search

Forschungsprojekt-Informatik / CompGCN

ICLR 2020: Composition-Based Multi-Relational Graph Convolutional Networks
Apache License 2.0
0 stars 0 forks source link

Evaluating results that are trained on a cuda device is not possible (out-of-the-box) on cpu #15

Closed 1Maxnet1 closed 1 year ago

1Maxnet1 commented 1 year ago

This is not a too big deal, but makes it hard to check the results, except if you have working cuda, which is not always the case. Workaround: Change line 256 in run.py to the following:

state           = torch.load(load_path, map_location=torch.device('cpu'))

It would be nice to have the run.py to check itself whether cuda is available and set the parameter accordingly if needed.