Lattice
A pure-Rust inference engine for transformer models on Apple Silicon, paired with a native macOS app for training, quantizing and chatting with local models.
Lattice is a pure-Rust inference engine for transformer models on Apple Silicon, implementing the full compute graph itself: weight loading, tokenization, the forward pass, quantization and LoRA training, with no dependency on ONNX, Python or CUDA. Hand-written Metal shaders accelerate inference on Apple GPUs, and SIMD kernels handle the CPU path.
It ships as five Rust crates for embedding, generation, quantization, LoRA fine-tuning and optimal transport, usable as a library or through a bundled binary with an OpenAI-compatible HTTP server. Lattice Studio, a companion SwiftUI macOS app, drives the same engine through a graphical panel for training LoRA adapters, comparing quantized models, and running multi-turn chat. It is Apache-2.0 licensed.