2026-10-04 · America/Los_Angeles · 社区动态 · #13

I made a free all-in-one LoRA trainer for consumer GPUs (Windows + Linux): Qwen-Image 2.1, FLUX.2 Klein 9B, Krea 2, Z-Image, Ideogram 4, Anima, SDXL/Pony/Illustrious, LTX 2.3 and MiniMax-H3 (video + audio) — from 4-8 GB VRAM

Hi everyone! I've been working on **AcademiaSD LoRAlab Trainer Studio**: one installer, one launcher and 9 trainers that share the same web interface. Every model is loaded in **4-bit NF4**, the text encoder and VAE run only once in a pre-cache stage, and the whole GPU goes to training. **Platform:** made for **Windows** (one-click `.bat` installer), with **Linux support just added** (`.sh` scripts, see `docs/Linux.md`). NVIDIA GPU, RTX 20xx / GTX 16xx or newer. **GitHub:** https://github.com/AcademiaSD/AcademiaSD_LoRAlab-TrainerStudio ## What it trains | Trainer | What you get | Min. VRAM |…

热度 38.4 / 100；排名与评分保留该期记录。