全链接层的建立准确率低

全连接层的建立

class Net(torch.nn.Module):

def __init__(self, hid):
    super(Net, self).__init__()
    self.linear1 = torch.nn.Linear(2, hid)
    self.linear2 = torch.nn.Linear(hid, hid)
    self.linear3 = torch.nn.Linear(hid, 1)

def forward(self,hid):
    linear1 = self.linear1(hid)
    self.hid1 = torch.tan(linear1)
    linear2 = self.linear2(self.hid1)
    self.hid2 = torch.tan(linear2)
    linear3 = self.linear3(self.hid2)
    self.output = torch.sigmoid(linear3)
    return self.output

运行结果及报错内容

我在尝试建立一个全连接层 这段代码是否存在错误呢，训练的准确率一直只有0.5

我想要达到的结果