[AMD/ROCm] kimik2.5-fp4-mi355x-atom. Bump ATOM image to rocm/atom:rocm7.2.3_ubuntu24.04_py3.12_pytorch_release_2.10.0_atom20260511

[seungrokj]

Summary

Bumps the ATOM docker image for kimik2.5-fp4-mi355x-atom from:

rocm/atom:rocm7.2.1-ubuntu24.04-pytorch2.9.1-atom0.1.2

to:

rocm/atom-dev:nightly_202605111702

Performance vs previous InferenceX data

Based on comparing InferenceX benchmark (date: 2026-04-01) against ATOM upstream run #25686894636 (2026-05-11, rocm/atom-dev:nightly_202605111702):

ISL	OSL	Conc	GPUs (tp)	IX tput/GPU	ATOM tput/GPU	Δ
1024	1024	4	4	189.05	212.56	+12.4%
1024	1024	8	4	287.02	376.15	+31.1%
1024	1024	16	4	456.55	546.70	+19.7%
1024	1024	128	4	1648.44	1645.68	~0%
8192	1024	4	4	680.48	849.95	+24.9%
8192	1024	8	4	1192.45	1402.76	+17.6%
8192	1024	16	4	1629.17	1960.34	+20.3%
8192	1024	128	4	3510.36	4088.46	+16.5%

No changes to the benchmark script or serve arguments — image bump only.

[github-actions]

Thanks for the contribution! For vLLM & SGLang, please ensure that your recipes is similar to the official vLLM recipes and/or the SGLang cookbook

• https://docs.sglang.io/cookbook/intro
• https://recipes.vllm.ai/

If it is not, please create a PR first before we can merge your single node PR into the master branch. Let's ensure that the documentation is first class such that the entire ML community can benefit from your hard work! Thank you

• https://github.com/vllm-project/recipes
• https://github.com/sgl-project/sglang/tree/main/docs_new

PR authors are responsible for ensuring that after merging, all GitHub Action jobs fully pass. A lot of the time, failures are just flakes and simply re-running the failed jobs will fix it. If re-running failed jobs is attempted, PR authors are responsible for ensuring it passes. See GitHub's docs on re-running failed jobs: https://docs.github.com/en/actions/how-tos/manage-workflow-runs/re-run-workflows-and-jobs#re-running-failed-jobs-in-a-workflow

As a rule of thumb, generally, PR authors should request a review & get a PR approval from the respective companies' CODEOWNERS before requesting a review from core maintainers.

If additional help is needed, PR authors can reach out to core maintainers over Slack.

[github-actions]

Thanks for the contribution! For vLLM & SGLang, please ensure that your recipes is similar to the official vLLM recipes and/or the SGLang cookbook

• https://docs.sglang.io/cookbook/intro
• https://recipes.vllm.ai/

If it is not, please create a PR first before we can merge your single node PR into the master branch. Let's ensure that the documentation is first class such that the entire ML community can benefit from your hard work! Thank you

• https://github.com/vllm-project/recipes
• https://github.com/sgl-project/sglang/tree/main/docs_new

PR authors are responsible for ensuring that after merging, all GitHub Action jobs fully pass. A lot of the time, failures are just flakes and simply re-running the failed jobs will fix it. If re-running failed jobs is attempted, PR authors are responsible for ensuring it passes. See GitHub's docs on re-running failed jobs: https://docs.github.com/en/actions/how-tos/manage-workflow-runs/re-run-workflows-and-jobs#re-running-failed-jobs-in-a-workflow

As a rule of thumb, generally, PR authors should request a review & get a PR approval from the respective companies' CODEOWNERS before requesting a review from core maintainers.

If additional help is needed, PR authors can reach out to core maintainers over Slack.

[seungrokj]

@functionstackx @cquil11
can you plz merge this ?
e2e perf / accuracy
https://github.com/SemiAnalysisAI/InferenceX/actions/runs/25962958917

https://inferencex.semianalysis.com/inference?unofficialRun=25962958917