Reproduce baseline result #371
zoexu119
started this conversation in
PCQM4M-LSC
Replies: 1 comment 1 reply
-
|
Hi! The default hyper-parameters are what we used. I am not sure what happens, maybe pytorch version? I am not really sure, and I do not have time to investigate it for the time being. Sorry! |
Beta Was this translation helpful? Give feedback.
1 reply
-
|
Thank you, Weihua! I believe it's just due to randomness, not a big problem. |
Beta Was this translation helpful? Give feedback.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Hi OGB team! I ran the GIN-virtual baseline using the official code here with default hyperparameters. However, the best validation MAE is 0.1159, and it's worse than 0.1083 as you reported. Did you use a different set of hyperparameters? Thank you!
Beta Was this translation helpful? Give feedback.
All reactions