search

amazon-science / tgl

Apache License 2.0
192 stars 31 forks source link

Issue with GDELT node features #10

Closed CongWeilin closed 2 years ago

CongWeilin commented 2 years ago

The node features are all the same:

import torch
node_feats = torch.load('DATA/GDELT/node_features.pt')
if node_feats.dtype == torch.bool:
    node_feats = node_feats.type(torch.float32)
torch.unique(node_feats.argmax(dim=1))

Gives us output tensor([79])

The node_features.pt is downloaded by wget -P ./DATA/GDELT https://s3.us-west-2.amazonaws.com/dgl-data/dataset/tgl/GDELT/node_features.pt

tedzhouhk commented 2 years ago

Hi, thanks for pointing this out. I believe I have uploaded the wrong version of the GDELT dataset to the s3 bucket. I will upload the correct version (this might take a while as these datasets are generated last year when I was a intern at AWS).

tedzhouhk commented 2 years ago

Hi, I have re-upload the dataset. Please download ALL FILES in the GDELT folder from the s3 bucket. I'll also notice it in the readme. Thanks.