[CI] Add Helion as an optional dependency (#32482)

Signed-off-by: Yanan Cao <gmagogsfm@gmail.com>

.buildkite/test-amd.yaml

@@ -703,6 +703,17 @@ steps:
     - pytest -v -s kernels/moe/test_batched_deepgemm.py
     - pytest -v -s kernels/attention/test_deepgemm_attention.py
 
+- label: Kernels Helion Test
+  timeout_in_minutes: 30
+  mirror_hardwares: [amdexperimental, amdproduction]
+  agent_pool: mi325_1
+  source_file_dependencies:
+  - vllm/utils/import_utils.py
+  - tests/kernels/helion/
+  commands:
+    - pip install helion
+    - pytest -v -s kernels/helion/
+
 - label: Model Executor Test # 23min
   timeout_in_minutes: 35
   torch_nightly: true

.buildkite/test-pipeline.yaml

@@ -624,6 +624,16 @@ steps:
     - pytest -v -s kernels/moe/test_batched_deepgemm.py
     - pytest -v -s kernels/attention/test_deepgemm_attention.py
 
+- label: Kernels Helion Test
+  timeout_in_minutes: 30
+  gpu: h100
+  source_file_dependencies:
+  - vllm/utils/import_utils.py
+  - tests/kernels/helion/
+  commands:
+    - pip install helion
+    - pytest -v -s kernels/helion/
+
 - label: Model Executor Test # 23min
   timeout_in_minutes: 35
   torch_nightly: true

setup.py

@@ -992,6 +992,8 @@ setup(
         "flashinfer": [],  # Kept for backwards compatibility
         # Optional deps for AMD FP4 quantization support
         "petit-kernel": ["petit-kernel"],
+        # Optional deps for Helion kernel development
+        "helion": ["helion"],
     },
     cmdclass=cmdclass,
     package_data=package_data,

tests/kernels/helion/test_helion_available.py

+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+"""
+Tests for Helion kernel availability and basic functionality.
+
+This module demonstrates the pattern for testing optional Helion kernels.
+Tests in this directory will be skipped if Helion is not installed.
+"""
+
+import pytest
+
+from vllm.utils.import_utils import has_helion
+
+# Skip entire module if helion is not available
+if not has_helion():
+    pytest.skip(
+        "Helion is not installed. Install with: pip install vllm[helion]",
+        allow_module_level=True,
+    )
+
+import helion
+import helion.language as hl
+import torch
+
+
+def test_helion_kernel_compilation_smoke():
+    """Smoke test: compile and run a simple Helion kernel."""
+
+    @helion.kernel(autotune_effort="none")
+    def add_kernel(x: torch.Tensor, y: torch.Tensor) -> torch.Tensor:
+        out = torch.empty_like(x)
+        for tile in hl.tile(x.size()):
+            out[tile] = x[tile] + y[tile]
+        return out
+
+    # Create test tensors
+    x = torch.randn(1024, device="cuda", dtype=torch.float32)
+    y = torch.randn(1024, device="cuda", dtype=torch.float32)
+
+    # Run the helion kernel
+    result = add_kernel(x, y)
+
+    # Verify correctness
+    expected = x + y
+    assert torch.allclose(result, expected), "Helion kernel output mismatch"

vllm/utils/import_utils.py

@@ -436,3 +436,18 @@ def has_arctic_inference() -> bool:
     """Whether the optional `arctic_inference` package is available."""
 
     return _has_module("arctic_inference")
+
+
+def has_helion() -> bool:
+    """Whether the optional `helion` package is available.
+
+    Helion is a Python-embedded DSL for writing ML kernels.
+    See: https://github.com/pytorch/helion
+
+    Usage:
+        if has_helion():
+            import helion
+            import helion.language as hl
+            # use helion...
+    """
+    return _has_module("helion")