[AMD/ROCM] GLM5/5.1 FP8 MTP Support on MI355X

[ajith-sirra-amd]

Overview

Add GLM-5 (GLM5.1 architecture) FP8 MTP benchmark configuration and testing support for AMD MI355X hardware.

Changes

• Added benchmark script for GLM-5.1 FP8 model on MI355X with MTP to run with Updated SGLang Image.
• Updated GitHub Actions configuration for AMD Master Yaml File.

Testing

• Verify benchmark execution on MI355X hardware
• Validate configuration settings

[claude]

Claude Code Review

This pull request is from a fork — automated review is disabled. A repository maintainer can comment @claude review to run a one-time review.

[seungrokj]

/sweep test-config --config-files .github/configs/amd-master.yaml --config-keys glm5.1-fp8-mi355x-sglang-mtp

[github-actions]

@seungrokj Kicking off a sweep.

Run: https://github.com/SemiAnalysisAI/InferenceX/actions/runs/24835029786
Command: test-config --config-files .github/configs/amd-master.yaml --config-keys glm5.1-fp8-mi355x-sglang-mtp
Pinned ref: 5a9c062
Approval: not required (trusted collaborator).

[functionstackx]

can u remove glm5.1 then? glm5.1 & glm5 is in the same class of architecture

#1086

[seungrokj]

@ajith-sirra-amd can you plz update glm5.1 to glm5 (so that this PR is an TP4 search space extension of existing PR #1086) ?

[chunfangamd]

lgtm

[chunfangamd]

lgtm

[functionstackx]

Thanks for the contribution! Can u add back chat templates as that more closely aligns the AR distribution with real world

[functionstackx]

lgtm, assuming there is an validation run

[seungrokj]

/sweep test-config --config-files .github/configs/amd-master.yaml --config-keys glm5-fp8-mi355x-sglang-mtp --evals-only

[github-actions]

@seungrokj Kicking off a sweep.

Run: https://github.com/SemiAnalysisAI/InferenceX/actions/runs/25197000070
Command: test-config --config-files .github/configs/amd-master.yaml --config-keys glm5-fp8-mi355x-sglang-mtp --evals-only
Pinned ref: 8504624
Approval: not required (trusted collaborator).