Mixture-of-experts (MoE) training megakernel for NVL72s

CODA: Rewriting Transformer Blocks as GEMM-Epilogue Programs

Official PyTorch re-implementation of MiniT2I.

Official inference code for Krea 2

Ideogram 4: Open image model at the forefront of design

[ICLR 2026] HILBERT-GUIDED SPARSE LOCAL ATTENTION

A storage solution for PyTorch tensors with distributed tensor support.

A trivial CLI library. https://docs.kidger.site/seali/

DeepGEMM: clean and efficient BLAS kernel library on GPU

MoE training for Me and You and maybe other people

A cli tool to browse and play anime

A simple, performant, and scalable Jax LLM!

Kubernetes-native Job Queueing

Diffusion-SDPO: Safeguarded Direct Preference Optimization for Diffusion Models

Official Implementation of "Maximum Likelihood Reinforcement Learning (MaxRL)"

[ICLR 2026 Oral] DiffusionNFT: Online Diffusion Reinforcement with Forward Process

Inference at the speed of light.

Official Python inference and LoRA trainer package for the LTX-2 audio–video generative model.

Directly Aligning the Full Diffusion Trajectory with Fine-Grained Human Preference

Miles is an enterprise-facing reinforcement learning framework for LLM and VLM post-training, forked from and co-evolving with slime.

Krea Realtime 14B. An open-source realtime AI video model.

VQVAEs, GumbelSoftmaxes and friends

✨ Making your shell magical

Multilingual Document Layout Parsing in a Single Vision-Language Model

TORAX: Tokamak transport simulation in JAX