Welcome to Phi-3-MLX’s documentation!

Phi-3-MLX is a versatile AI framework that leverages both the Phi-3-Vision multimodal model and the recently updated Phi-3-Mini-128K language model, optimized for Apple Silicon using the MLX framework.

View the project on GitHub

Features

• Support for the Phi-3.5-Mini (language-only) model

• Integration with Phi-3.5-Vision (multimodal) model

• Optimized performance on Apple Silicon using MLX

• Batched generation for processing multiple prompts

• Flexible agent system for various AI tasks

• Custom toolchains for specialized workflows

• Model quantization for improved efficiency

• LoRA fine-tuning capabilities

• API integration for extended functionality (e.g., image generation, text-to-speech)

Usage

• Installation
• Text Generation
  • Visual Question Answering
  • Batch Text Generation
  • Structured Generation Using Constrained Decoding
  • Multiple Choice Selection
  • Model and Cache Quantization
• LoRA Fine-tuning
  • Training a LoRA Adapter
  • Generating Text with LoRA
  • Comparing LoRA Adapters
• Agent Interactions
  • Multi-turn Conversation
  • Generative Feedback Loop
  • External API Tool Use
• Custom Toolchains
  • Example 1: In-Context Learning Agent
  • Example 2: Retrieval Augmented Coding Agent
  • Example 3: Multi-Agent Interaction
  • Example 4. External LLM Integration
• Benchmark

API Reference

• Classes
  • Agent
• Functions
  • load
  • generate
  • choose
  • constrain
  • execute
  • train_lora
  • test_lora
  • benchmark
  • chat_ui
  • get_api
  • add_code
  • add_text
  • rag

License

This project is licensed under the MIT License.

Citation