<div data-youtube-video><iframe width="640" height="480" allowfullscreen="true" autoplay="false" disablekbcontrols="false" enableiframeapi="false" endtime="0" ivloadpolicy="0" loop="false" modestbranding="false" origin playlist src="https://www.youtube.com/embed/QQ3RaRDW344" start="0"></iframe></div>V1 evaluation lora used to create a baseline. Training settings were refined for the next versions; V2 - ADAFACTOR, Rank 16, Alpha 1 V3 - Rank 32, Alpha 1 v3b - rank 32, alpha 8 v3c - rank 32, alpha 4 V4 - were all trained with ADAFACTOR using revised settings. Rank 64, Alpha 1 Round 2V5 - Trained with ADAFACTOR with final settings, Rank 32, Alpha 1, reduced learning rate, LR 2e3Switched to PRODIGYV6-early = ADAFACTOR, Rank 32, Alpha 1 V6 = ADAFACTOR, Rank 32, Alpha 1offset noise 0.06 V7-early = PRODIGY, Rank 64, Alpha 1 V7 = PRODIGY, Rank 64, Alpha 1initialD 1e-06V8-early = PRODIGY, Rank 64, Alpha 1 V8-mid = PRODIGY, Rank 64, Alpha 1V8 = PRODIGY, Rank 64, Alpha 1 as shown in the video, early Epochs seem to perform better beyond txt2img. Study presentation: <a target="_blank" rel="ugc" href="https://www.figma.com/file/NM9dSIwcKoyZpOSYqi4pKs/OneTrainer-Stable-Cascade-(LR-adjustment)">https://www.figma.com/file/NM9dSIwcKoyZpOSYqi4pKs/OneTrainer-Stable-Cascade-(LR-adjustment)</a> Models: <a target="_blank" rel="ugc" href="https://civarchive.com/models/320332">https://civarchive.com/models/320332</a> Dataset: <a target="_blank" rel="ugc" href="https://github.com/MushroomFleet/assassinKahb-1024">https://github.com/MushroomFleet/assassinKahb-1024</a> Article: Training Stable Cascade with OneTrainer OneTrainer: <a target="_blank" rel="ugc" href="https://github.com/Nerogar/OneTrainer">https://github.com/Nerogar/OneTrainer</a> my preset for OneTrainer: <a target="_blank" rel="ugc" href="https://pastebin.com/tLri1HdU">https://pastebin.com/tLri1HdU</a> .json file, place inside OneTrainer\training_presets which are trained to compare the Rank setting.<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/4ba182f3-50c1-4cb7-be0f-b5f6c30037b5/width=525/4ba182f3-50c1-4cb7-be0f-b5f6c30037b5.jpeg" /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/d5ec7e85-bc7b-4fe7-b00f-a5d16d0cd675/width=525/d5ec7e85-bc7b-4fe7-b00f-a5d16d0cd675.jpeg" /><h3 id="heading-7">test prompt:</h3><ul><li>AssassinKahb style a demonic looking skeleton holding a sword with red hair</li></ul><a target="_blank" rel="ugc" href="https://civarchive.com/models/310409?modelVersionId=359739">txt2img-lora: workflow</a> <img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/bb0227e0-359f-42a7-af7a-a9b0d4dff6e6/width=525/bb0227e0-359f-42a7-af7a-a9b0d4dff6e6.jpeg" /><a target="_blank" rel="ugc" href="https://civarchive.com/models/310409?modelVersionId=359520">img2img-lora: workflow</a><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/6d42f571-ee4d-4243-a2ec-7288bca2af13/width=525/6d42f571-ee4d-4243-a2ec-7288bca2af13.jpeg" /><a target="_blank" rel="ugc" href="https://civarchive.com/models/310409?modelVersionId=359663">txt2img-vision-lora: workflow</a><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/1aa772f2-a3d7-44fa-9ba4-c93354893f81/width=525/1aa772f2-a3d7-44fa-9ba4-c93354893f81.jpeg" /><a target="_blank" rel="ugc" href="https://civarchive.com/models/310409?modelVersionId=359716">txt2img-remix-lora: workflow</a><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/bb76f8a6-4568-4047-819f-2543ca6ade3e/width=525/bb76f8a6-4568-4047-819f-2543ca6ade3e.jpeg" />At the time of posting the Lora implementation is incomplete in most webUI's You will be able to use this if you have updated your ComfyUI to the latest version.

DJZ AssassinKahb [Cascade Test Lora]

same test model with relaxed training settings.

Files with SHA256: 0ee0ab27b9e8731d8678e17f93fa28f18376f87f786e952160fed63e11b5d3b7