<h1 id="qwen-2.1-finalcut-(6.6gb)-25-step">Qwen 2.1 FinalCut (6.6GB) 25 Step</h1>After 20 failed trainings including a training that took 4 A100 80GB a few days I finally have a version I am happy to release.<ul><li>25 Steps</li><li>Reduced "QWEN Canvas" issue.</li><li>Works with most systems without issues and has lora support.</li></ul>Note: I settled on group size 32 and standard qwen exclusions for INT8I found the single DiT blocks of this model way worse to train then the double/single Krea2 - I personally do not like this model to train, and for use I would rate it above SDXL and Flux2. It handles text placement and guidance very well, but seems to fight new concepts.

Qwen 2.1 FinalCut (3.8-6.6GB) 25 Step

Files with SHA256: eb07113d11f8a95aefa290fff774030cfc53559cea0ef5797a14275527e3d9f2