标签：LoRA

把 100+ LLM/VLM 的 SFT、DPO、PPO 收敛到一套配置和 CLI 上，省掉每个模型重写训练脚本

把论文公式翻译成可逐行对照的 PyTorch 代码，用来读算法而不是直接上生产

用磁盘换显存，让单卡 4GB 跑 70B 的分层流式推理库

按模型分章节的国产开源大模型部署与微调操作手册，不是可安装的库

在 Transformers 模型上注入 LoRA 等适配器，只训练千分之一参数并单独保存适配器权重

在 LLaMA 上扩中文词表再预训练的中文基座，产出的是 LoRA 增量而不是可直接加载的模型

一个仓库覆盖 600+ LLM 与 400+ 多模态模型的 SFT/DPO/GRPO 训练到量化部署，省掉自己拼 PEFT+并行+推理后端

在免费 Colab GPU 上跑起 Stable Diffusion WebUI 的笔记本集合，省掉本地显卡和装环境

给多步 Agent 做 GRPO 在线强化学习，用 LoRA 热加载把训练结果直接送回推理端

一份 YAML 加层流式，让 4 GB 笔记本显卡也能跑 8B 模型的 LoRA 微调

把 SFT/DPO/GRPO 训练、评测、推理和上云打包成 recipe 配置的开源训练编排层

LoRa 硬件上的去中心化网状网固件，替代自建点对点无线链路和依赖基站的通信方案

在 LoRa 和分组无线电上跑的非 IP 加密网络栈，顶替 Meshtastic 那类固件里的传输层

把 kohya-ss sd-scripts 那堆命令行参数收进 GUI 的本地 LoRA/Dreambooth 训练台

High Performance Open Source Radio Control Link

The open source research environment for AI researchers to seamlessly train, evaluate, and scale models from local hardware to GPU clusters.

MQTT gateway for ESP8266 or ESP32 with bidirectional 433mhz/315mhz/868mhz, Infrared communications, BLE, Bluetooth, beacons detection, mi flora, mi jia, LYWSD02, LYWSD03MMC, Mi Scale, TPMS, BBQ thermometer compatibility & LoRa.

[ICLR2025 Spotlight] SVDQuant: Absorbing Outliers by Low-Rank Components for 4-Bit Diffusion Models

Multi-LoRA inference server that scales to 1000s of fine-tuned LLMs

🦞 Just talk to your agent — it learns and EVOLVES 🧬.

Generate, animate and schedule your AI characters 🤖

OneTrainer is a one-stop solution for all your Diffusion training needs.

LLM Finetuning with peft

PJON (Padded Jittering Operative Network) is an experimental, arduino-compatible, multi-master, multi-media network protocol.

A Unified Library for Parameter-Efficient and Modular Transfer Learning

Build, personalize and control your own LLMs. From data pre-processing to fine-tuning, xTuring provides an easy way to personalize open-source LLMs. Join our discord community: https://discord.gg/TgHXuSJEk6

IoT Platform Framework

Communicate Freely

Wifi & BLE driven passenger flow metering with cheap ESP32 boards

OneDiff: An out-of-the-box acceleration library for diffusion models.

Android application for Meshtastic

Large-scale LLM inference engine

LXMF client for Android, Linux and macOS allowing you to communicate with people or LXMF-compatible systems over Reticulum networks using LoRa, Packet Radio, WiFi, I2P, or anything else Reticulum supports.

Fine-tune LLMs on your Mac with Apple Silicon. SFT, DPO, GRPO, Vision, TTS, STT, Embedding, and OCR fine-tuning — natively on MLX. Unsloth-compatible API.

Research of DeepSeek Engram Architecture based on Qwen-3 and Stable Diffusion series.

Lua RTOS for ESP32

Hypernetworks that adapt LLMs for specific benchmark tasks using only textual task description as the input

A simple mesh network communications app powered by the Reticulum Network Stack.

Native Android messaging app using Bluetooth LE, TCP, or RNode (LoRa) over LXMF and Reticulum

LLM model quantization (compression) toolkit with HW acceleration support for Nvidia (CUDA), AMD (ROCm), Intel GPU (XPU), Huawei Ascent (NPU), and Intel/AMD/Apple CPU via HF, vLLM, and SGLang.

Fine-tune the Whisper speech recognition model to support training without timestamp data, training with timestamp data, and training without speech data. Accelerate inference and support Web deployment, Windows desktop deployment, and Android deployment

The Things Stack, an Open Source LoRaWAN Network Server

A foundation model that generates synchronized video and audio in a single model

ChirpStack open-source LoRaWAN Network Server

从 NLP 到 LLM 的算法全栈教程，在线阅读地址：https://datawhalechina.github.io/base-llm/

AI-powered multi-voice audiobook generator — LLM script annotation, voice cloning, voice design, LoRA training, per-line style control, and export to MP3, chaptered M4B, or Audacity multi-track. Built on Qwen3-TTS.

A Framework for Speech, Language, Audio, Music Processing with Large Language Model

:airplane: Multi-functional, compatible DIY general aviation proximity awareness system

This is the fully-functional GNU Radio software-defined radio (SDR) implementation of a LoRa transceiver with all the necessary receiver components to operate correctly even at very low SNRs. This work has been conducted at the Telecommunication Circuits Laboratory, EPFL.

[ICML2024 (Oral)] Official PyTorch implementation of DoRA: Weight-Decomposed Low-Rank Adaptation