search

BorgwardtLab / TOGL

Topological Graph Neural Networks (ICLR 2022)
https://openreview.net/pdf?id=oxxUMeFwEHd
BSD 3-Clause "New" or "Revised" License
105 stars 20 forks source link

TopoGNN MUTAG error #10

Closed soham0209 closed 1 year ago

soham0209 commented 1 year ago

Hi, When running TopoGNN with poetry run python topognn/train_model.py --model TopoGNN --dataset MUTAG --max_epochs 10, the run fails with: /pytorch/aten/src/THCUNN/ClassNLLCriterion.cu:108: cunn_ClassNLLCriterion_updateOutput_kernel: block: [0,0,0], thread: [0,0,0] Assertiont >= 0 && t < n_classesfailed. I am almost sure that this is not an environment issue since other datasets such as DD, IMDB-MULTI runs successfully. My guess is that the labels for MUTAG is incorrect.

Pseudomanifold commented 1 year ago

I can't reproduce this error—does it still occur? We are using theTUDatasets version of MUTAG and are not updating labels ourselves.

soham0209 commented 1 year ago

Thank you, Bastien for replying. I am guessing that the version of pyg is different for me since I had to change pyg version to make it suitable for my environment.