Harry

The bug happens when loading the DUD-E and scPDB dataset

2

**Describe the bug** The bug happens when loading the DUD-E and scPDB dataset **To Reproduce** Steps to reproduce the behavior: 1. Just run the demo from your websites from tdc.generation...

Will you pre-train the supervised task with unsupervised tasks? Otherwise, how can you be sure that the content you learned before has not been forgotten? How do you think? :)

The values used to normalize the molecular descriptors.

8

Great work! But I have a sample question. Why do we use these values for normalization? How to get these values? ``` dists={ 'VSA_EState1': ('betaprime', (0.12979362790686721, 2.0084510281921832, -1.7874327445880742e-34, 0.4022005514286884), 0.0,...

CUDA out of memory

4

When I am reproducing your work, I find always CUDA out of memory. like this: Traceback (most recent call last): File "run.py", line 62, in trainer.train_and_test() File "/export/disk7/why/workbench/MERGE/v0/GLAM_repeat/GLAM/src_1gp_EmbeddingCompare/trainer.py", line 101,...

Maybe ComENet is not structure invariant?

1

When we use original xyz and xyz + 10 (just translation) The output is different.

Error happened with load data

5

**Describe the bug** The bug was happened while loading the data **To Reproduce** Steps to reproduce the behavior: from tdc.single_pred import Yields data = Yields(name = 'Buchwald-Hartwig') split = data.get_split()...

Why don't update the pre-trained model?

3

Hello, great work! Could you update the pre-trianed model to reproduce the result?