[Thangnguyenvn98]

Overview

Enrolls the remaining Kernels jobs in automatic_shard: 30 jobs (18 steps and 12 AMD mirrors). Part of a series that enrolls every remaining test step, split by test area for review. Commands change only where noted below.

Claims

Step	Plan	AMD mirror
kernels-fp8-moe-test-1xh100	1 shard	none
kernels-mhc-test-b200	1 shard	1 shard
kernels-minimax-m3-attention-b200	1 shard	1 shard
kernels-fla-ops-test-b200	1 shard	1 shard
deepseek-v4-kernel-test-b200	1 shard	1 shard
kernels-fp4-moe-test-b200	1 shard	1 shard
kernels-flashmla-test-h100	1 shard	1 shard
kernels-minimax-reduce-rms-test-2-gpus	1 shard	1 shard
vllm-ir-tests	1 shard	1 shard
kernels-gate-linear-cutedsl-b200	1 shard	1 shard
kernels-bf16x3-router-gemm-h100	1 shard	none
kernels-attention-diffkv-test-h100	1 shard	1 shard
minimax-m3-model-tests-b200	1 shard	none
kernels-minimax-m3-sparse-prefill-h100	1 shard	none
minimax-m3-mega-moe-test-b200	1 shard	none
amd-fp8-moe-kernels-mi355	1 shard	none
amd-native-quantization-kernels-mi355	1 shard	none
amd-kernels-mi355	1 shard	none
kernels-b200	already enrolled	1 shard
kernels-fusedmoe-layer-test-2-b200s	already enrolled	1 shard

• One-shard and one-job plans run the same tests as on main, as one job, and split by themselves once their tests grow.

Validation

• Planned each step and mirror with ci-infra main's planner against main's timings; every plan is within 10% of the same plan run as one job.
• The sharding harness (each shard's commands run against a mock test tree, as a PR build and as main with the OTel wrapper) shows every test runs exactly once wherever a plan has 2 or more shards.
• Every step and mirror passes on recent main runs, nightly included.
• /ci run (98456, and 98510 on this head) plus a scoped build (98467) for the optional steps and AMD mirrors: every enrolled job passed, and every shard finished under 22 minutes.

Details

• DeepSeek V4 Kernels (B200) names the one file its glob matched (test_fused_deepseek_v4_qnorm_rope_kv_insert.py), as [CI] Enroll five steps and four AMD mirrors in automatic sharding #64355 did for the H100 step: the planner reads the files a command names and doesn't expand shell globs.
• Not plain pytest steps (shell scripts, torchrun, examples) are enrolled too; the generator keeps them as one job. Standalone AMD steps plan like any other since Fix automatic sharding for standalone AMD jobs and selected tests ci-infra#871.
• caplog. Files that assert on logs sit in one-shard plans here, so their order in the process doesn't change.
• 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.

BEFORE SUBMITTING, PLEASE READ https://docs.vllm.ai/en/latest/contributing (anything written below this line will be removed by GitHub Actions)

[Thangnguyenvn98]

/ci run

[github-actions]

✅ Triggered Buildkite CI #98456 for commit d8ab8e057349.

[Thangnguyenvn98]

/ci run

[github-actions]

✅ Triggered Buildkite CI #98510 for commit 0d343ffd3b64.

[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.