[Feature Request] Is there any plan for stable diffusion running under this project?

[felixslu]

🚀 Feature

Support stable diffusion models!
SnapFusion models can run in the phone within 2 seconds!
https://arxiv.org/pdf/2306.00980.pdf

Motivation

Text-to-image diffusion models can create stunning images from natural language descriptions. It is an important node in large model ecology！

Alternatives

should we use weapon of tvm unity to reach the deployment of stable diffusion model on any devices with gpu?

Additional context

[yzh119]

Thanks for suggesting the paper, we'll read it and see how can we benefit from SnapFusion.

Yes, TVM Unity could be used to deploy Stable-Diffusion models, we already have a web-stable-diffusion repo which uses TVM Unity for SD models on web browsers.

[felixslu]

Thanks for your advise! I will try this repo.(web-stable-diffusion)

As I known,3-4 bit quantization technology (such as gptq) have not been used in the web-stable-diffusion project , maybe it could benefit from the optimization methods in mlc-llm project ? (Maybe your team could support a quantization version for SD to accelerate infer cost time. )

[felixslu]

Hi,MLC team! I have try to run this repo ,but failed!(web-stable-diffusion).

Firstly, I try to use pre-auto-tuned schedule params in directory of log_db ,but get errors. "https://github.com/mlc-ai/web-stable-diffusion/issues/38".

Then, I try to do auto-tuning by myself, but get other errors. "Did you forget to bind? RuntimeError: Memory verification failed with the following errors".

Can I get any furture help from your team? Thanks!

[felixslu]

hi,MLC Team!
Could you have time to update the stable diffusion log_db ? (web-stable-diffusion)
Because, we can not get auto-tunning params for newly introduced purity flag. We get error below when do "ms.database.create" ops.

/workspace/tvm/src/node/serialization.cc:375: JSONReader: cannot find field purity
#407

[Jack47]

is there any update? it would be really awesome if stable diffusion can be run on mobile phones

[Fcucgvhhhvjv]

is there any update? it would be really awesome if stable diffusion can be run on mobile phones

https://github.com/ZTMIDGO/Android-Stable-diffusion-ONNX
here