zhaobin Chu

@hadipash hello，I tried using LoRA to fine-tune the U-Net with SVD, and even with a batch size of 1, memory overflow occurs on the A100 GPU when the dataset consists...

I also tried training with LoRA, and even with a batch size of 1, there was still memory overflow. It seems that model parallel training is needed.

我测试用的是lora微调，使用部分模块是可以运行的，但微调后效果没出来；分辨率如果调低训练lora，在推理的时候，生成效果是不是就不太行了？

A100 80G

> 我之前试过在低分辨率上微调，高分辨率上推理，没有发现效果有明显下降。不过没有进行规模化的实验验证。 是motion lora么？我这边找了30张图片，然后规则做的zoom，微调的，batch_size=1就崩了，都还没用Classifier-free guidance. 哈哈

将一张图片用规则做成zoom in/zoom out的视频进行motion lora微调的

I am experimenting with fine-tuning motion LoRA and need to generate videos at a resolution of 1024x576. Do you mean that training motion LoRA at a lower resolution can achieve...

https://github.com/alibaba/animate-anything/blob/main/train_svd.py