10x faster matrix and vector operations

Accelerating Diffusion Transformers with Token-wise Feature Caching

Efficient AI Inference & Serving

A Python library transfers PyTorch tensors between CPU and NVMe

✈️ Accelerating Vision Diffusion Transformers with Skip Branches.

A very simple and barebones tensor decomposition library for CP decomposition a.k.a. PARAFAC a.k.a. TCA

TensorLy: Tensor Learning in Python.

Stateful load balancer custom-tailored for llama.cpp 🏓🦙

Making large AI models cheaper, faster and more accessible

Open-Sora: Democratizing Efficient Video Production for All

jax-triton contains integrations between JAX and OpenAI Triton

Zero-copy MPI communication of JAX arrays, for turbo-charged HPC applications in Python ⚡

RUDOLPH: One Hyper-Tasking Transformer can be creative as DALL-E and GPT-3 and smart as CLIP

Tensor library for machine learning

Efficient LLM Inference over Long Sequences

PyTorch bindings for CUTLASS grouped GEMM.

A tool for bandwidth measurements on NVIDIA GPUs.

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

Library for reading and processing ML training data.

a language for fast, portable data-parallel computation

Best practices & guides on how to write distributed pytorch training code

A native PyTorch Library for large model training

A Pytorch implementation of Sparsely-Gated Mixture of Experts, for massively increasing the parameter count of language models

PyTorch library for cost-effective, fast and easy serving of MoE models.

Development repository for the Triton language and compiler

Build system, successor to Buck

On-device AI across mobile, embedded and edge for PyTorch

LiteRT is the new name for TensorFlow Lite (TFLite). While the name is new, it's still the same trusted, high-performance runtime for on-device AI, now with an expanded vision.