vla-training

What problem does it solve?

This Skill simplifies the complex process of setting up and managing the training of Vision-Language-Action (VLA) models, particularly for autonomous driving applications.

Core Features & Use Cases

• End-to-End Workflow: Covers data preparation, distributed training configuration, and hyperparameter tuning.
• Multi-Modal Data Handling: Supports standard datasets like nuScenes and custom data formats.
• Distributed Training: Integrates with DeepSpeed and PyTorch FSDP for efficient large-model training.
• Use Case: When you need to train a VLA model for autonomous driving, this Skill provides the necessary scripts and configurations to handle data loading, set up distributed training across multiple GPUs, and define optimal training recipes.

Quick Start

Use the vla-training skill to set up a DeepSpeed training pipeline for autonomous driving VLA models.