Educational repo: building a tiny MolFormer-style encoder transformer from scratch for chemists. Sister course to GNNs-For-Chemists.
-
Updated
Sep 2, 2026 - Jupyter Notebook
Educational repo: building a tiny MolFormer-style encoder transformer from scratch for chemists. Sister course to GNNs-For-Chemists.
Fine-tuned chemical language model for predicting molecular lipophilicity in drug design. Explores parameter-efficient fine-tuning strategies (LoRA, BitFit, IA3), layer freezing techniques, and influence-based data selection. Balances accuracy and computational efficiency for molecular property prediction tasks.
CytoBridge: single-cell perturbation prediction with pathway-gated cross-attention; frozen benchmark for 'Beyond per-pair correlation: a drug-discrimination control exposes prediction collapse in single-cell multi-drug perturbation models'
Reproducible fluorescence lifetime prediction on unseen scaffolds: pretrained molecular models, solvent conditioning, and multitask learning.
Predicting Critical Micelle Concentration (CMC) of Surfactants Using Machine Learning
To associate your repository with the molformer topic, visit your repo's landing page and select "manage topics."