[Kimi-K3] Add AttnRes kernels

.buildkite/test_areas/models_basic.yaml

@@ -61,6 +61,18 @@ steps:
     # FA4 kernel tests require SM100; the suite skips them elsewhere.
     - pytest -v -s models/inkling
 
+- label: Kimi K3 Unit Tests (B200)
+  key: kimi-k3-unit-tests-b200
+  timeout_in_minutes: 40
+  device: b200-k8s
+  source_file_dependencies:
+  - vllm/models/kimi_k3/
+  - csrc/libtorch_stable/kimi_k3/
+  - tests/models/kimi_k3/
+  commands:
+    # The native NVIDIA AttnRes kernel requires the SM100 family.
+    - pytest -v -s models/kimi_k3
+
 - label: Basic Models Test (Other CPU) # 5min
   key: basic-models-test-other-cpu
   depends_on:

CMakeLists.txt

@@ -1074,6 +1074,24 @@ if(VLLM_GPU_LANG STREQUAL "CUDA" OR VLLM_GPU_LANG STREQUAL "HIP")
     set(MLA_ARCHS)
   endif()
 
+  if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 13.0)
+    cuda_archs_loose_intersection(KIMI_K3_ATTN_RES_ARCHS
+      "10.0f" "${CUDA_ARCHS}")
+  endif()
+  if(KIMI_K3_ATTN_RES_ARCHS)
+    set(KIMI_K3_ATTN_RES_SRC
+      "csrc/libtorch_stable/kimi_k3/attn_res_kernel.cu")
+    set_gencode_flags_for_srcs(
+      SRCS "${KIMI_K3_ATTN_RES_SRC}"
+      CUDA_ARCHS "${KIMI_K3_ATTN_RES_ARCHS}")
+    set_property(SOURCE ${KIMI_K3_ATTN_RES_SRC} APPEND PROPERTY
+      COMPILE_OPTIONS
+      "$<$<COMPILE_LANGUAGE:CUDA>:--expt-relaxed-constexpr;--expt-extended-lambda;--use_fast_math>")
+    list(APPEND VLLM_STABLE_EXT_SRC "${KIMI_K3_ATTN_RES_SRC}")
+    message(STATUS
+      "Building Kimi K3 AttnRes for archs: ${KIMI_K3_ATTN_RES_ARCHS}")
+  endif()
+
   # Hadacore kernels
   cuda_archs_loose_intersection(HADACORE_ARCHS "8.0+PTX;9.0+PTX" "${CUDA_ARCHS}")
   if(HADACORE_ARCHS)
@@ -1115,6 +1133,10 @@ if(VLLM_GPU_LANG STREQUAL "CUDA" OR VLLM_GPU_LANG STREQUAL "HIP")
       target_compile_definitions(_C_stable_libtorch PRIVATE
         VLLM_ENABLE_COOPERATIVE_TOPK=1)
     endif()
+    if(KIMI_K3_ATTN_RES_ARCHS)
+      target_compile_definitions(_C_stable_libtorch PRIVATE
+        VLLM_ENABLE_KIMI_K3_ATTN_RES=1)
+    endif()
     # Needed by CUTLASS kernels
     target_compile_definitions(_C_stable_libtorch PRIVATE
       CUTLASS_ENABLE_DIRECT_CUDA_DRIVER_CALL=1)