Kohaku-Blueleaf

1Cat-vLLM. vLLM fork for Tesla V100 (SM70) with AWQ 4-bit support, CUDA 12.8 build flow, and validated Qwen3.5 27B/35B deployment on multi-GPU V100.

flash-muon. Flash-Muon: An Efficient Implementation of Muon Optimizer

Muon. Muon is an optimizer for hidden layers in neural networks

verilator. Verilator open-source SystemVerilog simulator and lint system

libjxl. JPEG XL image format reference implementation

min. A fast, minimal browser that protects your privacy

claw-code. An agent-managed museum exhibit, built in Rust with Gajae-Code / LazyCodex — developed and maintained with no human intervention.

ccproxy-api. is a local reverse proxy providing unified access to multiple AI providers (Claude, Codex) through a single interface, using your existing subscriptions without API key billing.

MSLK. MSLK (Meta Superintelligence Labs Kernels) is a collection of PyTorch GPU operator libraries that are designed and optimized for GenAI training and inference, such as FP8 row-wise quantization and collective communications.

tt-forge. Tenstorrent's MLIR Based Compiler. We aim to enable developers to run AI on all configurations of Tenstorrent hardware, through an open-source, general, and performant compiler.

taxi. AXI, AXI stream, Ethernet, and PCIe components in System Verilog

KohakuTerrarium. KohakuTerrarium is a general-purpose AI agent framework and batteries-included app for building, running, and composing self-contained agents and multi-agent teams, with built-in tools, sub-agents, persistent sessions, TUI, and web UI.

ComfyUI-LG_SamplingUtils. Python

LightX2V. Lightweight Image Video Action Generation Inference Framework

OptimFactory. Python

claudish. Claude Code. Any Model. The most powerful AI coding agent now speaks every language.

torchcodec. PyTorch media decoding and encoding

S3impleClient. Python

GRAPE. [ICLR 2026] GRAPE: Group Representational Position Encoding (https://arxiv.org/abs/2512.07805)

manim. A community-maintained Python framework for creating mathematical animations.

KohakuCaption. Python

KohakuRAG. Simple Hierarchical RAG Framework

decord2. An efficient video loader for deep learning with smart shuffling that's super easy to digest

friendly-words. The Glitch word list, packaged into an NPM module for easy use.

KohakuBoard. High efficiency Local/self-hosted ML Experiment Tracking System

KohakuVault. Simple Single File Key-Value Store DB Based on SQLite

OpenDeepWiki. OpenDeepWiki is the open-source version of the DeepWiki project, aiming to provide a powerful knowledge management and collaboration platform. The project is mainly developed using C# and TypeScript, supporting modular design, and is easy to expand and customize.

ImagenWorld. Stress-Testing Image Generation Models with Explainable Human Evaluation on Open-ended Real-World Tasks [ICLR 2026]

peewee-async. Asynchronous interface for peewee ORM powered by asyncio

clash-for-linux-install. 😼 优雅地使用基于 clash/mihomo 的代理环境

KohakuHub. A self-hosted HuggingFace alternative

valkey. A flexible distributed key-value database that is optimized for caching and other realtime workloads.

RamTorch. RAM is all you need

intro-to-omniverse-isaac-supp. Supplementary materials for the talk "Introduction to the Omniverse and Isaac Platform for Robotics"

HDM-ext. HDM model loader for ComfyUI

HDM. Home Made Diffusion Models

M-YGO-Agent. AI-driven Yu-Gi-Oh! bot using deep reinforcement learning and LLMs

edopro. A script engine for "yu-gi-oh!" and sample gui (former "ygopro")

ygo-agent. AI-driven Yu-Gi-Oh! bot using deep reinforcement learning and LLMs

IGS. Image Gaussian Splatting

decord. An efficient video loader for deep learning with smart shuffling that's super easy to digest

sn-sm. Subset-Norm and Subset-Momentum. This repo is built on top of https://github.com/jiaweizzhao/GaLore.

flash-linear-attention. 🚀 Efficient implementations for emerging model architectures

shell. A fluid, morphing shell for your Linux desktop

TIPO-Agent. Python

overleaf. A web-based collaborative LaTeX editor

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

xet-core. xet client tech, used in huggingface_hub

KohakuRiver. A lightweight cluster manager that turns your small fleet of nodes into one powerful computer, using Docker for environment consistency without the overhead of enterprise orchestration systems.

astrowind. ⭕️ AstroWind: A free template using Astro v6 and Tailwind CSS v4. Astro starter theme.

ratarmount. Access large archives as a filesystem efficiently, e.g., TAR, RAR, ZIP, GZ, BZ2, XZ, ZSTD archives

unidisc. UniDisc: A discrete diffusion model for joint multimodal generation, enabling controllable and efficient text-image synthesis, editing, and inpainting.

promoplot. Corner plots -- now with ads!

cloudscraper. A Python module to bypass Cloudflare's anti-bot page.

Depth-Anything-V2. [NeurIPS 2024] Depth Anything V2. A More Capable Foundation Model for Monocular Depth Estimation

Depth-Anything. [CVPR 2024] Depth Anything: Unleashing the Power of Large-Scale Unlabeled Data. Foundation Model for Monocular Depth Estimation

FastSAM. Fast Segment Anything

KBLaM. Official Implementation of "KBLaM: Knowledge Base augmented Language Model"

Dzip-torch. Dzip: improved general-purpose lossless compression based on novel neural network modeling

ptx-tutorial-by-aislop. PTX-Tutorial Written Purely By AIs (Deep Research of Openai and Claude 3.7)

layered_vectorization. JavaScript

ComfyUI. The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface.

modelcontextprotocol. Specification and documentation for the Model Context Protocol

docs. Documentation for the Model Context Protocol (MCP)

thera. [TMLR 2025] Thera: Aliasing-Free Arbitrary-Scale Super-Resolution with Neural Heat Fields

imm. Official implementation of Inductive Moment Matching

slop. The place for SLOP

ideas.

SageAttention. [ICLR2025, ICML2025, NeurIPS2025 Spotlight] Quantized Attention achieves speedup of 2-5x compared to FlashAttention, without losing end-to-end metrics across language, image, and video models.

NotaGen. NotaGen: Advancing Musicality in Symbolic Music Generation with Large Language Model Training Paradigms

ml-gbc. Python

ai-deadlines. ⏰ AI conference deadline countdowns

UwUDiff. Codebase/framework for training Diffusion Model.

sd-scripts. Python

LoCon. LoRA for convolution layer

guanaco-lora. Instruct-tune LLaMA on consumer hardware

ml-mdm. Train high-quality text-to-image diffusion models in a data & compute efficient manner

native-sparse-attention-pytorch. Implementation of the sparse attention pattern proposed by the Deepseek team in their "Native Sparse Attention" paper

viztracer. A debugging and profiling tool that can trace and visualize python code execution

convnext_perceptual_loss. A PyTorch implementation of perceptual loss using ConvNeXt feature extractors.

Direct_Ascent_Synthesis. A demo for the Direct Ascent Synthesis: Hidden Generative Capabilities in Discriminative Models paper (https://arxiv.org/abs/2502.07753)

NeuralSVG. Official implementation of NerualSVG

HakuLatent. A comprehensive codebase for training and finetuning Image <> Latent models.

torrentp. Python torrent downloader - Download from torrent with .torrent file or magnet link, with just 3 lines of python code.

trackerslist. Updated list of public BitTorrent trackers

ktransformers. A Flexible Framework for Experiencing Heterogeneous LLM Inference/Fine-tune Optimizations

deep-research. An AI-powered research assistant that performs iterative, deep research on any topic by combining search engines, web scraping, and large language models. The goal of this repo is to provide the simplest implementation of a deep research agent - e.g. an agent that can refine its research direction overtime and deep dive into a topic.

triton-windows. Fork of the Triton language and compiler for Windows support and easy installation

funboost. pip install funboost，python全功能分布式函数调度框架,funboost的功能是全面性重量级，用户能想得到的功能99%全都有;funboost的使用方式是轻量级，只有@boost一行代码需要写。支持python所有类型的并发模式和一切知名消息队列中间件，支持如 celery dramatiq等框架整体作为funboost中间件，python函数加速器，框架包罗万象，用户能想到的控制功能全都有。一统编程思维，兼容50% python业务场景，适用范围广。只需要一行代码即可分布式执行python一切函数，99%用过funboost的pythoner 感受是　简易 方便 自由 强大，相见恨晚 。

Zero. Experience email the way you want with Mail0 – the first open source email app that puts your privacy and safety first. Join the discord: https://mail0.link/discord

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

ml-gsm-symbolic. GSM-Symbolic templates and generated data

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

rag-web-ui. RAG Web UI is an intelligent dialogue system based on RAG (Retrieval-Augmented Generation) technology.

RAGEN. RAGEN leverages reinforcement learning to train LLM reasoning agents in interactive, stochastic environments.

AnySchedule. A composer tool for hyperparemeter schedule in PyTorch or more.

otk-pyoptix. Complete Python bindings for the OptiX host API

Multi-Head-Latent-Attention-MLA-. working implimention of deepseek MLA

linux. Linux kernel source tree

torchft. Fault tolerance for PyTorch (HSDP, LocalSGD, DiLoCo, Streaming DiLoCo)

gtkwave. GTKWave is a fully featured GTK+ based wave viewer for Unix and Win32 which reads LXT, LXT2, VZT, FST, and GHW files as well as standard Verilog VCD/EVCD files and allows their viewing.

vivado-risc-v. Xilinx Vivado block designs for FPGA RISC-V SoC running Debian Linux distro

biriscv. 32-bit Superscalar RISC-V CPU

difflogic. A Library for Differentiable Logic Gate Networks

verilog-axi. Verilog AXI components for FPGA implementation

HakuTPU. An AI accelerator implementation with Xilinx FPGA

TRELLIS. Official repo for paper "Structured 3D Latents for Scalable and Versatile 3D Generation" (CVPR'25 Spotlight).

flash-attention. Fast and memory-efficient exact attention

Qart. Python

ASCII-generator. ASCII generator (image to text, image to image, video to video)

MinDiff. Some minimal implementation of some Diffusion Models. Try to use as less code and as simple arch as possible

paper-reviewer. Generate a comprehensive review from an arXiv paper, then turn it into a blog post. This project powers the website below for the HuggingFace's Daily Papers (https://huggingface.co/papers).

cursor.

Light-L16-Archive. Repository for archiving everything Light L16 related before it is all lost form the internet.

ComfyUI_frontend. Official front-end implementation of ComfyUI

core-to-core-latency. Measures the latency between CPU cores

Cello. Higher level programming in C

UwULLM. Codebase/framework for training Large Language Model.

SimpleTuner. A general fine-tuning kit geared toward image/video/audio diffusion models.

vllm. A high-throughput and memory-efficient inference and serving engine for LLMs

topic-modeling-parquet. topic modeling & inference scripts

PyQt-SiliconUI. A powerful and artistic UI library based on PyQt5，基于 PyQt5 的UI框架，灵动、优雅而轻便

Sketch-Gen. Understand Human Behavior to Align True Needs

GaussianImage. 🏠[ECCV 2024] GaussianImage: 1000 FPS Image Representation and Compression by 2D Gaussian Splatting

Paints-UNDO. Understand Human Behavior to Align True Needs

drum-game. Free rhythm game for electronic drummers

LyCORIS. Lora beYond Conventional methods, Other Rank adaptation Implementations for Stable diffusion.

RustPython. A Python Interpreter written in Rust

sdfx. The ultimate no-code platform to build and share AI apps with beautiful UI.

MyVectorizationWithSAM. Jupyter Notebook

cinny. Yet another matrix client

triton. Development repository for the Triton language and compiler

openui. OpenUI let's you describe UI using your imagination, then see it rendered live.

fish-speech. SOTA Open Source TTS

MiniGPT-4. Open-sourced codes for MiniGPT-4 and MiniGPT-v2 (https://minigpt-4.github.io, https://minigpt-v2.github.io/)

Semi-Auto-NovelAI-to-Pixiv. 带有 WebUI 的 NovelAI 量产工具, 实现了批量文生图; 批量图生图; 视频转绘; 分块重绘; 批量 Vibe; 批量局部重绘; 批量超分降噪; 批量自动打码; 批量添加水印; 批量上传 Pixiv; 图片筛选; 批量抹除, 还原或导出生成信息; 法术解析; 多模型反推提示词; ChatGPT; 动态加载插件; 自动 roll 画风串; 批量 Enhance; tag选择器; 涂鸦重绘; 图片压缩整理; 批量AI工具; wildcard; ...

BFSR. The official repository of BFSR: "Boosting Flow-based Generative Super-Resolution Models via Learned Prior" [CVPR 2024]

KGen. A system for Prompt generation to improve Text-to-Image performance.

DiffSynth-Studio. Enjoy the magic of Diffusion models!

sd-webui-infotext-example. This is an example of how to read and write Infotext / PNG Info when developing an extension for Stable Diffusion web UI

Content-Adaptive-Image-Downsampling. This contains Matlab implementation of Johannes Kopf's image processing paper which deals with the adaptive downsampling of images. It gives better results than existing methods like subsampling and bicubic.

PixelOE. Detail-Oriented Pixelization based on Contrast-Aware Outline Expansion.

sd-mecha. Executable State Dict Recipes

z-tipo-extension. A sd-webui extension for utilizing DanTagGen to "upsample prompts".

ComfyUI_DanTagGen. ComfyUI node of DTG

grok-1. Grok open release

regional-attn. Python

oci-arm-host-capacity. This script allows to bypass Oracle Cloud Infrastructure 'Out of host capacity' error immediately when additional OCI capacity will appear in your Home Region / Availability domain.

GaLore. GaLore: Memory-Efficient LLM Training by Gradient Low-Rank Projection

llama.cpp. LLM inference in C/C++

stable-diffusion-webui-forge. Python

ComfyUI_Comfyroll_CustomNodes. Custom nodes for SDXL and SD1.5 including Multi-ControlNet, LoRA, Aspect Ratio, Process Switches, and many more nodes.

HakuBooru. text-image dataset maker for anime-style images

PalWorld-Server-Unoffical-Fix. This is an unofficial palworld server binary distribution project that fixes some problems with the original server.

multidiffusion-upscaler-for-automatic1111. Tiled Diffusion and VAE optimize, licensed under CC BY-NC-SA 4.0

anime_screenshot_pipeline. A 99% automatized pipeline to construct training set from anime and more for text-to-image model training

UniDiffusion. A Diffusion training toolbox based on diffusers and existing SOTA methods, including Dreambooth, Texual Inversion, LoRA, Custom Diffusion, XTI, ....

TKinterModernThemes. A collection of modern themes with code that makes it easy to integrate into a tkinter project.

icetk. A unified tokenization tool for Images, Chinese and English.

Kohaku-NAI. A Novel-AI client with more utilities built in it

oneflow. OneFlow is a deep learning framework designed to be user-friendly, scalable and efficient.

textgen. Open-source desktop app for local LLMs. Text, vision, tool-calling, OpenAI/Anthropic-compatible API. 100% private.

exllama. A more memory-efficient rewrite of the HF transformers implementation of Llama for use with quantized weights.

kohya-trainer. Adapted from https://note.com/kohya_ss/n/nbf7ce8d80f29 for easier cloning

bitsandbytes-windows-webui. Windows compile of bitsandbytes for use in text-generation-webui.

bitsandbytes. Accessible large language models via k-bit quantization for PyTorch.

tiny-stable-diffusion. Tiny optimized Stable-diffusion that can run on GPUs with just 1GB of VRAM. (Beta)

PTQD. The official implementation of PTQD: Accurate Post-Training Quantization for Diffusion Models

HyperKohaku. A diffusers based implementation of HyperDreamBooth

FLsun-V400. Everything you need for quality prints

vuera. :eyes: Vue in React, React in Vue. Seamless integration of the two. :dancers:

million. Optimizing compiler for React

IF. Python

AutoGPT. AutoGPT is the vision of accessible AI for everyone, to use and to build on. Our mission is to provide the tools, so that you can focus on what matters.

HCP-Diffusion. A universal Stable-Diffusion toolbox

sdnext. SD.Next: All-in-one WebUI for AI generative image and video creation, captioning and processing

arozos. Web Desktop Operating System for low power platforms, Now written in Go!

a1111-sd-webui-lycoris. An extension for stable-diffusion-webui to load lycoris models.

segment-anything. The repository provides code for running inference with the SegmentAnything Model (SAM), links for downloading the trained model checkpoints, and example notebooks that show how to use the model.

losalina. Hypernetworks for kohya's sd-scripts

whisper_real_time. Real time transcription with OpenAI Whisper.

BigKnow2022. BigKnow2022: Bringing Language Models Up to Speed

transformers. 🤗 Transformers: the model-definition framework for state-of-the-art machine learning models in text, vision, audio, and multimodal models, for both inference and training.

alpaca-lora. Instruct-tune LLaMA on consumer hardware

a1111-sd-webui-locon. A extension for loading LyCORIS model in sd-webui

aicommit. An AI-powered git commit message generator written in python.

ControlNet. Let us control diffusion models!

SW-CV-ModelZoo. Repo for my Tensorflow/Keras CV experiments. Mostly revolving around the Danbooru20xx dataset

gradio. Build and share delightful machine learning apps, all in Python. 🌟 Star to support our work!

a1111-sd-webui-haku-img. An Image utils extension for A1111's sd-webui

lsun. LSUN Dataset Documentation and Demo Code

TTUR. Two time-scale update rule for training GANs

pytorch. Tensors and Dynamic neural networks in Python with strong GPU acceleration

stable-diffusion-webui. Stable Diffusion web UI

verilog-mersenne. Verilog implementation of Mersenne Twister PRNG

verilog-uart. Verilog UART

cpython. The Python programming language

YaneuraOu. YaneuraOu is the World's Strongest Shogi engine(AI player) , WCSC29 1st winner , educational and USI compliant engine.

lima. Linux virtual machines, with a focus on running containers

flask. The Python micro framework for building web applications.