[AMD][Qwen3.5] Bump MI355X AgentX image to v0.5.20-rocm720-mi35x-20260924

[yichiche]

Summary

• Bump qwen3.5-fp4-mi355x-sglang-agentic-mtp from lmsysorg/sglang-rocm:v0.5.19-rocm720-mi35x-20260915 to lmsysorg/sglang-rocm:v0.5.20-rocm720-mi35x-20260924 (Docker Hub tag HTTP 200, digest sha256:baadaba198e23c46c1dd651b5edefae8000bd5430943bd10b246b563905b3ecf).
• Keep TP4 GPU-resident conc [1, 4, 8, 12, 16]. Move TP2 conc 16 onto HiCache and drop TP2 GPU-resident conc 20. TP2 resident is [1, 4, 8, 12]; TP2 HiCache is [16, 20, 24, 28, 32, 36, 40].
• Pin the HiCache host pool with --hicache-size 253 (overrides --hicache-ratio). Other recipe flags are unchanged. Sweep image comes from configs/amd-master.yaml.

Details

origin/main still pins the 20260915 AgentX image and the older TP2 grid (resident [1, 4, 8, 12, 16, 20], HiCache starting at 20). This PR only updates that AgentX arm. Fixed-seq MI355X SGLang arms are a separate PR.

AI model disclosure

Cursor Grok 4.6 prepared the image bump, HiCache size pin, changelog, and this PR text.

[github-actions]

Thanks for the contribution!

• Review: If this PR changes files owned by someone other than a repository admin or @SemiAnalysisAI/core, ask one eligible CODEOWNER to complete the latest PR_REVIEW_CHECKLIST.md before contacting a core maintainer on Slack. Follow the template exactly, including As a PR reviewer and CODEOWNER, I have reviewed this and have, so sign-off verification triggers.
• PR verification: Sweeps only run on labeled PRs. Add full-sweep-fail-fast (strongly recommended); use full-sweep-enabled only when matrix jobs should continue after a failure.
• After merging: PR authors must ensure all GitHub Actions jobs pass. Transient failures often pass on rerun; see how to rerun failed jobs.

中文

感谢你的贡献！

• **审阅：**如果 PR 修改的文件归属于仓库管理员及 @SemiAnalysisAI/core 之外的 CODEOWNER，请先联系一位有资格的 CODEOWNER 填写最新的 PR_REVIEW_CHECKLIST.md，再通过 Slack 联系核心维护者。必须严格遵循模板，并保留 As a PR reviewer and CODEOWNER, I have reviewed this and have，才能触发签核验证。
• **PR 验证：**扫描仅在带有标签的 PR 上运行。强烈建议添加 full-sweep-fail-fast；仅当需要矩阵任务在失败后继续运行时才使用 full-sweep-enabled。
• **合并后：**PR 作者必须确保所有 GitHub Actions 任务通过。临时性失败通常可以通过重新运行恢复；参见重新运行失败任务的说明。

[functionstackx]

InferenceX has switched away from unmaintainable bash scripts to YAML files that don't repeat the same stuff over and over again. Please merge the latest main into this PR: we have migrated single-node AgentX onto native srt-slurm (#3428), so AgentX configs are now declarative YAML recipes, not per-config 1000+ line bash slop scripts. Please also delete the old benchmarks/single_node/** scripts (see this recipe for the new format).

[functionstackx]

Closing in favor of #3455, which ports this PR's changes onto the srt-slurm recipe that #3428 switched qwen3.5-fp4-mi355x-sglang-agentic-mtp to: the v0.5.20-rocm720-mi35x-20260924 image, TP2 resident conc [1, 4, 8, 12], TP2 HiCache conc [16, 20, 24, 28, 32, 36, 40], and hicache-size: 253 on every HiCache variant. The legacy benchmarks/single_node/agentic/qwen3.5_fp4_mi355x_sglang_mtp.sh script this PR edits is no longer used by that config. Thanks for the tuning!