API Overview

This section serves as the navigation hub for the FluxVLA API Reference, covering the following areas:

• Training interfaces (scripts/train.sh and training configuration)

• Evaluation interfaces (scripts/eval.sh and evaluation configuration)

• Inference interfaces (Runner, Operator, and data processing for real-robot and simulation deployment)

• Configuration interfaces (model, train_dataloader, runner, inference, eval)

What Do You Want to Do?

Train a Model

• Start with the training entry point and parameters: train_eval_scripts

• Then refer to the quick-start training guide: ../start/vla.md

• For private datasets: ../tutorials/private_dataset_config.md

Evaluate a Model

• Start with the evaluation entry point and parameters: train_eval_scripts

• Then refer to the quick-start evaluation guide: ../start/vla-eval.md

Deploy for Inference

• Start with the inference interface overview: inference_interfaces

• Then refer to the platform deployment tutorial: ../tutorials/inference/index.rst

Modify Configuration

• Start with the configuration field index: config_schema

• Then refer to the complete configuration tutorial: ../tutorials/config/index

Reference Structure

• General Conventions: Parameter notation, path and naming conventions, input/output field conventions

• Training/Evaluation Script Interfaces: Command entry points, positional arguments, environment variables, common override methods

• Model Interfaces: Model components, registration mechanism, configuration integration

• Dataset Interfaces: Data structures, field mappings, transform pipelines

• Inference Interfaces: Runner/Operator, platform deployment entry points

• Configuration Schema: Key top-level configuration keys and common fields

Related Resources

• Quick Start: ../start/index.rst

• Advanced Tutorials: ../tutorials/index.rst

• FAQ: ../faq.md