啥都融进去了，用了5.3万张图耗费3天使用A100显卡炼制基于sd1.5炼制，所有图像经人工筛选，大模型炼制感觉没有必要，微调能力不如lora，如果要炼建议学习率搞大一点，几个回合出效果，不要打持久战based on SD1.5，use 53000 high qulity picture to train this model两轮保存一次模型（every 2 epoch saved modles）<a target="_blank" rel="ugc" href="https://huggingface.co/HOPStudio/Train_Everything-V0">HOPStudio/Train_Everything-V0 · Hugging Face</a>感谢internstudio提供显卡支持，欢迎关注报名书生·浦语大模型实战营，学习大语言模型的底层与部署，课程全免费，详见<a target="_blank" rel="ugc" href="https://github.com/InternLM/Tutorial">InternLM/Tutorial: LLM&amp;VLM Tutorial (</a><a target="_blank" rel="ugc" href="http://github.com">github.com</a><a target="_blank" rel="ugc" href="https://github.com/InternLM/Tutorial">)</a>。If you want to learn LLM for free，looking <a target="_blank" rel="ugc" href="https://github.com/InternLM/Tutorial">InternLM/Tutorial: LLM&amp;VLM Tutorial (</a><a target="_blank" rel="ugc" href="http://github.com">github.com</a><a target="_blank" rel="ugc" href="https://github.com/InternLM/Tutorial">) </a>for more information.

Files with SHA256: 012a7d05bbe0c4a24f83bce0e79fb0ae67d4575b964d0cf2a26bdf1c9d85305f