[ASPLOS 2026] CLM: Removing the GPU Memory Barrier for 3D Gaussian Splatting with CPU Offloading

Quantized LLM training in pure CUDA/C++.

Accelerating MoE with IO and Tile-aware Optimizations

DGGT: Feedforward 4D Reconstruction of Dynamic Driving Scenes using Unposed Images

A TUI-based utility for real-time monitoring of InfiniBand traffic and performance metrics on the local node

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

A fast communication-overlapping library for tensor/expert parallelism on GPUs.

StreamingVLM: Real-Time Understanding for Infinite Video Streams

Ola: Pushing the Frontiers of Omni-Modal Language Model

DiffusionDriveV2: Reinforcement Learning-Constrained Truncated Diffusion Modeling in End-to-End Autonomous Driving

FlagGems is an operator library for large language models implemented in the Triton Language.

Collection of kernels written in Triton language

A lightweight profiler for NCCL

Data processing for and with foundation models! 🍎 🍋 🌽 ➡️ ➡️🍸 🍹 🍷

Learning to Drive via Real-World Simulation at Scale

A NCCL extension library, designed to efficiently offload GPU memory allocated by the NCCL communication library.

A portable Multimodal Lakehouse powered by Ray that brings exabyte-level scalability and fast, ACID-compliant, change-data-capture to your ML and analytics workloads.

Accelerated Computer Vision Lab (ACCV-Lab) is a systematic collection of packages with the common goal to facilitate end-to-end efficient training in the ADAS domain, each package offering tools & …

NVIDIA Resiliency Extension is a python package for framework developers and users to implement fault-tolerant features. It improves the effective training time by minimizing the downtime due to fa…

SAM 3D Objects

Discover Unknown Unsafe Events via Generative Simulation

High-performance data engine for AI and multimodal workloads. Process images, audio, video, and structured data at any scale

DriveArena: A Closed-loop Generative Simulation Platform for Autonomous Driving

Depth Anything 3

[ECCV 2024] DriveDreamer: Towards Real-world-driven World Models for Autonomous Driving

[CVPR 2025] OmniSplat: Taming Feed-Forward 3D Gaussian Splatting for Omnidirectional Images with Editable Capabilities