search

invoke-ai / InvokeAI

Invoke is a leading creative engine for Stable Diffusion models, empowering professionals, artists, and enthusiasts to generate and create visual media using the latest AI-driven technologies. The solution offers an industry leading WebUI, and serves as the foundation for multiple commercial products.
https://invoke-ai.github.io/InvokeAI/
Apache License 2.0
23.78k stars 2.44k forks source link

[enhancement]: need to support fp8 #7148

Open licyk opened 1 month ago

licyk commented 1 month ago

Is there an existing issue for this?

Contact Details

licyk233@gmail.com

What should this feature add?

Need to support fp8 to reduce vram usage.

My graphics card is 4060 8GB. When I use the SDXL model and controlnet openpose in invokeai, I encounter a "torch.cuda.OutOfMemoryError" error.

However, in the stable diffusion webui, I can reduce the video memory usage by enabling fp8, and using the SDXL model and controlnet openpose, so I can generate images normally without any "torch.cuda.OutOfMemoryError" issues.

Alternatives

No response

Additional Content

No response

licyk commented 1 month ago

This is the fp8 implementation in stable diffusion webui: https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14031