ValueError:only one element tensors can be converted to Python scalars

问题描述

    使用超图卷积时报错

    源程序如下所示：

class HyperConv(nn.Module):
    def __init__(self, layers, emb_dim):
        super(HyperConv, self).__init__()
        self.layers = layers
        self.emb_dim = emb_dim
        self.hypergraph_conv = nn.ModuleList([])
        for i in range(self.layers):
            self.hypergraph_conv.append(HGNN_conv(self.emb_dim, False))

    def forward(self, adjacency, embedding):
        item_embeddings = embedding
        item_embedding_layer0 = item_embeddings
        final = [item_embedding_layer0]
        for i in range(self.layers):
            item_embeddings = self.hypergraph_conv[i](item_embeddings, adjacency)
            final.append(item_embeddings)

        item_embeddings = np.sum(final, 0) / (self.layers+1)
        return item_embeddings

---

解决方案：

    final为包含多个tensor的多维list，先将其转化为torch.Tensor类型，再用torch.sum()进行求和，即可解决问题。修改代码如下所示：

item_embeddings = torch.sum(torch.stack(final), 0)
item_embeddings = item_embeddings / (self.layers+1)