API Reference

Complete API reference for the Molfun library. Each page documents the public classes, functions, and registries exposed by a subpackage.

Core API

Page	Description
MolfunStructureModel	Main facade for training, prediction, export, and hub operations
Predict Functions	High-level predict_structure, predict_properties, predict_affinity
PEFT	Parameter-Efficient Fine-Tuning utilities (LoRA, IA3)

Training

Page	Description
Strategies	Fine-tuning strategies: Full, HeadOnly, LoRA, Partial
Distributed	DDP and FSDP distributed training support

Modular Architecture

Page	Description
Registry	Generic ModuleRegistry plugin system
Attention	Attention implementations: Flash, Standard, Linear, Gated
Blocks	Trunk blocks: Pairformer, Evoformer, SimpleTransformer
Structure Modules	IPA and Diffusion structure modules
Embedders	Input and ESM embedders
ModelBuilder	Programmatic model assembly
ModuleSwapper	Hot-swap modules in a live model

Data

Page	Description
Fetchers	PDBFetcher, AffinityFetcher, MSAProvider
Datasets	StructureDataset, AffinityDataset, StreamingStructureDataset
Splits	DataSplitter: random, temporal, identity
Parsers	PDB, mmCIF, A3M, FASTA, SDF, Mol2 parsers
Collections	Curated dataset collections

Losses, Tracking, Export & CLI

Page	Description
Losses	Loss registry and built-in loss functions
Tracking	Experiment trackers: W&B, Comet, MLflow, Langfuse, HuggingFace, Console
Export	ONNX and TorchScript export utilities
CLI	Command-line interface reference