安装了requirements.txt，除了用了较小的glove以外，别的地方没有改动。

[gpf951101]

v = torch.tensor(v, dtype=torch.float32).to(self.opt.device) 我再for循环的第一行添加了上面这行代码

[Gaoshuang77]

@gpf951101 您加了这行是可以运行的吗？我刚刚去试了一下，但是报错“RuntimeError: Expected object of scalar type Double but got scalar type Float for argument #3 'mat1' in call to _thaddmm”

[GeneZC]

@Gaoshuang77 你是在cpu上跑的吧

[Gaoshuang77]

@GeneZC 是的，在CPU上跑的。

[GeneZC]

@Gaoshuang77 那就检查一下你在哪里引入了double（float64）类型的变量，按理说这份代码都应该是float（float32）类型

[dongyang12138]

请问您解决了您的问题吗，我也遇到了同样的问题

[GeneZC]

@dongyang12138 @Gaoshuang77 try removing line 27 in tfnet_lf.py: pos_inx = pos_inx.cpu().numpy()

[GeneZC]

Or downgrade Pytorch version <= 1.4.0

[GeneZC]

And the best solution may be change weight = torch.tensor(weight) to weight = torch.tensor(weight).float()

[Ottehack]

我在别的模型上也出现了Float和Double类型的问题，后来我去降级了pytorch，从1.5.1降到1.4就没问题了