Fix ONNX Runtime version to rel-0.5.0 and modify files according to it (

#529)

Dockerfile

@@ -99,16 +99,13 @@ FROM ${BASE_IMAGE} AS trtserver_onnx
 # needs to be built from source
 
 # Onnx Runtime release version
-ARG ONNX_RUNTIME_VERSION=0.4.0
+ARG ONNX_RUNTIME_VERSION=0.5.0
 
 # Get release version of Onnx Runtime
 WORKDIR /workspace
 RUN apt-get update && apt-get install -y --no-install-recommends git
 
-# Check out stable commit on master until new release
-# to support cloud-based filesystems
-RUN git clone --recursive https://github.com/Microsoft/onnxruntime && \
-    (cd onnxruntime && git checkout 2f698bd54b713bb87dbd0bbb913e94bcf7fd480c)
+RUN git clone -b rel-${ONNX_RUNTIME_VERSION} --recursive https://github.com/Microsoft/onnxruntime
 
 ENV PATH="/opt/cmake/bin:${PATH}"
 ARG SCRIPT_DIR=/workspace/onnxruntime/tools/ci_build/github/linux/docker/scripts
@@ -234,8 +231,14 @@ COPY --from=trtserver_caffe2 /opt/conda/lib/python3.6/site-packages/torch/lib/li
       /opt/tensorrtserver/lib/
 
 # Onnx Runtime headers and library
-ARG ONNX_RUNTIME_VERSION=0.4.0
-COPY --from=trtserver_onnx /workspace/onnxruntime/include/onnxruntime \
+# Put include files to same directory as ONNX Runtime changed the include path
+# https://github.com/microsoft/onnxruntime/pull/1461
+ARG ONNX_RUNTIME_VERSION=0.5.0
+COPY --from=trtserver_onnx /workspace/onnxruntime/include/onnxruntime/core/session/onnxruntime_c_api.h \
+     /opt/tensorrtserver/include/onnxruntime/
+COPY --from=trtserver_onnx /workspace/onnxruntime/include/onnxruntime/core/providers/cpu/cpu_provider_factory.h \
+     /opt/tensorrtserver/include/onnxruntime/
+COPY --from=trtserver_onnx /workspace/onnxruntime/include/onnxruntime/core/providers/cuda/cuda_provider_factory.h \
      /opt/tensorrtserver/include/onnxruntime/
 COPY --from=trtserver_onnx /workspace/build/Release/libonnxruntime.so.${ONNX_RUNTIME_VERSION} \
      /opt/tensorrtserver/lib/

src/backends/onnx/loader.h

@@ -25,7 +25,7 @@
 // OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #pragma once
 
-#include <core/session/onnxruntime_c_api.h>
+#include <onnxruntime_c_api.h>
 #include "src/core/status.h"
 
 namespace nvidia { namespace inferenceserver {

src/backends/onnx/onnx_backend.cc

@@ -38,7 +38,7 @@
 #include "src/core/server_status.h"
 
 #ifdef TRTIS_ENABLE_GPU
-#include <core/providers/cuda/cuda_provider_factory.h>
+#include <cuda_provider_factory.h>
 #include <cuda_runtime_api.h>
 #endif  // TRTIS_ENABLE_GPU
 

src/backends/onnx/onnx_backend.h

@@ -25,7 +25,7 @@
 // OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #pragma once
 
-#include <core/session/onnxruntime_c_api.h>
+#include <onnxruntime_c_api.h>
 #include "src/core/backend.h"
 #include "src/core/model_config.pb.h"
 #include "src/core/scheduler.h"

src/backends/onnx/onnx_utils.h

@@ -26,7 +26,7 @@
 
 #pragma once
 
-#include <core/session/onnxruntime_c_api.h>
+#include <onnxruntime_c_api.h>
 #include "src/core/model_config.h"
 #include "src/core/status.h"