Repository navigation
[AgentX] MiniMax-M3.1-Flash GB300 SGLang: add TP2 configs / [AgentX] MiniMax-M3.1-Flash GB300 SGLang：新增 TP2 配置 - #4511
Conversation
Adds TP2 with L2 HiCache at conc 32, 40 and 44, and raises mem-fraction-static to 0.87 on the DRAM-offload cells. 为 MiniMax-M3.1-Flash GB300 SGLang AgentX 新增 conc 32、40、44 的 TP2 + L2 HiCache 点，并把 DRAM offload 配置的 mem-fraction-static 提高到 0.87。
ispobock
added
the
full-sweep-fail-fast
|
Thanks for the contribution!
中文感谢你的贡献！
|
![claude[bot]](https://avatars.githubusercontent.com/in/1236702?s=60&v=4){kind=small}
There was a problem hiding this comment.
Nothing blocking. The comments below are optional suggestions. There is no need to push a fix for them before merging.
Beyond the inline comment, I checked three other things and none needs a fix. The pull/XXX pr-link is accepted by validate_perf_changelog.py, and existing entries already use it. The new changelog entry is a regular one, not append-only, which fits because existing hicache points change. The three zipped lists in zip_override_tp2_hicache all have length 3 and line up with conc-list [32, 40, 44].
Extended reasoning...
The change only touches benchmark config: one srt-slurm recipe, one master-config search-space row and one perf-changelog entry for minimaxm31flash-fp4-gb300-sglang-agentic-dspark. It touches no security-sensitive code. The only finding is a nit about comments, which is posted inline. This note just records the extra checks that passed.
| search-space: | ||
| # Balanced TP4 with L2 HiCache: 180 GB per rank at conc 64. | ||
| - { tp: 4, spec-decoding: mtp, kv-offloading: dram, kv-offload-backend: { name: hicache }, conc-list: [64], srt-recipe: agentic.yaml } | ||
| # TP2 with L2 HiCache: 180 GB per rank at conc 32-44. |
There was a problem hiding this comment.
🟡 nit (optional): The new comments label the TP2 row and override, or restate values the YAML already shows, which AGENTS.md's mandatory comment rules forbid. sweep:# TP2 with L2 HiCache Example: nvidia-master.yaml:8957 only repeats tp 2 and conc-list [32, 40, 44]. Fix: drop the label and restated values, and keep only the reason, such as why 180 GB per rank fits the two-GPU DRAM allocation.
Why this was flagged
AGENTS.md:56 bans header comments that describe a config, override or section. AGENTS.md:58 allows only technical reasons for a line and says never restate the code. The new comment at nvidia-master.yaml:8957 says what the next line already encodes: tp 2, 180 GB, conc 32-44. The new block at agentic.yaml:90-91 opens with "TP2 with L2 HiCache: two-GPU engines", which labels zip_override_tp2_hicache. It also restates gpus: 2 and the 2x CONC batch. Neighbouring overrides use the same style, but that is precedent, not an exemption. No runtime effect.
Verification: AGENTS.md:54-58 forbids "header comments describing a ... config, recipe, override, or section" and says "never restate the code". nvidia-master.yaml:8957 adds # TP2 with L2 HiCache: 180 GB per rank at conc 32-44., which labels the row and restates tp: 2 and conc-list: [32, 40, 44]. agentic.yaml:90-91 labels zip_override_tp2_hicache and restates gpus: 2. Nothing breaks at runtime.
|
View unofficial run (performance): https://inferencex.semianalysis.com/inference?unofficialRun=40211526054 View unofficial run (accuracy): https://inferencex.semianalysis.com/evaluation?unofficialRun=40211526054 |
|
/use 40211526054 |
|
@edwingao28 staged run 40211526054. This run remains available across future |
|
Ingest report — ✅ Results ingested into the Ingest workflow · Merge Ingest run · Benchmark source run
benchmarkFiles 13, rawBenchmarkRows 13, coveredByPerJobFiles 0, purgedPoints 0, failedRunRows 0, runStatsNew 0, runStatsDuplicate 0, evalSamplesNew 0, gpuMetricSamplesNew 0, powerPublicationPoints 13, changelogEntriesWritten 1 Counts cover only this run. Rows: benchmark_results, eval_results, agentic rows linked to a trace, gpu_metrics series. This comment updates if the ingest is rerun. 中文Ingest 报告 — ✅ 结果已写入 上表仅统计本次运行写入的数据，各行单位依次为 benchmark_results 行、eval_results 行、已关联 trace 的 agentic 行和 gpu_metrics 序列。重新运行 ingest 时本评论会原地更新。 |
Add TP2 points with DRAM KV offload at conc 32, 40 and 44 to the GB300 SGLang AgentX config, and raise
mem-fraction-staticto 0.87 on the DRAM KV offload points. The TP2 host cache is 180 GB per rank, inside the 385 GB allocation of a two-GPU point.中文
为 GB300 SGLang AgentX 配置新增 conc 32、40、44 的 TP2 + DRAM KV offload 点，并把 DRAM KV offload 点的
mem-fraction-static提高到 0.87。TP2 每个 rank 的 host cache 为 180 GB，在两卡点 385 GB 的分配额度以内。AI model disclosure
Prepared with Claude Code using
claude-opus-5-5. Reviewed by the human author before submission.