知识图谱补全实验微调记录

四个数据集：FB5k，FB15k-237，WN18，WN18RR
model = TuckER
iteration%50==0 torch.save(model.sate_dict(), PATH)
训练模型，并保存记录。(√)

TuckER模型代码

class TuckER(torch.nn.Module):
    def __init__(self, d, d1, d2, **kwargs):
        super(TuckER, self).__init__()

        self.E = torch.nn.Embedding(len(d.entities), d1, padding_idx=0)
        self.R = torch.nn.Embedding(len(d.relations), d2, padding_idx=0)
        self.W = torch.nn.Parameter(torch.tensor(np.random.uniform(-1, 1, (d2, d1, d1)), 
                                    dtype=torch.float, device="cuda", requires_grad=True))

        self.input_dropout = torch.nn.Dropout(kwargs["input_dropout"])
        self.hidden_dropout1 = torch.nn.Dropout(kwargs["hidden_dropout1"])
        self.hidden_dropout2 = torch.nn.Dropout(kwargs["hidden_dropout2"])
        self.loss = torch.nn.BCELoss()

        self.bn0 = torch.nn.BatchNorm1d(d1)
        self.bn1 = torch.nn.BatchNorm1d(d1)
        

    def init(self):
        xavier_normal_(self.E.weight.data)
        xavier_normal_(self.R.weight.data)

    def forward(self, e1_idx, r_idx):
        e1 = self.E(e1_idx)
        x = self.bn0(e1)
        x = self.input_dropout(x)
        x = x.view(-1, 1, e1.size(1))

        r = self.R(r_idx)
        W_mat = torch.mm(r, self.W.view(r.size(1), -1))
        W_mat = W_mat.view(-1, e1.size(1), e1.size(1))
        W_mat = self.hidden_dropout1(W_mat)

        x = torch.bmm(x, W_mat) 
        x = x.view(-1, e1.size(1))      
        x = self.bn1(x)
        x = self.hidden_dropout2(x)
        x = torch.mm(x, self.E.weight.transpose(1,0))
        pred = torch.sigmoid(x)
        return pred

加载预训练好的模型，并在此基础上建立其他模型来微调，使得补全结果变得更好。
1)将原始模型的参数固化，即，让这些参数不用被反向调参。
2)在原始模型的基础之上再搭建一个模型，让这个新模型进一步优化评价指标。

RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation

加载固化的参数

class TuckER(torch.nn.Module):
    def __init__(self, nfeat, nhid, dropout, d, d1, d2, **kwargs):
        super(TuckER, self).__init__()

        ###用加载的模型继续训练或操作, model_current_dict
        dir = os.getcwd() + "/save_models"
        PATH = os.path.join(dir,  "WN18tucker_200_weight_gpu.pth")
        pretrained_dict = torch.load(PATH)

        self.E = torch.nn.Embedding(len(d.entities), d1, padding_idx=0)
        self.E.weight.data.copy_(pretrained_dict['E.weight'])
        self.R = torch.nn.Embedding(len(d.relations), d2, padding_idx=0)
        self.R.weight.data.copy_(pretrained_dict['R.weight'])

        self.W = pretrained_dict['W']
        # self.W = torch.nn.Parameter(torch.tensor(np.random.uniform(-1, 1, (d2, d1, d1)),
        #                                          dtype=torch.float, requires_grad=False))

        # self.E = torch.nn.Embedding(len(d.entities), d1, padding_idx=0)
        # self.R = torch.nn.Embedding(len(d.relations), d2, padding_idx=0)
        # self.W = torch.nn.Parameter(torch.tensor(np.random.uniform(-1, 1, (d2, d1, d1)),
        #                                          dtype=torch.float, requires_grad=False))

        self.E.weight.requires_grad = False
        self.R.weight.requires_grad = False

        self.bn0 = torch.nn.BatchNorm1d(d1)
        self.bn1 = torch.nn.BatchNorm1d(d1)

        self.input_dropout = torch.nn.Dropout(kwargs["input_dropout"])
        self.hidden_dropout1 = torch.nn.Dropout(kwargs["hidden_dropout1"])
        self.hidden_dropout2 = torch.nn.Dropout(kwargs["hidden_dropout2"])
        self.loss = torch.nn.BCELoss()