Add experiment scripts

[ekarais]

Description

• The main purpose of this PR is to adapt the repo such that experiments can be run on Google Colab. These adaptations will likely be removed from the final version as many people do not have access to Colab.
• The secondary purpose is to add the experiment scripts which we reported in the paper.
• This PR also adds model checkpointing, which is crucial for longer training runs.
• Finally, this PR changes the preprocessed data storage format from a directory of .json files to a single .jsonl file. We made this change because QM9 used to result in 130K .json files, which are both inefficient to iterate over and use a lot of resources on the cluster.

List of commits

1. Adaptation to Google Colab:
   • Create the Colab notebook: a33765a, 2053817, 72e5783, 3330acb, 3ae43fb, 911bccf, dde56aa, 9c418b5
   • Adapt the conda environment: 0df6c78, 72e5783, 742b018, 1c00dd3
   • Adapt existing training scripts
     • Remove env activation commands: d530f04
     • Remove torch.compile: ad228f4
     • Launch multiple jobs from the same script: a8dfd09
   • Upload metadata of QM9: 95b7e39, 34e595c, 7ac3037, 773fa5a
2. Add experiment scripts: 2995a0e, 21d07dc, 38c8d52, a25d167, 774e5fc, 11208cb, 5f9a58c, 8dc3110, d3502de, 2b14826, 232dcfd, 755d0dc, f3e6303, 54d157c, 8ef81b9, 42917f2, bfc936a, c574002
3. Model checkpointing: 9ac82ce, aa6b86c, bddccde
4. Miscellanous:
   • Memory optimization: b82f92d
   • Update README with instruction to run experiments: d140c88
   • Anonymize repository: 9904f7b, 9ce3486, 99c4d1c, 0c1b2d7, d1930d4
   • Change data storage format to .jsonl: 52e0e3d
   • Rename the wandb project: 9645d82

Notes

• We replaced torch_scatter with a custom scatter_add to remove the dependency, but ultimately brought back the torch_scatter version.