[CI Perf] Prune tests in tests/kernels/moe/

[mgoin]

Purpose

pytest --collect-only -qq tests/kernels/moe

# main
tests/kernels/moe/test_batched_deepgemm.py: 32
tests/kernels/moe/test_batched_moe.py: 3936
tests/kernels/moe/test_block_fp8.py: 450
tests/kernels/moe/test_block_int8.py: 224
tests/kernels/moe/test_count_expert_num_tokens.py: 853
tests/kernels/moe/test_cutlass_grouped_gemm.py: 6
tests/kernels/moe/test_cutlass_moe.py: 462
tests/kernels/moe/test_deepep_deepgemm_moe.py: 34
tests/kernels/moe/test_deepep_moe.py: 56
tests/kernels/moe/test_deepgemm.py: 12
tests/kernels/moe/test_modular_kernel_combinations.py: 108
tests/kernels/moe/test_moe.py: 9105
tests/kernels/moe/test_moe_align_block_size.py: 1125
tests/kernels/moe/test_moe_permute_unpermute.py: 3456
tests/kernels/moe/test_mxfp4_moe.py: 3
tests/kernels/moe/test_nvfp4_moe.py: 180
tests/kernels/moe/test_pplx_cutlass_moe.py: 32
tests/kernels/moe/test_pplx_moe.py: 986
tests/kernels/moe/test_rocm_aiter_topk.py: 4
tests/kernels/moe/test_silu_mul_fp8_quant_deep_gemm.py: 5
tests/kernels/moe/test_triton_moe_ptpc_fp8.py: 32

# pr
tests/kernels/moe/test_batched_deepgemm.py: 32
tests/kernels/moe/test_batched_moe.py: 1104
tests/kernels/moe/test_block_fp8.py: 450
tests/kernels/moe/test_block_int8.py: 224
tests/kernels/moe/test_count_expert_num_tokens.py: 146
tests/kernels/moe/test_cutlass_grouped_gemm.py: 6
tests/kernels/moe/test_cutlass_moe.py: 462
tests/kernels/moe/test_deepep_deepgemm_moe.py: 34
tests/kernels/moe/test_deepep_moe.py: 56
tests/kernels/moe/test_deepgemm.py: 12
tests/kernels/moe/test_modular_kernel_combinations.py: 108
tests/kernels/moe/test_moe.py: 3701
tests/kernels/moe/test_moe_align_block_size.py: 325
tests/kernels/moe/test_moe_permute_unpermute.py: 432
tests/kernels/moe/test_mxfp4_moe.py: 3
tests/kernels/moe/test_nvfp4_moe.py: 180
tests/kernels/moe/test_pplx_cutlass_moe.py: 32
tests/kernels/moe/test_pplx_moe.py: 800
tests/kernels/moe/test_rocm_aiter_topk.py: 4
tests/kernels/moe/test_silu_mul_fp8_quant_deep_gemm.py: 5
tests/kernels/moe/test_triton_moe_ptpc_fp8.py: 32

Test Plan

Test Result

(Optional) Documentation Update

---

Essential Elements of an Effective PR Description Checklist

• The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
• The test plan, such as providing test command.
• The test results, such as pasting the results comparison before and after, or e2e results
• (Optional) The necessary documentation update, such as updating supported_models.md and examples for a new model.

[github-actions]

👋 Hi! Thank you for contributing to the vLLM project.

💬 Join our developer Slack at https://slack.vllm.ai to discuss your PR in #pr-reviews, coordinate on features in #feat- channels, or join special interest groups in #sig- channels.

Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run fastcheck CI which starts running only a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of those by going to your fastcheck build on Buildkite UI (linked in the PR checks section) and unblock them. If you do not have permission to unblock, ping simon-mo or khluu to add you in our Buildkite org.

Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging.

To run CI, PR reviewers can either: Add ready label to the PR or enable auto-merge.

🚀

[gemini-code-assist]

Code Review

This pull request aims to improve CI performance by pruning the test suite for MoE kernels. The changes effectively reduce the number of test combinations by limiting the values in pytest.mark.parametrize and replacing Cartesian products of parameters with a curated list of test cases.

However, a recurring pattern in this pruning is the removal of torch.float16 and torch.float32 data types from several tests. While this reduces test execution time, it also significantly reduces test coverage for these supported dtypes. This is risky as it could mask precision-related regressions, especially on hardware where float16 is common. I have added comments highlighting these specific cases, recommending to re-introduce some of the removed dtypes to maintain a safer level of test coverage.

[yewentao256]

LGTM, Thanks for the work!