diff --git a/fbgemm_gpu/experimental/example/test/triton_example_test.py b/fbgemm_gpu/experimental/example/test/triton_example_test.py
index 8e043656ad..3e134835cc 100644
--- a/fbgemm_gpu/experimental/example/test/triton_example_test.py
+++ b/fbgemm_gpu/experimental/example/test/triton_example_test.py
@@ -14,7 +14,7 @@
 @triton.jit
 # fmt: off
 def triton_add_kernel(x_ptr, y_ptr, z_ptr, n_elements, BLOCK_SIZE: tl.constexpr) -> None:
-# fmt: on
+# fmt: on  # noqa E115
     # We use a 1D launch grid so axis is 0.
     pid = tl.program_id(axis=0)
 
diff --git a/fbgemm_gpu/experimental/gen_ai/CMakeLists.txt b/fbgemm_gpu/experimental/gen_ai/CMakeLists.txt
index a82236cbba..a95759befb 100644
--- a/fbgemm_gpu/experimental/gen_ai/CMakeLists.txt
+++ b/fbgemm_gpu/experimental/gen_ai/CMakeLists.txt
@@ -35,14 +35,14 @@ set(quantize_ops_sources
     src/quantize/quantize.cu
     src/quantize/quantize.cpp)
 
-# set(comm_ops_sources
-#     src/comm/car.cu
-#     src/comm/car.cpp)
+set(comm_ops_sources
+    src/comm/car.cu
+    src/comm/car.cpp)
 
 set(experimental_gen_ai_cpp_source_files
     ${attention_ops_sources}
-    ${quantize_ops_sources})
-#     ${comm_ops_sources})
+    ${quantize_ops_sources}
+    ${comm_ops_sources})
 
 set_source_files_properties(${experimental_gen_ai_cpp_source_files}
     PROPERTIES INCLUDE_DIRECTORIES