<h1 id="moss-4-step-v2.0-qkv-forge-neo-fix">MOSS 4-Step v2.0 — QKV Forge Neo Fix</h1>4-Step Turbo • QKV Fusion • Low VRAM FriendlyA compatibility-optimized version of MOSS-Image-SFT 4-Step v2.0, adapted for Forge Neo.This experimental conversion fixes the separate Q/K/V LoRA projections by combining them into fused QKV layers, allowing Forge Neo to recognize all 168 LoRA modules instead of only 126.<h3 id="features">Features</h3><ul><li>4-step fast generation</li><li>6 steps also tested with excellent visual results</li><li>168/168 LoRA modules recognized — 0 skipped</li><li>Improved QKV compatibility with Forge Neo</li><li>Excellent photorealism, cinematic lighting and detailed textures</li><li>Impressive text rendering and typography</li><li>Tested on RTX 3060 6GB VRAM</li><li>Compatible with MOSS-Image SFT INT8</li></ul><h3 id="recommended-settings">Recommended Settings</h3>SettingValueSamplerEulerSteps4 (6 optional)CFG1.0LoRA Strength1.0Base ModelMOSS-Image SFT<h3 id="credits-and-license">Credits &amp; License</h3>Original 4-Step LoRA by akhaliq.Based on MOSS-Image SFT by OpenMOSS-Team, with upstream DMD contributions from lodestones.This is an experimental QKV compatibility conversion, not a newly trained LoRA.Licensed under Apache 2.0. Original authors retain their respective rights.Original project: <a target="_blank" rel="ugc" href="https://huggingface.co/akhaliq/MOSS-Image-SFT-4-steps">https://huggingface.co/akhaliq/MOSS-Image-SFT-4-steps</a>QKV conversion for Forge Neo by VINCE.If you enjoy this work, consider supporting future projects with a coffee or Civitai Buzz! ☕

Files with SHA256: 90a5cbed73b98d769fdb6694d7beaa6c95f05d21ce42de8bb5d3e643b00998ae