API Reference

OmniGenBench is a comprehensive toolkit for genomic foundation models. It provides automated benchmarking and training pipelines, a hub for accessing pre-trained models, datasets, and pipelines, and a flexible framework for building custom models and tasks.

Tip

How to Use This Reference:

• Browse API documentation by functional category below.

• Each section contains detailed classes, functions, usage examples, and extension guidelines.

• Start with Abstract Classes to understand the core interfaces (OmniModel, OmniDataset, OmniTokenizer, OmniMetric).

• Explore Auto-Pipelines, CLI Commands, and Online Hubs for high-level workflows.

• Dive into Low-level Modules for detailed implementation and customization.

Quick Links:

• Trainers - Complete trainer documentation (Native, Accelerate, HuggingFace)

• Downstream Datasets - Dataset classes and usage

• Downstream Models - Model architectures

• CLI Commands - CLI command reference

Main API Categories:

• Abstract Classes: Core interfaces (OmniModel, OmniDataset, OmniTokenizer, OmniMetric) - use as base classes for extensions

• Auto-Pipelines: Automated benchmarking (AutoBench) and training (AutoTrain) workflows

• CLI Commands: Command-line tools for codeless operations (ogb, autobench, autotrain, autoinfer)

• Online Hubs: Access pre-trained models, datasets, and pipelines from HuggingFace Hub

• Low-level Modules: Detailed modules for models, datasets, metrics, tokenizers, trainers, and utilities

Refer to each section below for detailed documentation and usage instructions.