基本参考这篇文章来训练的：<a target="_blank" rel="ugc" href="https://civarchive.com/articles/7097/flux-complete-lora-settings-and-dataset-guide-post-mortem-of-two-weeks-of-learning">https://civarchive.com/articles/7097/flux-complete-lora-settings-and-dataset-guide-post-mortem-of-two-weeks-of-learning</a>使用了40张图片来训练，2.5e-5的学习率训练了8000步左右，DIM 32，Alpha 16.用joycaption自然语言打标加上关键触发词：Gwenfl6我感觉Flux的动漫风格基础比较原始，尽量配合其他优秀的Style lora使用会更好 我觉得我训练Lora的方法还有提升的空间，如果有建议请告诉我 Basically refer to this article for training: <a target="_blank" rel="ugc" href="https://civarchive.com/articles/7097/flux-complete-lora-settings-and-dataset-guide-post-mortem-of-two-weeks-of-learning">https://civarchive.com/articles/7097/flux-complete-lora-settings-and-dataset-guide-post-mortem-of-two-weeks-of-learning</a>40 images were used for training, and the learning rate of 2.5e-5 was trained for about 8,000 steps, DIM 32, Alpha 16. Joycaption natural language labeling was used, and the key trigger word was added: Gwenfl6I feel that the anime style of Flux is relatively primitive, and it would be better to use it with other excellent Style lora I think there is room for improvement in my training method for Lora. If you have any suggestions, please let me know.

Files with SHA256: f29d70144c680ed1cabacfa374b009e2ac17dfca32844985a463a4e1cce2beed