Kohya 发布混元视频模型 Lora 训练脚本
据X用户 TTPlanet 报道,该脚本将需要 16GB 显存,仅使用图像输入来训练视频 Lora。
@Kohya released the training code for Hunyuan Video. It will allow 16GB VRAM Graphic card to train video Lora with image input only.
Due to there is no GUI, and with a lot of job to setup. I have made a package with all repo, models, python and GUI included.
You can find it here http://pan.baidu.com/s/1EzlY728byW5…
please let me know which platform you can quickly download. I will upload to it. The password for the file is 汤团猪TTPlanet just copy to unzip by 7z application.
该用户还制作了一个包含所有 repo、模型、python 和 GUI 的包。
混元视频模型(HunyuanVideo) 是腾讯发布的开源文生视频模型。