Analyze computation-communication overlap in V3/R1.

Implementation for SimDINO/SimDINOv2

DeepEP: an efficient expert-parallel communication library

Quantized Attention that achieves speedups of 2.1-3.1x and 2.7-5.1x compared to FlashAttention2 and xformers, respectively, without lossing end-to-end metrics across various models.

Wan: Open and Advanced Large-Scale Video Generative Models

PyTorch implementation of FractalGen https://arxiv.org/abs/2502.17437

An obsidian template vault for tracking your academic life.

Redis for humans. 🌎🌍🌏

Cost-efficient and pluggable Infrastructure components for GenAI inference

Official repo of paper LM2

A token pruning method that accelerates ViTs for various tasks while maintaining high performance.

[Oral; Neurips OPT2024 ] μLO: Compute-Efficient Meta-Generalization of Learned Optimizers

Official implementation of the paper "You Do Not Fully Utilize Transformer's Representation Capacity"

Towards Economical Inference: Enabling DeepSeek's Multi-Head Latent Attention in Any Transformer-based LLMs

I Think, Therefore I Diffuse: Enabling Multimodal In-Context Reasoning in Diffusion Models

YuE: Open Full-song Music Generation Foundation Model, something similar to Suno.ai but open

Official code for paper: F3D-Gaus: Feed-forward 3D-aware Generation on ImageNet with Cycle-Consistent Gaussian Splatting

LookHere position encoding for ViTs (NeurIPS 2024)

[3DV'25] 3D Reconstruction with Spatial Memory

Efficient End2End Compiler for Mixed-Precision Deep Learning

Reproducible evaluation of NeRF and 3DGS methods

This repository provides tutorials and implementations for various Generative AI Agent techniques, from basic to advanced. It serves as a comprehensive guide for building intelligent, interactive A…

📦 Repomix (formerly Repopack) is a powerful tool that packs your entire repository into a single, AI-friendly file. Perfect for when you need to feed your codebase to Large Language Models (LLMs) o…

FlashMLA: Efficient MLA Decoding Kernel for Hopper GPUs

Domain-specific language designed to streamline the development of high-performance GPU/CPU/Accelerators kernels

A CPU Realtime VLM in 500M. Surpassed Moondream2 and SmolVLM. Training from scratch with ease.

MoH: Multi-Head Attention as Mixture-of-Head Attention