ray icon indicating copy to clipboard operation
ray copied to clipboard

[Serve] Add inference serve example to run Stable Diffusion Inference using AWS Inferentia2

Open ratnopam opened this issue 1 year ago • 9 comments

Description

Add example of how to run inference using RayServe for Stable Diffusion model using AWS Inferentia2. The example will showcase how to leverage AWS Inf2 instances to set up neuron libraries and use RayServe to run inference on a neuron compatible Stable Diffusion model.

Link

Add example for Stable Diffusion based inference similar to aws-neuron BERT-base example

ratnopam avatar Feb 06 '24 21:02 ratnopam

You can assign this issue to me.

ratnopam avatar Feb 06 '24 22:02 ratnopam

Sg, thanks @ratnopamc Feel free to let us know when the PR is ready or if you runs into any issue

GeneDer avatar Feb 07 '24 00:02 GeneDer

@GeneDer , I've raised a PR https://github.com/ray-project/ray/issues/43018 to store the output image. The github url of the image is referenced in the docs. Can you please merge the above PR? Once it's merged, I can push the PR for the doc changes. Thanks.

ratnopam avatar Feb 07 '24 20:02 ratnopam

@ratnopamc I think you are referring to this PR? https://github.com/ray-project/ray/pull/43046 I see that DCO, premerge, and docs build all failed. Can you try to fix them? I think you tagged the right person for reviews. I can also ping them internally once I see those passing :)

GeneDer avatar Feb 08 '24 01:02 GeneDer

@GeneDer, Thanks for looking into this. I raised https://github.com/ray-project/ray/issues/43018 for storing the output image. And in https://github.com/ray-project/ray/pull/43046, I've referenced that image URL in the doc. I checked the failures; looks like premerge and doc build are failing due to "(serve/tutorials/aws-neuron-core-inference-stablediffusion: line 122) broken https://raw.githubusercontent.com/ray-project/images/master/docs/serve/stable_diffusion_inferentia2_output.png - 404 Client Error: Not Found for url: https://raw.githubusercontent.com/ray-project/images/master/docs/serve/stable_diffusion_inferentia2_output.png". I am thinking we need to merge https://github.com/ray-project/ray/issues/43018 first to get the jobs passed to have the image available. Please let me know your thoughts.

ratnopam avatar Feb 08 '24 15:02 ratnopam

Those links are confusing lol Got this merged https://github.com/ray-project/images/pull/18 But there are more issue with your PR tho, make sure to fix the rest

GeneDer avatar Feb 10 '24 21:02 GeneDer

@GeneDer thanks a lot for your review. All checks are passing now.

ratnopam avatar Feb 11 '24 14:02 ratnopam

@ratnopamc just in case you missed the notification, can you address a few comments suggested by the docs team?

GeneDer avatar Feb 15 '24 16:02 GeneDer

@GeneDer, thanks. I resubmitted for review.

ratnopam avatar Feb 17 '24 05:02 ratnopam