<h2 id="qwen-3_4_b-trained-text-encoder-for-z-image">Qwen 3_4_B Trained Text Encoder for Z-Image</h2><p></p><h3 id="fp32">FP32</h3><ul><li><p>Full Finetune at FP32 (Full Model Finetune - All Parameters &amp; All layers)</p></li><li><p>FP32 Finetune of QWEN3_4b focusing on describing human features SFW/NSFW captions.</p></li><li><p>Can be run in FP32 with no time loss on most machines that use CPU offloading.</p></li></ul><h3 id="bf16">BF16</h3><ul><li><p>Full Finetune at BF16 (20 Layers)</p></li><li><p>Long Text descriptions 500-1000 token length focusing on describing human features.</p></li><li><p>For use with Z-Image or Z-Image Turbo</p></li></ul><p></p><hr /><p></p><ul><li><p>Comparison Images showing QWEN base VS Human Corpus <a target="_blank" rel="ugc" href="https://civarchive.com/articles/26226/qwen-for-z-image-in-fp32">HERE</a></p></li></ul><p></p><p></p>

Files with SHA256: 43c96d1b7dbb218c885c9a65afd94b91f61bc553a91cdf64fff38ddf4346a986