Skip to content

MoLFormer-data Box link returns 404; request canonical finetune_datasets.zip / BBBP split source #29

Description

@CREATEDCHARACTR

Hi IBM/MoLFormer maintainers,

I am trying to reproduce the BBBP split used by the MoLFormer/MoleculeNet fine-tuning path so we can run a clean external anchor against the public MAMMAL BBBP checkpoint:

  • model: ibm-research/biomed.omics.bl.sm.ma-ted-458m.moleculenet_bbbp
  • model revision: 4f249585b27907f0b5f191dba87cf3ecf2c23ef5
  • needed split: data/bbbp/test.csv from finetune_datasets.zip

The current README points to:

https://ibm.box.com/v/MoLFormer-data

That URL currently returns HTTP/2 404 from our environment. The README still describes the expected finetune_datasets.zip hierarchy, including:

data/
  bbbp/
    test.csv
    train.csv
    valid.csv

Could you provide one of the following?

  1. A working public URL for finetune_datasets.zip.
  2. The canonical data/bbbp/test.csv split used for the published MoLFormer MoleculeNet BBBP evaluation.
  3. A checksum or commit/source reference that lets downstream reviewers verify the split identity.

We are intentionally not substituting a third-party BBBP mirror because the review target is the canonical MoLFormer predefined split.

Contact: saul@projecttfb.com

Thank you.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions