The repo is finally unlocked. enjoy the party! The fastest repo in history to surpass 100K stars ⭐. Join Discord: https://discord.gg/5TUQKqFWd Built in Rust using oh-my-codex.

A Distributed Attention Towards Linear Scalability for Ultra-Long Context, Heterogeneous Data Training

Speed of Light Analysis for ML Model Runtime

A benchmark of real-world DL kernel problems

Universal LLM Deployment Engine with ML Compilation

Collection of memory microbenchmarks to investigate NVIDIA GPUs Network on Chip architectures

Learn Lean 4 with PLFA proofs.

A garden of small programming language implementations 🪴

Modify implementations for Pierce' Types and Programming Languages to add a REPL, convert into dune projects, and provide preconfigured development containers based on devfiles

dLLM: Simple Diffusion Language Modeling

Userspace eBPF runtime for Observability, Network, GPU & General Extensions Framework

Perplexity open source garden for inference technology

DELTA-pytorch：DELTA: Dynamically Optimizing GPU Memory beyond Tensor Recomputation

We invite you to visit and follow our new repository at https://github.com/microsoft/TileFusion. TiledCUDA is a highly efficient kernel template library designed to elevate CUDA C’s level of abstra…

TileFusion is an experimental C++ macro kernel template library that elevates the abstraction level in CUDA C for tile processing.

Tile-Based Runtime for Ultra-Low-Latency LLM Inference

gLLM: Global Balanced Pipeline Parallelism System for Distributed LLM Serving with Token Throttling

Curated collection of papers in machine learning systems

Rust version of THU uCore OS. Linux compatible.

[NeurIPS 2025] ClusterFusion: Expanding Operator Fusion Scope for LLM Inference via Cluster-Level Collective Primitive

CUDA Templates and Python DSLs for High-Performance Linear Algebra

The Higher-Order Intermediate Representation

Random stuff.

Training neural networks in TensorFlow 2.0 with 5x less memory