Would you mind provide the *-textual.pth and *-visual.pth

[zhiweihu1103]

I see MMRNS only provide the .h5 embedding files, would you mind provide the .pth you used in the paper. Thx.

[Zhang-Each]

You can simply process the h5 embeddings into a tensor in the shape of [num_entity, embedding_dim]

[Zhang-Each]

I've been busy to find an summer intern and will update this repository lately .

[12hsk]

I tried to convert the embeddings from the H5 file into tensor format of [num_entity, embedding_dim], but I still encountered various errors afterwards. Could you please provide the .pth source file used in the project?

您可以简单地将 h5 嵌入处理成 [num_entity， embedding_dim] 形状的张量

[Zhang-Each]

I tried to convert the embeddings from the H5 file into tensor format of [num_entity, embedding_dim], but I still encountered various errors afterwards. Could you please provide the .pth source file used in the project?

您可以简单地将 h5 嵌入处理成 [num_entity， embedding_dim] 形状的张量

Hello, we have upload the embeddings and you can download from the link provided in README.

[zhiweihu1103]

why the test set and valid set is reversed for DB15K dataset? The statistics of paper said that the test set is 9904, however, the test set in this github is 9902, Which test set did you use?

[Zhang-Each]

why the test set and valid set is reversed for DB15K dataset? The statistics of paper said that the test set is 9904, however, the test set in this github is 9902, Which test set did you use?

Sorry, this might be a typo in the paper. Our data split follows the setting of MMRNS https://github.com/quqxui/MMRNS

[zhiweihu1103]

test set and valid set for DB15K are 9902 and 9904, respectively, right?

[Zhang-Each]

test set and valid set for DB15K are 9902 and 9904, respectively, right?

yes

[zhiweihu1103]

Thx.