Inference Deployment

This chapter provides a comprehensive introduction to the inference deployment workflow under the FluxVLA framework, applicable to multiple platforms including Aloha, Tron2, and the UR3 robotic arm. By following the standardized procedures, you can efficiently deploy trained models onto physical or simulated robots to perform end-to-end task inference and validation.

The content of this chapter is organized as follows:

• Aloha Physical Deployment A detailed guide on setting up the inference environment, loading models, and launching inference scripts on the Aloha platform. This covers the complete workflow from hardware connections and software configuration to task execution. See Aloha Physical Deployment for details.

• Tron2 Platform Deployment An introduction to the model inference deployment workflow on the Tron2 simulation and physical platforms, including environment preparation, parameter configuration, and debugging recommendations to help users quickly achieve model migration and multi-environment adaptation. See Tron2 Physical Deployment for details.

• UR3 Robotic Arm Deployment A detailed guide for UR3 physical robot inference deployment, covering data formats, environment setup, ROS Topic configuration, motion control interfaces, and a quick-start checklist to help users efficiently complete end-to-end deployment. See UR3 Physical Deployment for details.

Note:

• It is recommended to ensure that the training pipeline (including data collection and model training) has been completed as required before proceeding with inference deployment.

• If common issues are encountered during inference deployment, please refer to the FAQ sections in the respective chapters or seek community support.

For specific deployment details of a particular platform, please click the corresponding section in the left-hand navigation to access the complete guide.

• Aloha Physical Deployment
  • Overview
  • Power-On
  • 2. Software Environment Preparation
  • 3. Launch the Robotic Arms
  • 4. Launch the Cameras
  • 5. Run FluxVLA Model Inference
  • 6. Robotic Arm Reset
• Tron2 Physical Deployment
  • Tron2 Infra
  • FluxVLA Inference
• UR3 Physical Deployment
  • Table of Contents
  • 1. System Requirements
  • 2. Environment Setup
  • 3. Real-Robot Startup Workflow
  • 4. FluxVLA Inference
• Inference Acceleration
  • Overview
  • Acceleration Techniques
  • Configuration
  • Benchmarks
• RTC (Real-Time Chunking)
  • Training-time RTC
  • Test-time RTC
  • Testing
  • Test visualization
  • Supported models