mlx-examples: Explore Machine Learning Models with MLX

Overview

mlx-examples is a collection of standalone implementations built with the MLX framework. It gives developers and researchers practical starting points for learning MLX and exploring model training, inference, and generation across several machine learning domains.

The repository is most useful as a set of focused examples rather than a complete application or a full-featured language-model package. For a more complete Python package for LLMs with MLX, the project points users to MLX LM.

Key Features

• A beginner-friendly MNIST example for getting started with MLX.
• Language-model examples including Transformer training, LLaMA, Mistral, Mixtral, T5, and BERT.
• LoRA and QLoRA examples for parameter-efficient fine-tuning.
• Image generation and vision examples, including FLUX, Stable Diffusion, SDXL, ResNets, and a CVAE.
• Video generation examples for text-to-video and image-to-video with Wan2.1.
• Audio examples for Whisper speech recognition, EnCodec, and MusicGen.
• Multimodal and other model examples, including CLIP, LLaVA, SAM, GCN, and Real NVP.
• Links to converted checkpoints from the MLX Community on Hugging Face.

Use Cases

• MLX newcomers can use the MNIST example to learn the framework through a small, focused project.
• ML practitioners can inspect or adapt model examples when exploring training or inference workflows in MLX.
• Developers experimenting with language models can compare examples for generation, fine-tuning, or different model families.
• Researchers exploring image, video, audio, or multimodal tasks can use the corresponding examples as implementation references.

Getting Started

Start with the MNIST example, or browse the repository for an example matching your task. See the README for the project overview and links to individual examples.

Considerations

• These are standalone examples, not a single integrated application or a full-featured LLM package.
• The README does not provide one universal installation or run command. Refer to each example for its own setup and requirements.
• The repository is built around MLX; check the MLX project and individual example documentation for hardware and compatibility requirements.