GRiT: A Generative Region-to-text Transformer for Object Understanding (https://arxiv.org/abs/2212.00280)

A library with extensible implementations of DPO, KTO, PPO, ORPO, and other human-aware loss functions (HALOs).

Synthetic data curation for post-training and structured data extraction

SkyReels V1: The first and most advanced open-source human-centric video foundation model

llama3 implementation one matrix multiplication at a time

Get up and running with Llama 3.3, DeepSeek-R1, Phi-4, Gemma 3, Mistral Small 3.1 and other large language models.

Maya: An Instruction Finetuned Multilingual Multimodal Model using Aya

Small collection of data structures for performing common mathematical operations in scientific computing.

Official Code for Paper "Think While You Generate: Discrete Diffusion with Planned Denoising" [ICLR 2025]

A collection of notebooks/recipes showcasing some fun and effective ways of using Claude.

Simple and Effective Masked Diffusion Language Model

A throughput-oriented high-performance serving framework for LLMs

A query and indexing engine for Redis, providing secondary indexing, full-text search, vector similarity search and aggregations.

[NeurIPS'23 Oral] Visual Instruction Tuning (LLaVA) built towards GPT-4V level capabilities and beyond.

SoTA production-ready AI retrieval system. Agentic Retrieval-Augmented Generation (RAG) with a RESTful API.

RL Agent to produce sounds using the famous Pink Trombone

llvm project

DenseFusion-1M: Merging Vision Experts for Comprehensive Multimodal Perception

[EMNLP 2024 Findings] OneGen: Efficient One-Pass Unified Generation and Retrieval for LLMs.

tiny vision language model

Implementation of Autoregressive Diffusion in Pytorch

Animation engine for explanatory math videos

Together Mixture-Of-Agents (MoA) – 65.1% on AlpacaEval with OSS models

Open-Sora: Democratizing Efficient Video Production for All

Fast stable diffusion on CPU

A replication of Diffusion-LM Improves Controllable Text Generation

LLM training in simple, raw C/CUDA

Minimal, clean code for the Byte Pair Encoding (BPE) algorithm commonly used in LLM tokenization.

Inference Llama 2 in one file of pure C