Model description
If you want to use fluxgym, you need to install python 3.10. If you are using RTX 50 series GPU, please install CUDA128 (Our GPU is RTX 5070Ti).
Please close all important process and let GPU work for training.
Please know python programming language and you may read all errors prompts.
Please use flux1-dev-fp8 model if you only have 16G VRAM or lower.
Please write less prompts to accelerate training and get less errors(more and many prompts for training images will get UTF-8 errors).
If you get new trainied lora, you need to work with workflow and flux1-dev-fp8 and flux1-dev-dev2pro models. We will upload wine bottles models and white-liquor-bottle models based on our business. We trained eight bottle images in this model.
如果你想安装fluxgym来训练新的Flux Loras,那么你需要安装CUDA128和python 3.10版本,其他版本目前感觉兼容性不太行。请学习python编程语言来帮助阅读所有的报错信息。
如果你已经开始使用,需要关闭所有的后台进程来加速GPU的训练,这就是专业显卡不能打游戏的一个原因,把图像处理的资源全部用来做训练,没有设计游戏驱动。
如果你的显卡只有16G及以下,请使用flux1-dev-fp8这个模型,亲测没有问题。每张图片的训练提示词用英文,你可以翻译一下嘛,抓住重点细节,太长直接报错,UTF-8 errors,其实就是提示词太长。提示词太短了,我感觉训练意义也不大。
如果你得到一个LORA,请配合工作流使用,我们已经上传。以及flux1-dev-fp8 和 flux1-dev-dev2pro两个模型最合适。
我们目前会上传一些酒瓶训练模型,这是基于我们的业务,我们感觉目前并没有太多的模型可以借鉴,大家可以看一下我们的模型。传统的白酒瓶和具有现代风格的新酒瓶我们都会分开训练。这个模型基于八张酒瓶图像生成。
Trigger words
You should use white-liquor-bottle to trigger the image generation.
Download model
Download them in the Files & versions tab.