Your own personal AI assistant. Any OS. Any Platform. The lobster way. 🦞

A high-throughput and memory-efficient inference and serving engine for LLMs

TPU inference for vLLM, with unified JAX and PyTorch support.

Composable transformations of Python+NumPy programs: differentiate, vectorize, JIT to GPU/TPU, and more

Fast, Flexible and Portable Structured Generation

A set of ready to use Agent Skills for research, science, engineering, analysis, finance and writing.

Tensors and Dynamic neural networks in Python with strong GPU acceleration

Enabling PyTorch on Google TPU

Scoreboard for ONNX Backend Compatibility

Orbax provides common utility libraries for JAX users.

Flax is a neural network library for JAX that is designed for flexibility.

Open Neural Network Exchange

ONNX Runtime: cross-platform, high performance scoring engine for ML models

Backward compatible ML compute opset inspired by HLO/MHLO

Representation and Reference Lowering of ONNX Models in MLIR Compiler Infrastructure

An Open Source Machine Learning Framework for Everyone

Differentiable, Hardware Accelerated, Molecular Dynamics

tree is a library for working with nested data structures

Light weight design document templates in Markdown format.

Markdown <=> IPython Notebook

Books with Jupyter notebooks

Dive into Deep Learning Compiler

Simple Project Management and Document Templates in Markdown for Compatibility and Portability

Visualizer for neural network, deep learning, and machine learning models

Compiler for Neural Network hardware accelerators