Merge branch 'master' into feature/20191121-batchinv-seno

CMakeLists.txt

@@ -87,20 +87,21 @@ endif()
 include(${PROJECT_SOURCE_DIR}/build-tools/cmake/get_cuda_version.cmake)
 
 set(CPACK_PACKAGE_NAME "nnabla-cpplib-cuda${CUDA_VERSION}-cudnn${CUDNN_VERSION}")
-find_program(LSB_RELEASE lsb_release)
-message("LSB_RELEASE" ${LSB_RELEASE})
-if(EXISTS ${LSB_RELEASE})
-
-  execute_process(COMMAND ${LSB_RELEASE} -si
-    OUTPUT_VARIABLE LSB_RELEASE_ID
-    OUTPUT_STRIP_TRAILING_WHITESPACE
-    )
-  execute_process(COMMAND ${LSB_RELEASE} -sr
-    OUTPUT_VARIABLE LSB_RELEASE_RELEASE
-    OUTPUT_STRIP_TRAILING_WHITESPACE
-    )
-  set(CPACK_SYSTEM_NAME ${LSB_RELEASE_ID}${LSB_RELEASE_RELEASE})
+if(UNIX)
+  find_program(LSB_RELEASE lsb_release)
+  if(EXISTS ${LSB_RELEASE})
+    execute_process(COMMAND ${LSB_RELEASE} -si
+      OUTPUT_VARIABLE LSB_RELEASE_ID
+      OUTPUT_STRIP_TRAILING_WHITESPACE
+      )
+    execute_process(COMMAND ${LSB_RELEASE} -sr
+      OUTPUT_VARIABLE LSB_RELEASE_RELEASE
+      OUTPUT_STRIP_TRAILING_WHITESPACE
+      )
+    set(CPACK_SYSTEM_NAME ${LSB_RELEASE_ID}${LSB_RELEASE_RELEASE})
+  endif()
 endif()
+
 set(CPACK_PACKAGE_VERSION_MAJOR ${version_major})
 set(CPACK_PACKAGE_VERSION_MINOR ${version_minor})
 set(CPACK_PACKAGE_VERSION_PATCH ${version_patch})

build-tools/code_generator/function_types.yaml

@@ -304,9 +304,9 @@ Broadcast:
 Tile:
   float: [float]
   half: [Half]
-# OneHot:
-#   float: [int, float]
-#   half: [int, Half]
+OneHot:
+  float: [int, float]
+  half: [int, Half]
 Flip:
   float: [float]
   half: [Half]
@@ -354,9 +354,9 @@ RandomChoice:
 # RandomCrop:
 #   float: [float]
 #   half: [Half]
-# RandomFlip:
-#   float: [float]
-#   half: [Half]
+RandomFlip:
+  float: [float]
+  half: [Half]
 # RandomShift:
 #   float: [float]
 #   half: [Half]
@@ -420,6 +420,9 @@ FixedPointQuantize:
 Pow2Quantize:
   float: [float]
   half: [Half]
+MinMaxQuantize:
+  float: [float]
+  half: [Half]
 TopNError:
   float: [float, int]
   half: [Half, int]

build-tools/code_generator/generate.py

@@ -26,17 +26,12 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import io
-import os
 from os.path import abspath, dirname, join
 import sys
 
 # Set path to <NNabla root>/build-tools/code_generator to import the following two
-from utils.common import check_update
-from utils.type_conv import type_from_proto
 import code_generator_utils as utils
 
-import itertools
 
 here = abspath(dirname(abspath(__file__)))
 base = abspath(here + '/../..')

build-tools/make/build-with-docker.mk

@@ -25,6 +25,7 @@ DOCKER_RUN_OPTS += -e NNABLA_DIRECTORY=$(NNABLA_DIRECTORY)
 NNABLA_EXT_CUDA_DIRECTORY ?= $(shell pwd)
 DOCKER_RUN_OPTS += -e NNABLA_EXT_CUDA_DIRECTORY=$(NNABLA_EXT_CUDA_DIRECTORY)
 DOCKER_RUN_OPTS += -e CMAKE_OPTS=$(CMAKE_OPTS)
+DOCKER_RUN_OPTS += -e INCLUDE_CUDA_CUDNN_LIB_IN_WHL=$(INCLUDE_CUDA_CUDNN_LIB_IN_WHL)
 
 include $(NNABLA_EXT_CUDA_DIRECTORY)/build-tools/make/options.mk
 ifndef NNABLA_BUILD_INCLUDED
@@ -35,7 +36,10 @@ ifndef NNABLA_BUILD_WITH_DOCKER_INCLUDED
   include $(NNABLA_DIRECTORY)/build-tools/make/build-with-docker.mk
 endif
 
+NVIDIA_DOCKER_WRAPPER=$(NNABLA_EXT_CUDA_DIRECTORY)/build-tools/scripts/nvidia-docker.sh
+
 CUDA_SUFFIX = $(CUDA_VERSION_MAJOR)$(CUDA_VERSION_MINOR)-cudnn$(CUDNN_VERSION)
+CUDA_VERSION_MAJOR_MINOR = $(CUDA_VERSION_MAJOR).$(CUDA_VERSION_MINOR)
 
 DOCKER_IMAGE_ID_BUILD_NNABLA_EXT_CUDA = $(shell md5sum $(NNABLA_EXT_CUDA_DIRECTORY)/docker/development/Dockerfile.build$(ARCH_SUFFIX) |cut -d \  -f 1)
 DOCKER_IMAGE_BUILD_NNABLA_EXT_CUDA ?= $(DOCKER_IMAGE_NAME_BASE)-build-cuda$(CUDA_SUFFIX)$(ARCH_SUFFIX):$(DOCKER_IMAGE_ID_BUILD_NNABLA_EXT_CUDA)
@@ -64,6 +68,8 @@ docker_image_build_cuda$(DOCKER_IMAGE_TARGET_SUFFIX):
 		docker pull $(DOCKER_IMAGE_BUILD_CUDA_BASE) && \
 		(cd $(NNABLA_EXT_CUDA_DIRECTORY) && docker build $(DOCKER_BUILD_ARGS)\
 			--build-arg BASE=$(DOCKER_IMAGE_BUILD_CUDA_BASE) \
+			--build-arg CUDA_VERSION_MAJOR_MINOR=$(CUDA_VERSION_MAJOR_MINOR) \
+			--build-arg ARCH_SUFFIX=$(ARCH_SUFFIX) \
 			-t $(DOCKER_IMAGE_BUILD_NNABLA_EXT_CUDA) \
 			-f docker/development/Dockerfile.build$(ARCH_SUFFIX) \
 			.) \
@@ -75,6 +81,8 @@ docker_image_build_cuda_multi_gpu$(DOCKER_IMAGE_TARGET_SUFFIX):
 		docker pull $(DOCKER_IMAGE_BUILD_CUDA_MULTI_GPU_BASE) && \
 		(cd $(NNABLA_EXT_CUDA_DIRECTORY) && docker build $(DOCKER_BUILD_ARGS) \
 			--build-arg BASE=$(DOCKER_IMAGE_BUILD_CUDA_MULTI_GPU_BASE) \
+			--build-arg CUDA_VERSION_MAJOR_MINOR=$(CUDA_VERSION_MAJOR_MINOR) \
+			--build-arg ARCH_SUFFIX=$(ARCH_SUFFIX) \
 			-t $(DOCKER_IMAGE_BUILD_NNABLA_EXT_CUDA_MULTI_GPU) \
 			-f docker/development/Dockerfile.build-multi-gpu$(ARCH_SUFFIX) \
 			.) \
@@ -147,17 +155,17 @@ bwd-nnabla-ext-cuda-wheel-multi-gpu-only: docker_image_build_cuda_multi_gpu
 .PHONY: bwd-nnabla-ext-cuda-test
 bwd-nnabla-ext-cuda-test: docker_image_build_cuda
 	cd $(NNABLA_EXT_CUDA_DIRECTORY) \
-	&& nvidia-docker run $(DOCKER_RUN_OPTS) $(DOCKER_IMAGE_BUILD_NNABLA_EXT_CUDA) make -f build-tools/make/build.mk nnabla-ext-cuda-test-local
+	&& ${NVIDIA_DOCKER_WRAPPER} run $(DOCKER_RUN_OPTS) $(DOCKER_IMAGE_BUILD_NNABLA_EXT_CUDA) make -f build-tools/make/build.mk nnabla-ext-cuda-test-local
 
 .PHONY: bwd-nnabla-ext-cuda-multi-gpu-test
 bwd-nnabla-ext-cuda-multi-gpu-test: docker_image_build_cuda_multi_gpu
 	cd $(NNABLA_EXT_CUDA_DIRECTORY) \
-	&& nvidia-docker run $(DOCKER_RUN_OPTS) $(DOCKER_IMAGE_BUILD_NNABLA_EXT_CUDA_MULTI_GPU) make -f build-tools/make/build.mk nnabla-ext-cuda-multi-gpu-test-local
+	&& ${NVIDIA_DOCKER_WRAPPER} run $(DOCKER_RUN_OPTS) $(DOCKER_IMAGE_BUILD_NNABLA_EXT_CUDA_MULTI_GPU) make -f build-tools/make/build.mk nnabla-ext-cuda-multi-gpu-test-local
 
 .PHONY: bwd-nnabla-ext-cuda-shell
 bwd-nnabla-ext-cuda-shell: docker_image_build_cuda
 	cd $(NNABLA_EXT_CUDA_DIRECTORY) \
-	&& nvidia-docker run $(DOCKER_RUN_OPTS) -it --rm ${DOCKER_IMAGE_BUILD_NNABLA_EXT_CUDA} make nnabla-ext-cuda-shell
+	&& ${NVIDIA_DOCKER_WRAPPER} run $(DOCKER_RUN_OPTS) -it --rm ${DOCKER_IMAGE_BUILD_NNABLA_EXT_CUDA} make nnabla-ext-cuda-shell
 
 ########################################################################################################################
 # Docker image with current nnabla
@@ -173,7 +181,10 @@ docker_image_nnabla_ext_cuda:
 	&& cp $(BUILD_EXT_CUDA_DIRECTORY_WHEEL)/dist/*.whl . \
 	&& echo ADD $(shell basename $(BUILD_EXT_CUDA_DIRECTORY_WHEEL)/dist/*.whl) /tmp/ >>Dockerfile \
 	&& echo RUN pip install /tmp/$(shell basename $(BUILD_EXT_CUDA_DIRECTORY_WHEEL)/dist/*.whl) >>Dockerfile \
-	&& docker build --build-arg BASE=$${BASE} $(DOCKER_BUILD_ARGS) -t $(DOCKER_IMAGE_NNABLA_EXT_CUDA) . \
+	&& docker build --build-arg BASE=$${BASE} $(DOCKER_BUILD_ARGS) \
+		--build-arg CUDA_VERSION_MAJOR_MINOR=$(CUDA_VERSION_MAJOR_MINOR) \
+		--build-arg ARCH_SUFFIX=$(ARCH_SUFFIX) \
+		-t $(DOCKER_IMAGE_NNABLA_EXT_CUDA) . \
 	&& rm -f $(shell basename $(BUILD_DIRECTORY_WHEEL)/dist/*.whl) \
 	&& rm -f $(shell basename $(BUILD_EXT_CUDA_DIRECTORY_WHEEL_MULTI_GPU)/dist/*.whl) \
 	&& rm -f Dockerfile
@@ -191,7 +202,10 @@ docker_image_nnabla_ext_cuda_multi_gpu: bwd-nnabla-ext-cuda-wheel-multi-gpu
 	&& cp $(BUILD_EXT_CUDA_DIRECTORY_WHEEL_MULTI_GPU)/dist/*.whl . \
 	&& echo ADD $(shell basename $(BUILD_EXT_CUDA_DIRECTORY_WHEEL_MULTI_GPU)/dist/*.whl) /tmp/ >>Dockerfile \
 	&& echo RUN pip install /tmp/$(shell basename $(BUILD_EXT_CUDA_DIRECTORY_WHEEL_MULTI_GPU)/dist/*.whl) >>Dockerfile \
-	&& docker build --build-arg BASE=$${BASE} $(DOCKER_BUILD_ARGS) -t $(DOCKER_IMAGE_NNABLA_EXT_CUDA_MULTI_GPU) . \
+	&& docker build --build-arg BASE=$${BASE} $(DOCKER_BUILD_ARGS) \
+		--build-arg CUDA_VERSION_MAJOR_MINOR=$(CUDA_VERSION_MAJOR_MINOR) \
+		--build-arg ARCH_SUFFIX=$(ARCH_SUFFIX) \
+		-t $(DOCKER_IMAGE_NNABLA_EXT_CUDA_MULTI_GPU) . \
 	&& rm -f $(shell basename $(BUILD_DIRECTORY_WHEEL)/dist/*.whl) \
 	&& rm -f $(shell basename $(BUILD_EXT_CUDA_DIRECTORY_WHEEL_MULTI_GPU)/dist/*.whl) \
 	&& rm -f Dockerfile

build-tools/make/build.mk

@@ -181,15 +181,17 @@ nnabla-ext-cuda-test:
 nnabla-ext-cuda-test-local: nnabla-install nnabla-ext-cuda-install
 	cd $(BUILD_EXT_CUDA_DIRECTORY_WHEEL) \
 	&& PYTHONPATH=$(NNABLA_EXT_CUDA_DIRECTORY)/python/test \
-		python -m pytest $(NNABLA_DIRECTORY)/python/test
+	&& $(NNABLA_DIRECTORY)/build-tools/make/pytest.sh $(NNABLA_DIRECTORY)/python/test \
+	&& $(NNABLA_DIRECTORY)/build-tools/make/pytest.sh $(NNABLA_EXT_CUDA_DIRECTORY)/python/test
 
 .PHONY: nnabla-ext-cuda-multi-gpu-test-local
 nnabla-ext-cuda-multi-gpu-test-local: nnabla-ext-cuda-multi-gpu-install
 	cd $(BUILD_EXT_CUDA_DIRECTORY_WHEEL_MULTI_GPU) \
 	&& PYTHONPATH=$(NNABLA_EXT_CUDA_DIRECTORY)/python/test:$(NNABLA_DIRECTORY)/python/test \
-		mpiexec -q -n 2 python -m pytest \
+		mpiexec -q -n 2 $(NNABLA_DIRECTORY)/build-tools/make/pytest.sh \
 			--test-communicator \
 			--communicator-gpus=0,1 \
 			$(NNABLA_DIRECTORY)/python/test/communicator
 	cd $(BUILD_EXT_CUDA_DIRECTORY_WHEEL_MULTI_GPU) \
-	&& python -m pytest $(NNABLA_DIRECTORY)/python/test
+	&& $(NNABLA_DIRECTORY)/build-tools/make/pytest.sh $(NNABLA_DIRECTORY)/python/test \
+	&& $(NNABLA_DIRECTORY)/build-tools/make/pytest.sh $(NNABLA_EXT_CUDA_DIRECTORY)/python/test

build-tools/make/options.mk

@@ -26,8 +26,8 @@ DOCKER_RUN_OPTS += -e CUDA_VERSION_MINOR=$(CUDA_VERSION_MINOR)
 export CUDNN_VERSION ?= 7
 DOCKER_RUN_OPTS += -e CUDNN_VERSION=$(CUDNN_VERSION)
 
-export WHL_NO_PREFIX ?= False
-DOCKER_RUN_OPTS += -e WHL_NO_PREFIX=$(WHL_NO_PREFIX)
+export WHL_NO_CUDA_SUFFIX ?= False
+DOCKER_RUN_OPTS += -e WHL_NO_CUDA_SUFFIX=$(WHL_NO_CUDA_SUFFIX)
 
 ifndef NNABLA_OPTIONS_INCLUDED
   include $(NNABLA_DIRECTORY)/build-tools/make/options.mk

build-tools/scripts/nvidia-docker.sh

@@ -0,0 +1,19 @@
+#!/bin/bash
+
+if which nvidia-docker >/dev/null
+then
+    nvidia-docker "$@"
+else
+    if [ "$1" == "run" ]
+    then
+        GPU=\"device=${NV_GPU}\" 
+        if [ "${NV_GPU}" == "" ]
+        then
+            GPU=all
+        fi
+        docker run --gpus=$GPU "${@:2}"
+    else
+        docker "$@"
+    fi
+fi
+

docker/development/Dockerfile.build

@@ -149,6 +149,8 @@ ENV PYVERNAME=${PYTHON_VERSION_MAJOR}.${PYTHON_VERSION_MINOR}
 # Limitation: numpy>=1.17 does not support python2.7
 ADD python/requirements.txt /tmp/deps/
 
+ARG CUDA_VERSION_MAJOR_MINOR
+ARG ARCH_SUFFIX
 RUN umask 0 \
     && cd /tmp/deps \
     && wget https://repo.continuum.io/miniconda/Miniconda3-latest-Linux-x86_64.sh \
@@ -159,6 +161,10 @@ RUN umask 0 \
     && conda activate nnabla-build \
     && pip install numpy\<1.17 \
     && pip install --only-binary -U -r /tmp/deps/requirements.txt \
+    && ( [ "$CUDA_VERSION_MAJOR_MINOR" = "10.0" ] || [ "$CUDA_VERSION_MAJOR_MINOR" = "9.0" ] \
+        && [ "x$ARCH_SUFFIX" = "x" ] \
+        && pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/cuda/${CUDA_VERSION_MAJOR_MINOR} nvidia-dali \
+        || echo "Skip DALI installation (CUDA=$CUDA_VERSION_MAJOR_MINOR ARCH=$ARCH_SUFFIX)" ) \
     && conda clean -y --all \
     && cd / \
     && rm -rf /tmp/*

docker/development/Dockerfile.build-multi-gpu

@@ -79,6 +79,8 @@ ENV PYVERNAME=${PYTHON_VERSION_MAJOR}.${PYTHON_VERSION_MINOR}
 # Limitation: numpy>=1.17 does not support python2.7
 ADD python/requirements.txt /tmp/deps/
 
+ARG CUDA_VERSION_MAJOR_MINOR
+ARG ARCH_SUFFIX
 RUN umask 0 \
     && cd /tmp/deps \
     && wget https://repo.continuum.io/miniconda/Miniconda3-latest-Linux-x86_64.sh \
@@ -89,6 +91,10 @@ RUN umask 0 \
     && conda activate nnabla-build \
     && pip install numpy\<1.17 \
     && pip install --only-binary -U -r /tmp/deps/requirements.txt \
+    && ( [ "$CUDA_VERSION_MAJOR_MINOR" = "10.0" ] || [ "$CUDA_VERSION_MAJOR_MINOR" = "9.0" ] \
+        && [ "x$ARCH_SUFFIX" = "x" ] \
+        && pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/cuda/${CUDA_VERSION_MAJOR_MINOR} nvidia-dali \
+        || echo "Skip DALI installation (CUDA=$CUDA_VERSION_MAJOR_MINOR ARCH=$ARCH_SUFFIX)" ) \
     && conda clean -y --all \
     && cd / \
     && rm -rf /tmp/*

docker/runtime/Dockerfile.runtime

@@ -13,6 +13,8 @@ ARG PYTHON_VERSION_MAJOR
 ARG PYTHON_VERSION_MINOR
 ENV PYVERNAME=${PYTHON_VERSION_MAJOR}.${PYTHON_VERSION_MINOR}
 
+ARG CUDA_VERSION_MAJOR_MINOR
+ARG ARCH_SUFFIX
 RUN umask 0 \
     && cd /tmp/deps \
     && wget https://repo.continuum.io/miniconda/Miniconda3-latest-Linux-x86_64.sh \
@@ -21,6 +23,10 @@ RUN umask 0 \
     && . /opt/miniconda3/bin/activate \
     && conda create -n nnabla-build python=${PYVERNAME} \
     && conda activate nnabla-build \
+    && ( [ "$CUDA_VERSION_MAJOR_MINOR" = "10.0" ] || [ "$CUDA_VERSION_MAJOR_MINOR" = "9.0" ] \
+        && [ "x$ARCH_SUFFIX" = "x" ] \
+        && pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/cuda/${CUDA_VERSION_MAJOR_MINOR} nvidia-dali \
+        || echo "Skip DALI installation (CUDA=$CUDA_VERSION_MAJOR_MINOR ARCH=$ARCH_SUFFIX)" ) \
     && conda clean -y --all \
     && cd / \
     && rm -rf /tmp/*

docker/runtime/Dockerfile.runtime-multi-gpu

@@ -15,6 +15,8 @@ ENV PYVERNAME=${PYTHON_VERSION_MAJOR}.${PYTHON_VERSION_MINOR}
 
 ADD python/requirements.txt /tmp/deps/
 
+ARG CUDA_VERSION_MAJOR_MINOR
+ARG ARCH_SUFFIX
 RUN umask 0 \
     && cd /tmp/deps \
     && wget https://repo.continuum.io/miniconda/Miniconda3-latest-Linux-x86_64.sh \
@@ -23,6 +25,10 @@ RUN umask 0 \
     && . /opt/miniconda3/bin/activate \
     && conda create -n nnabla-build python=${PYVERNAME} \
     && conda activate nnabla-build \
+    && ( [ "$CUDA_VERSION_MAJOR_MINOR" = "10.0" ] || [ "$CUDA_VERSION_MAJOR_MINOR" = "9.0" ] \
+        && [ "x$ARCH_SUFFIX" = "x" ] \
+        && pip install --extra-index-url https://developer.download.nvidia.com/compute/redist/cuda/${CUDA_VERSION_MAJOR_MINOR} nvidia-dali \
+        || echo "Skip DALI installation (CUDA=$CUDA_VERSION_MAJOR_MINOR ARCH=$ARCH_SUFFIX)" ) \
     && pip install --only-binary -U -r /tmp/deps/requirements.txt \
     && conda clean -y --all \
     && cd / \

examples/cpp/mnist_collection/train_dcgan_cuda.cpp

@@ -19,6 +19,7 @@ using namespace std;
 using namespace nbla;
 
 #include <dcgan_training.hpp>
+#include <string.h>
 
 /******************************************/
 // Example of mnist training
@@ -31,8 +32,21 @@ int main(int argc, char *argv[]) {
       {"cudnn:float", "cuda:float", "cpu:float"}, "CudaCachedArray", "0"};
 
   // Execute training
-  if (!dcgan_training(ctx)) {
-    return (-1);
+  if (argc < 2 || strcmp(argv[1], "--static") == 0) {
+    std::cout << "Execute training with static graph" << std::endl;
+    if (!dcgan_training_with_static_graph(ctx)) {
+      return -1;
+    }
+  } else if (strcmp(argv[1], "--dynamic") == 0) {
+    std::cout << "Execute training with dynamic graph" << std::endl;
+    if (!dcgan_training_with_dynamic_graph(ctx)) {
+      return -1;
+    }
+  } else {
+    std::cerr << std::endl;
+    std::cerr << "Usage: " << argv[0] << " --static / --dynamic " << std::endl;
+    std::cerr << std::endl;
+    return -1;
   }
 
   return 0;

examples/cpp/mnist_collection/train_lenet_classifier_cuda.cpp

@@ -19,6 +19,7 @@ using namespace std;
 using namespace nbla;
 
 #include <lenet_training.hpp>
+#include <string.h>
 
 /******************************************/
 // Example of lenet training
@@ -31,8 +32,21 @@ int main(int argc, char *argv[]) {
       {"cudnn:float", "cuda:float", "cpu:float"}, "CudaCachedArray", "0"};
 
   // Execute training
-  if (!lenet_training(ctx)) {
-    return (-1);
+  if (argc < 2 || strcmp(argv[1], "--static") == 0) {
+    std::cout << "Execute training with static graph" << std::endl;
+    if (!lenet_training_with_static_graph(ctx)) {
+      return -1;
+    }
+  } else if (strcmp(argv[1], "--dynamic") == 0) {
+    std::cout << "Execute training with dynamic graph" << std::endl;
+    if (!lenet_training_with_dynamic_graph(ctx)) {
+      return -1;
+    }
+  } else {
+    std::cerr << std::endl;
+    std::cerr << "Usage: " << argv[0] << " --static / --dynamic " << std::endl;
+    std::cerr << std::endl;
+    return -1;
   }
 
   return 0;

examples/cpp/mnist_collection/train_resnet_classifier_cuda.cpp

@@ -19,6 +19,7 @@ using namespace std;
 using namespace nbla;
 
 #include <resnet_training.hpp>
+#include <string.h>
 
 /******************************************/
 // Example of resnet training
@@ -31,8 +32,21 @@ int main(int argc, char *argv[]) {
       {"cudnn:float", "cuda:float", "cpu:float"}, "CudaCachedArray", "0"};
 
   // Execute training
-  if (!resnet_training(ctx)) {
-    return (-1);
+  if (argc < 2 || strcmp(argv[1], "--static") == 0) {
+    std::cout << "Execute training with static graph" << std::endl;
+    if (!resnet_training_with_static_graph(ctx)) {
+      return -1;
+    }
+  } else if (strcmp(argv[1], "--dynamic") == 0) {
+    std::cout << "Execute training with dynamic graph" << std::endl;
+    if (!resnet_training_with_dynamic_graph(ctx)) {
+      return -1;
+    }
+  } else {
+    std::cerr << std::endl;
+    std::cerr << "Usage: " << argv[0] << " --static / --dynamic " << std::endl;
+    std::cerr << std::endl;
+    return -1;
   }
 
   return 0;