[CI] Recipe bring-up for /overview combos with no data / [CI] 缺数据 combo 的 recipe bring-up 跟踪

Open

#2,587 opened on Aug 12, 2026 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Assessment

Difficulty: 5/5
Estimated time: Over a week
Newbie friendliness: 35/100

Issue type: Feature
Clarity: Mostly clear
Activity status: Quiet
Tech stack: docker, python
Domain: ci-cd, machine-learning

Research direction

Start by reviewing parent issue #2304 and the referenced bring-up issues #2403, #2496, #2508, and #2551, then inspect the closed test PRs for the kimik3 × gb300 path. Done means benchmark recipes exist for each still-missing combo and scheduled runs can populate the corresponding /overview data.

Written by the indexing model from the issue text.

Description

These /overview model×hardware combos have no benchmark recipe, so no cadence of scheduled runs can populate them (parent: #2304). Each needs config bring-up:

kimik3 × mi355x — active bring-up in #2403 / #2496 / #2508 / #2551 (ROCm DSpark HSA faults, cudagraph modes, vllm #51682)
kimik3 × gb300 — only closed [Test] PRs so far; likely blocked on aarch64 build of the pre-release vllm/vllm-openai:kimi-k3 image
glm5.2 × gb200 — not started
glm5.2 × gb300 — not started

glm5.2 × mi355x landed in #2488 (2026-08-11) and is no longer missing.

以下 /overview model×hardware combo 没有 benchmark recipe，定时跑多少次都补不上数据（父 issue：#2304），需要逐个 bring-up：

kimik3 × mi355x — #2403 / #2496 / #2508 / #2551 正在攻坚（ROCm DSpark HSA fault、cudagraph mode、vllm #51682）
kimik3 × gb300 — 只有已关闭的 [Test] PR；疑似卡在 pre-release vllm/vllm-openai:kimi-k3 镜像无 aarch64 build
glm5.2 × gb200 — 未启动
glm5.2 × gb300 — 未启动

glm5.2 × mi355x 已由 #2488（2026-08-11）落地，不再缺失。

Dominant language: Python
Stars: 1.7k
Forks: 303
Avg merge: 1d 13h
Merged PRs (30d): 284

Contributor guide

Open the contributing guide

First steps

Read the whole issue, then the project's contributing guide.
Comment on the issue to say you are picking it up — it saves two people doing the same work.
Fork the repository and make your change on a branch.
Open a pull request that references the issue number.