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

Could x-flux run on RTX 3090? #115

Open
xingyouxin opened this issue Sep 24, 2024 · 4 comments
Open

Could x-flux run on RTX 3090? #115

xingyouxin opened this issue Sep 24, 2024 · 4 comments

Comments

@xingyouxin
Copy link

I try to train x-flux with LoRA on RTX 3090 GPU. But it seems that out of memory.

@LixDemon
Copy link

I met this error on 3090 GPU.
NotImplementedError: Cannot copy out of meta tensor; no data! Please use torch.nn.Module.to_empty() instead of torch.nn.Module.to() when moving module from meta to a different device.

@xingyouxin
Copy link
Author

I met this error on 3090 GPU. NotImplementedError: Cannot copy out of meta tensor; no data! Please use torch.nn.Module.to_empty() instead of torch.nn.Module.to() when moving module from meta to a different device.

Yes, so do I when using the flux-dev-fp8.safetensors (11 GB). I also tried flux1-dev.safetensors (22 GB), resulting in a memory error.

@LixDemon
Copy link

I met this error on 3090 GPU. NotImplementedError: Cannot copy out of meta tensor; no data! Please use torch.nn.Module.to_empty() instead of torch.nn.Module.to() when moving module from meta to a different device.

Yes, so do I when using the flux-dev-fp8.safetensors (11 GB). I also tried flux1-dev.safetensors (22 GB), resulting in a memory error.

Have you tried sd_scripts for FLUX Lora training? https://github.com/kohya-ss/sd-scripts/tree/99744af53afcb750b9a64b7efafe51f3f0da8826 I see it works with 24GB VRAM GPUs.

@xingyouxin
Copy link
Author

I met this error on 3090 GPU. NotImplementedError: Cannot copy out of meta tensor; no data! Please use torch.nn.Module.to_empty() instead of torch.nn.Module.to() when moving module from meta to a different device.

Yes, so do I when using the flux-dev-fp8.safetensors (11 GB). I also tried flux1-dev.safetensors (22 GB), resulting in a memory error.

Have you tried sd_scripts for FLUX Lora training? https://github.com/kohya-ss/sd-scripts/tree/99744af53afcb750b9a64b7efafe51f3f0da8826 I see it works with 24GB VRAM GPUs.

Sorry, I didn't. But I have tried ai-toolkit for Flux LoRA training, and tested on web-ui successfully. https://github.com/ostris/ai-toolkit/ Now, I am testing 青龙圣者(Qinglongshengzhe)'s framework. If you are interested in it, don't hesitate to get in touch with me at QQ (593851428). https://www.bilibili.com/video/BV1RW421977P/?spm_id_from=333.788&vd_source=85be3ec5e95fd23384a835c58edf1296 We can do a great job together.

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

2 participants