Fix iGPU CMakeFile tags (#6695) (#6698)

* Unify iGPU test build with x86 ARM

* adding TRITON_IGPU_BUILD to core build definition; adding logic to skip caffe2plan test if TRITON_IGPU_BUILD=1

* re-organizing some copies in Dockerfile.QA to fix igpu devel build

* Pre-commit fix

---------

Co-authored-by: kyle <[email protected]>

CMakeLists.txt

@@ -232,6 +232,7 @@ ExternalProject_Add(triton-server
     -Daws-c-common_DIR:PATH=${TRITON_THIRD_PARTY_INSTALL_PREFIX}/aws-sdk-cpp/${LIB_DIR}/aws-c-common/cmake
     -Daws-checksums_DIR:PATH=${TRITON_THIRD_PARTY_INSTALL_PREFIX}/aws-sdk-cpp/${LIB_DIR}/aws-checksums/cmake
     -Dopentelemetry-cpp_DIR:PATH=${_FINDPACKAGE_OPENTELEMETRY_CONFIG_DIR}
+    -DTRITON_IGPU_BUILD:BOOL=${TRITON_IGPU_BUILD}
     -DTRITON_THIRD_PARTY_REPO_TAG:STRING=${TRITON_THIRD_PARTY_REPO_TAG}
     -DTRITON_COMMON_REPO_TAG:STRING=${TRITON_COMMON_REPO_TAG}
     -DTRITON_CORE_REPO_TAG:STRING=${TRITON_CORE_REPO_TAG}

Dockerfile.QA

@@ -246,11 +246,13 @@ RUN mkdir -p qa/L0_decoupled/models/repeat_int32/1 && \
     mkdir -p qa/L0_decoupled/models/fan_repeat/1 && \
     mkdir -p qa/L0_decoupled/models/sequence_repeat/1 && \
     mkdir -p qa/L0_decoupled/models/repeat_square/1 && \
-    mkdir -p qa/L0_decoupled/models/nested_square/1
+    mkdir -p qa/L0_decoupled/models/nested_square/1 && \
+    mkdir -p qa/L0_grpc_state_cleanup/models/repeat_int32/1
 
 RUN if [ "$IGPU_BUILD" == "0" ]; then \
         cp backends/repeat/libtriton_repeat.so qa/L0_model_config && \
         cp backends/repeat/libtriton_repeat.so qa/L0_decoupled/models/repeat_int32/1 && \
+        cp backends/repeat/libtriton_repeat.so qa/L0_grpc_state_cleanup/models/repeat_int32/1/. && \
         cp backends/square/libtriton_square.so qa/L0_decoupled/models/square_int32/1; \
     fi
 
@@ -264,10 +266,6 @@ RUN cp -r qa/L0_decoupled/models qa/L0_decoupled/python_models/ && \
     cp /workspace/tritonbuild/python/examples/decoupled/square_config.pbtxt \
         qa/L0_decoupled/python_models/square_int32/.
 
-RUN mkdir -p qa/L0_grpc_state_cleanup/models/repeat_int32/1 && \
-    cp backends/repeat/libtriton_repeat.so \
-        qa/L0_grpc_state_cleanup/models/repeat_int32/1/.
-
 RUN mkdir -p qa/L0_repoagent_checksum/models/identity_int32/1 && \
     cp tritonbuild/identity/install/backends/identity/libtriton_identity.so \
         qa/L0_repoagent_checksum/models/identity_int32/1/.

build.py

@@ -616,9 +616,7 @@ def pytorch_cmake_args(images):
     if "pytorch" in images:
         image = images["pytorch"]
     else:
-        image = "nvcr.io/nvidia/pytorch:{}-py3".format(
-            FLAGS.upstream_container_version
-        )
+        image = "nvcr.io/nvidia/pytorch:{}-py3".format(FLAGS.upstream_container_version)
     cargs = [
         cmake_backend_arg("pytorch", "TRITON_PYTORCH_DOCKER_IMAGE", None, image),
     ]
@@ -651,7 +649,6 @@ def onnxruntime_cmake_args(images, library_paths):
             )
         )
 
-
     if target_platform() == "windows":
         if "base" in images:
             cargs.append(
@@ -767,9 +764,7 @@ def tensorflow_cmake_args(images, library_paths):
             FLAGS.upstream_container_version
         )
     extra_args = [
-        cmake_backend_arg(
-            backend_name, "TRITON_TENSORFLOW_DOCKER_IMAGE", None, image
-        )
+        cmake_backend_arg(backend_name, "TRITON_TENSORFLOW_DOCKER_IMAGE", None, image)
     ]
     return extra_args
 

src/CMakeLists.txt

@@ -774,7 +774,5 @@ endif() # NOT WIN32
 
 # Currently unit tests do not build for windows...
 if ( NOT WIN32)
-  if ( NOT EXISTS "/etc/ld.so.conf.d/nvidia-tegra.conf" )
-    add_subdirectory(test test)
-  endif()
+  add_subdirectory(test test)
 endif() # NOT WIN32

src/test/CMakeLists.txt

@@ -34,7 +34,7 @@ find_package(GTest REQUIRED)
 #
 # caffe2plan
 #
-if(${TRITON_ENABLE_TENSORRT})
+if(${TRITON_ENABLE_TENSORRT} AND NOT ${TRITON_IGPU_BUILD})
   add_executable(caffe2plan caffe2plan.cc)
   target_include_directories(caffe2plan PRIVATE ${CUDA_INCLUDE_DIRS})
   target_link_libraries(