Zum Inhalt

Developer Guide Overview

Welcome to the Developer Guide for DeepExtension — a flexible, extensible platform for building domain-specific AI models. This guide is designed to help developers, engineers, and technically curious users understand how to extend and customize DeepExtension for real-world usage.

Whether you're deploying on a local machine or preparing your own training logic, this section provides everything you need to set up, integrate, and build with confidence — even if you're new to LLM fine-tuning.

What You'll Learn

This guide will walk you through:

  • Supported Hardware & Performance Notes
    Understand which hardware configurations work best for inference, training, and deployment — and how they affect speed and capacity.

  • Installation Guide
    How to set up DeepExtension from scratch on supported platforms like Linux or Windows via WSL (CUDA-enabled) and macOS.

  • Base Model Management
    Learn how to register and organize local base models that serve as the foundation for all fine-tuning activities.

  • Implementing Your Own Training (CUDA（Linux or Windows via WSL))
    Extend DeepExtension by integrating your own training script (LoRA, SFT, GRPO, etc.) using our built-in templates.

  • Implementing Your Own Training (macOS + MLX)
    If you're developing on macOS, we offer support for lightweight training using Apple Silicon and MLX. You'll learn how to integrate and test custom MLX-based training workflows.

Who This Is For

This section is for:

  • Developers with Python experience who want to extend model training logic
  • MLOps engineers managing training environments and hardware
  • AI researchers prototyping domain-specific models
  • Technical users seeking to customize DeepExtension for internal workflows

You do not need to be an AI expert to follow along — we provide templates, configuration tips, and clear entry points so you can build progressively.

Ready to Start?

Choose a chapter below to dive in: