Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

svd的微调脚本可以用在25-frames的视频数据上么? #62

Open
DataAIPlayer opened this issue May 8, 2024 · 9 comments
Open

Comments

@DataAIPlayer
Copy link

我在用Accelerator的deepspeed做u-net微调时,即使batch_size=1,仍会出现显存溢出

@sculmh
Copy link
Collaborator

sculmh commented May 9, 2024

你可以尝试将训练的分辨率调低、trainable_modules只微调一部分模块而不是'all'、cache_latents设为True等方法降低显存

@DataAIPlayer
Copy link
Author

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?

@sculmh
Copy link
Collaborator

sculmh commented May 10, 2024

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。

@xiangweifeng
Copy link

我在用Accelerator的deepspeed做u-net微调时,即使batch_size=1,仍会出现显存溢出

多大显存啊

@DataAIPlayer
Copy link
Author

A100 80G

@DataAIPlayer
Copy link
Author

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

@sculmh
Copy link
Collaborator

sculmh commented May 10, 2024

不是lora。SVD这部分工作我参与不多,不是很熟。

@sculmh
Copy link
Collaborator

sculmh commented May 10, 2024

你用的图片还是视频微调motion?

@DataAIPlayer
Copy link
Author

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

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants