bprmf

    def evaluate(self, u_ids):
        u_e = self.user_embeddings(u_ids)
        return torch.matmul(u_e, self.item_embeddings.weight.t())

    def forward(self, u_ids, i_ids):
        u_e = self.user_embeddings(u_ids)  # [batch_size, embedding_size]
        i_e = self.item_embeddings(i_ids)   # [batch_size, embedding_size]
        return torch.bmm(u_e.unsqueeze(1), i_e.unsqueeze(2)).squeeze()      # [batch_size]