LNP Modeling Components

This repository contains reusable neural network architecture components for molecular and formulation modeling.

Included

• Molecular feature encoders
• Token projection layers
• Set-based attention modules
• Mixture-of-experts layers
• Feature fusion modules
• Prediction heads

Excluded

This repository does not include:

• Training or evaluation datasets
• Processed data
• Model checkpoints
• Precomputed embeddings
• Experimental results
• Prediction outputs
• Server-specific configuration
• Private file paths

Users must provide their own authorized data and runtime configuration.

Installation

pip install -e .

License

See LICENSE.