[Thangnguyenvn98]

Overview

Enrolls five steps and four of their AMD mirrors in automatic_shard, continuing the earlier enrollment batches (#64115, #64174, #64181). Each runs as one job today, unchanged, and splits without another CI change once its tests outgrow one job. One command changes: DeepSeek V4 Kernels (H100) names the single file its glob matched, so the shard planner can find it.

Claims

Step	Plan	AMD mirror
language-models-test-ppl	1 shard	1 shard
multi-modal-models-standard-1-qwen2	1 shard	1 shard
engine	1 shard	1 shard
deepseek-v4-kernel-test-h100	1 shard	1 shard
kernels-deepgemm-test-h100	1 shard	none

• Each job runs the same tests, environment and timeout as on main.

Validation

• Planned each step and mirror with ci-infra main's planner against main's timings: every plan is one job, within main's machine time.
• Every step and mirror passes on recent main runs, nightly included.
• kernels/test_fused_deepseek_v4_*.py matches only test_fused_deepseek_v4_qnorm_rope_kv_insert.py today, the one test file the step's source_file_dependencies lists. In build 98080 that command collected and passed the same 215 tests as on main.
• /ci run on the previous head (98766) plus a build scoped to the AMD mirrors it leaves behind the block step (98793): every job passed. A fresh /ci run follows on this head.

Details

• Left out: steps the generator can't shard. It only shards a step whose commands after the first pytest are all plain pytest. An earlier revision also enrolled Python 3.14 Smoke, Python-only Installation (shell scripts) and Plugin Integration (pip install between pytest commands), with their mirrors; the generator ignored the flag on them, so this PR no longer touches them.
• DeepSeek V4 glob. The planner reads the files a pytest command names from the checkout and doesn't expand shell globs, so with the glob the step got no plan. The B200 DeepSeek V4 step keeps its glob; it isn't in this batch.
• Engine Core caplog. test_arg_utils.py, test_config.py and test_logger.py assert on logs; none starts an engine (test_logger.py's LLM(...) raises before engine init), and the step stays one job in its current order.
• Timeouts unchanged. timeout_in_minutes stays per job. Fallback-aware timeouts belong in a central timeout setting.

🤖 Generated with Claude Code

---

Pull Request Checklist

• I used vLLM's /pr-checklist skill. (Mandatory for agents, optional for humans).

• AI assistance was used during the creation of this PR.

• Design Fit: Minimizes impact on core components, reuses existing functionality, and justifies added complexity.

• Testing and Validation: Validates the change and ensures any added tests are meaningful and reliable, with CI coverage or documented CI resource constraints and validation performed outside CI.

• Code Quality and Style: Keeps code and comments clear and concise, and updates relevant documentation and examples.

• Pull Request Contents: Includes a brief summary and relevant links, supports claims with evidence, explains root causes and implementation trade-offs, and follows the contributing guide.

[Thangnguyenvn98]

/ci run

[github-actions]

✅ Triggered Buildkite CI #98080 for commit 1367a9781048.

[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.

[Thangnguyenvn98]

/ci run

[github-actions]

✅ Triggered Buildkite CI #98766 for commit 73207f4fd6fd.

[Thangnguyenvn98]

/ci run

[github-actions]

✅ Triggered Buildkite CI #98860 for commit 7799a545eb62.