[Kernel] Remove unused fake implementation

[jeejeelee]

Purpose

Test Plan

Test Result

---

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.

[claude]

Claude Code Review

This repository is configured for manual code reviews. Comment @claude review for a one-time review, or @claude review always to subscribe this PR to a review on every future push.

_{Tip: disable this comment in your organization's Code Review settings.}

[jeejeelee]

/ci run

[github-actions]

✅ Triggered Buildkite CI #87418 for commit b02568e35a2a.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Team

Run ID: 8ba2cfdc-d8be-4fd8-bd2a-1e8183ec62e7

📥 Commits

Reviewing files that changed from the base of the PR and between f2e2936 and b02568e.

📒 Files selected for processing (39)

• vllm/_aiter_ops.py
• vllm/_custom_ops.py
• vllm/_xpu_ops.py
• vllm/compilation/passes/fusion/allreduce_rms_fusion.py
• vllm/kernels/helion/ops/dynamic_per_token_scaled_fp8_quant.py
• vllm/kernels/helion/ops/fused_qk_norm_rope.py
• vllm/kernels/helion/ops/per_token_group_fp8_quant.py
• vllm/kernels/helion/ops/rms_norm_dynamic_per_token_quant.py
• vllm/kernels/helion/ops/rms_norm_per_block_quant.py
• vllm/kernels/helion/ops/silu_and_mul_per_block_quant.py
• vllm/kernels/helion/register.py
• vllm/lora/ops/triton_ops/fused_moe_lora_fp8_op.py
• vllm/lora/ops/triton_ops/fused_moe_lora_op.py
• vllm/lora/ops/triton_ops/lora_expand_fp8_op.py
• vllm/lora/ops/triton_ops/lora_expand_op.py
• vllm/lora/ops/triton_ops/lora_shrink_fp8_op.py
• vllm/lora/ops/triton_ops/lora_shrink_op.py
• vllm/lora/ops/xpu_ops/lora_ops.py
• vllm/model_executor/kernels/linear/scaled_mm/deep_gemm.py
• vllm/model_executor/layers/attention/attention.py
• vllm/model_executor/layers/attention/mla_attention.py
• vllm/model_executor/layers/attention/static_sink_attention.py
• vllm/model_executor/layers/hpc/rope_norm.py
• vllm/model_executor/layers/mamba/gdn/olmo_gdn_linear_attn.py
• vllm/model_executor/layers/mamba/gdn/qwen_gdn_linear_attn.py
• vllm/model_executor/layers/mamba/linear/minimax_linear_attn.py
• vllm/model_executor/layers/mamba/mamba_mixer.py
• vllm/model_executor/layers/mamba/mamba_mixer2.py
• vllm/model_executor/layers/mamba/ops/cpu/gdn_attention.py
• vllm/model_executor/layers/mamba/short_conv.py
• vllm/model_executor/layers/rotary_embedding/common.py
• vllm/model_executor/models/bailing_moe_v3.py
• vllm/model_executor/offloader/prefetch_ops.py
• vllm/models/deepseek_v4/xpu/model.py
• vllm/models/qwen4_exp/amd/ple_layer.py
• vllm/models/qwen4_exp/amd/qsa.py
• vllm/models/qwen4_exp/nvidia/ple_layer.py
• vllm/models/qwen4_exp/nvidia/qsa.py
• vllm/utils/flashinfer.py

💤 Files with no reviewable changes (38)

• vllm/lora/ops/triton_ops/lora_expand_fp8_op.py
• vllm/model_executor/layers/rotary_embedding/common.py
• vllm/model_executor/layers/attention/static_sink_attention.py
• vllm/kernels/helion/ops/fused_qk_norm_rope.py
• vllm/model_executor/kernels/linear/scaled_mm/deep_gemm.py
• vllm/model_executor/layers/mamba/mamba_mixer2.py
• vllm/models/qwen4_exp/nvidia/ple_layer.py
• vllm/kernels/helion/ops/dynamic_per_token_scaled_fp8_quant.py
• vllm/kernels/helion/ops/rms_norm_per_block_quant.py
• vllm/model_executor/layers/mamba/short_conv.py
• vllm/model_executor/offloader/prefetch_ops.py
• vllm/models/qwen4_exp/amd/qsa.py
• vllm/models/qwen4_exp/nvidia/qsa.py
• vllm/model_executor/layers/mamba/mamba_mixer.py
• vllm/lora/ops/triton_ops/lora_shrink_fp8_op.py
• vllm/lora/ops/triton_ops/lora_shrink_op.py
• vllm/models/qwen4_exp/amd/ple_layer.py
• vllm/model_executor/models/bailing_moe_v3.py
• vllm/model_executor/layers/mamba/linear/minimax_linear_attn.py
• vllm/model_executor/layers/mamba/gdn/olmo_gdn_linear_attn.py
• vllm/kernels/helion/ops/silu_and_mul_per_block_quant.py
• vllm/lora/ops/triton_ops/fused_moe_lora_fp8_op.py
• vllm/model_executor/layers/hpc/rope_norm.py
• vllm/models/deepseek_v4/xpu/model.py
• vllm/lora/ops/triton_ops/fused_moe_lora_op.py
• vllm/utils/flashinfer.py
• vllm/lora/ops/triton_ops/lora_expand_op.py
• vllm/lora/ops/xpu_ops/lora_ops.py
• vllm/model_executor/layers/attention/mla_attention.py
• vllm/model_executor/layers/mamba/ops/cpu/gdn_attention.py
• vllm/_aiter_ops.py
• vllm/kernels/helion/ops/per_token_group_fp8_quant.py
• vllm/_custom_ops.py
• vllm/compilation/passes/fusion/allreduce_rms_fusion.py
• vllm/model_executor/layers/attention/attention.py
• vllm/kernels/helion/ops/rms_norm_dynamic_per_token_quant.py
• vllm/model_executor/layers/mamba/gdn/qwen_gdn_linear_attn.py
• vllm/_xpu_ops.py

Included review availability: Your plan provides up to 10 included reviews per hour; 9 remain after this review.

---

📝 Summary

Summary by CodeRabbit

• Refactor

  • Streamlined custom operator registration across attention, quantization, routing, recurrent, LoRA, and device-specific operations.
  • Removed redundant compile-time fallback implementations while preserving runtime kernel behavior and mutation handling.
  • Improved kernel registration flexibility for operations that mutate outputs without returning values.
  • Updated generated-input handling for per-block quantization scenarios.

• Compatibility

  • Existing operator interfaces and runtime implementations remain unchanged; no end-user functionality was removed.

Walkthrough

The change removes redundant fake implementations from custom operators and Helion kernels. Registrations retain runtime implementations and mutation metadata. Helion registration now detects void-returning mutating kernels and permits missing fake implementations.

Changes

Fake implementation cleanup

Layer / File(s)	Summary
Backend operator registrations vllm/_aiter_ops.py, vllm/_custom_ops.py, vllm/_xpu_ops.py, vllm/compilation/passes/fusion/allreduce_rms_fusion.py	Selected AITER, Torch, XPU, and fusion operators no longer register fake implementations.
Helion registration handling vllm/kernels/helion/register.py, vllm/kernels/helion/ops/*	Helion removes redundant fake implementations. Registration skips inferred fakes for void-returning kernels that mutate arguments.
LoRA operator registrations vllm/lora/ops/triton_ops/*, vllm/lora/ops/xpu_ops/lora_ops.py	LoRA custom operators retain their runtime functions and mutation metadata without fake implementations.
Model operator registrations vllm/model_executor/*, vllm/models/*, vllm/utils/flashinfer.py	Attention, Mamba, rotary, offloader, model, and FlashInfer operators no longer register removed fake implementations.

Estimated code review effort: 3 (Moderate) | ~20 minutes

Merge Risk: ⚪ Minimal · up to b0256

This change removes redundant fake custom-operator implementations while retaining runtime implementations and mutation metadata. No concrete current-head merge-blocking risk remains.

🚥 Pre-merge checks | ✅ 3 | ❌ 2

❌ Failed checks (1 warning, 1 inconclusive)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 33.33% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 3 functions across 1 files.	Write docstrings for the functions missing them to satisfy the coverage threshold.
Description check	❓ Inconclusive	The description is an unfilled template. It does not provide a specific purpose, test plan, or test results.	Replace the template with a concise purpose, the exact tests or commands executed, and the resulting test status or output.

✅ Passed checks (3 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title accurately summarizes the main change: removal of unused fake implementations from kernel-related code.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

• Fix all pre-merge checks with AI

✨ Finishing Touches 💡 1

📝 Generate docstrings 💡

• Create stacked PR
• Commit on current branch

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}