这个Lora我使用了正则化训练，避免了权重过溢问题，可以使用大于1以上的权重，泛用性极强，它适用于所有大模型。使用建议：1. CLIP终止层：22.权重范围：0.7 至 2，可以使用正负权重，建议使用0.7以上的权重，1.3以上画风会突变，朝着狂野迸发！示例图我分别用了两个大模型进行演示，权重依次是：0.8，1，1.4推荐使用“<a target="_blank" rel="ugc" href="https://www.liblib.ai/modelinfo/8c8f1c7f64d943ebb6b57833d5d042d2">lbc_Simple_简约</a>”或“<a target="_blank" rel="ugc" href="https://www.liblib.ai/modelinfo/76b421c56a3b43a8a300d01e040e6c37">lbc_Fantasy_幻想_2.7D</a>”大模型效果更佳。如果大家喜欢我训练的模型风格，还请多多支持！小弟感激不尽，如有想要训练的画风或角色也可以留言在下方，我可以免费为大家训练，一起分享美好事物！

惣流·明日香·兰格雷

lbc_souryuu_asuka.safetensors

v1.0

Rei Ayanami Langley

I used regularization training for this Lora to avoid weight overflow issues, allowing weights greater than 1, highly versatile, suitable for all large models.Recommendations:1. CLIP termination layer: 22. Weight range: 0.7 to 2, positive and negative weights can be used, recommended to use weights above 0.7, styles will drastically change above 1.3, bursting into a wild and explosive direction!In the example images, I used two large models for demonstration with weights of 0.8, 1, 1.4 respectively.Recommended to use "<a href="https://www.liblib.ai/modelinfo/8c8f1c7f64d943ebb6b57833d5d042d2" target="_blank">lbc_Simple_Concise</a>" or "<a href="https://www.liblib.ai/modelinfo/76b421c56a3b43a8a300d01e040e6c37" target="_blank">lbc_Fantasy_2.7D</a>" for better results with large models.If you like the style of the models I trained, please show your support! I am very grateful. If you have any styles or characters you want to train, feel free to leave a message below, I can train them for you for free, let's share beautiful things together!For those interested in downloading other models that are not suitable for publication in China, please go to the C site for support: https://civitai.com/user/fuaneng/models

Files with SHA256: 537fdd5b5f10817230d6641cdabb72c45a0299a4df4494eee78bd972443765cb