Shuffling error in test submission

[Open Graph Benchmark]

Hi OGB-LSC participants,

We have noticed several initial test submissions got unexpectedly low accuracy since test predictions were mistakingly shuffled, e.g., setting shuffle=True in the test data loader.

For the final test submission, please be careful about this error as well as other potential errors. The final submission will be strict, and we cannot resolve individual errors after the deadline.

Please kindly keep this in mind. We are looking forward to your final results.

Best,

OGB-LSC Team