[BUG] Cannot read downloaded .npz files successfully for OpenKE datasets

[tingwl0122]

Describe the bug I cannot read in downloaded .npz files for some OpenKE datasets at /GLB-Repo/datasets, such as FB13and WN18RR. I didn't test all of the OpenKE datasets, but I could successfully run codes for cora, citeseer, and PubMed.

(A PR #86 is created to update dataset preparation in /GLB-Repo/glb/tags.py )

To Reproduce Run python3 tags.py --metadata FB13--task task at /GLB-Repo/glb

Expected behavior

Comments The error occurs at _dfs_read_file in graph.py. array = file_reader.get(path, d.get("key"), device) cannot merge path, keys and device name successfully.

[tingwl0122]

I think this is mainly due to data type? For OpenKE datasets, the code is stuck at reading the first key: node_name, where it supposes to be lots of characters; while for datasets like cora, the information is mainly tensor or csr_matrix.

[xingjian-zhang]

I will have a look at that! Thank you for reporting this issue.

[xingjian-zhang]

I have reproduced the error. The problem is that

1. We did not define how to handle string arrays, and
2. DGL does not support string tensors as node attributes (Neither does PyTorch.)

[xingjian-zhang]

After discussion, we decide to skip non-numeric attributes. I am opening a new PR.

[tingwl0122]

Got it! I'll test this after that. Thank you!

[tingwl0122]

Hi, I am not sure whether this issue was resolved? I tried to run tests on these OpenKE datasets today, but the same problem still occurred.

[xingjian-zhang]

Hi @tingwl0122 ! Could you please try this new PR on your machine?

[tingwl0122]

Sure. I'll try this in a second.