Backport of updates from release 2021.3 branch (openvinotoolkit#72)

* Azure CI: Add "ref: releases/2021/3" * Fix Klocwork issues in Java * Updated arm plugin build script (openvinotoolkit#65) * Fix some Klocwork issues * Update README.md (openvinotoolkit#76) * Fixed Convolution fusion (openvinotoolkit#84) * Remove MO deps from mo_pytorch module (openvinotoolkit#93) Co-authored-by: azhogov <[email protected]> Co-authored-by: Dmitry Kurtaev <[email protected]> Co-authored-by: Aleksandr Voron <[email protected]> Co-authored-by: Liubov Batanina <[email protected]>
eshoguli · Mar 25, 2021 · dd55883 · dd55883
1 parent 2fbbd7a
commit dd55883
Show file tree

Hide file tree

Showing 11 changed files with 79 additions and 42 deletions.
diff --git a/.ci/azure/linux.yml b/.ci/azure/linux.yml
@@ -121,6 +121,7 @@ jobs:
   - script: |
       python3 -m pip install -U setuptools==51.0.0
       python3 -m pip install -r requirements.txt
+      python3 -m pip install -r $(OPENVINO_REPO_DIR)/model-optimizer/requirements_onnx.txt
     workingDirectory: $(REPO_DIR)/modules/mo_pytorch/test
     displayName: 'Install PyTorch tests dependencies'
 

diff --git a/modules/arm_plugin/README.md b/modules/arm_plugin/README.md
@@ -43,7 +43,7 @@ docker container run --rm -ti -v $PWD/build:/armcpu_plugin arm-plugin
 or export only the archive with artifacts:
 ```
 docker container run --rm -ti --tmpfs /armcpu_plugin:rw -v $PWD:/remote \
-                     arm-plugin sh -c "/armplg_build.sh && cp ./OV_ARM_package.tar.gz /remote"
+                     arm-plugin sh -c "sh /armplg_build.sh && cp ./OV_ARM_package.tar.gz /remote"
 ```
 > **NOTE**: There are a few environment variables that control `/armplg_build.sh` script execution.
 >
@@ -189,16 +189,16 @@ On the output video you should see people enclosed in red rectangles:
 ![](docs/img/object_detection_demo_yolo.gif)
 
 ### OpenVINO™ samples
-You could verify the plugin by running [OpenVINO™ samples]. You can find C++ samples in `deployment_tools/inference_engine/bin` directory (if you build the plugin using approach #1) or `openvino/bin/armv7l/Release` directory (if you build the plugin using approach #2 or #3). The following procedure assumes the approach #1 is used.  
-OpenVINO™ samples require OpenCV libraries. If you build the plugin using approach #1 all needed OpenCV libraries are already placed in `build\lib` directory. If you build the plugin using approach #2 or #3 you need to install OpenCV or [build it from source].  
+You could verify the plugin by running [OpenVINO™ samples]. You can find C++ samples in `deployment_tools/inference_engine/bin/<platform_type>` directory (if you build the plugin using approach #1) or `openvino/bin/<platform_type>/Release` directory (if you build the plugin using approach #2 or #3). The following procedure assumes the approach #1 is used.  
+OpenVINO™ samples require OpenCV libraries. If you build the plugin using approach #1 all needed OpenCV libraries are already placed in `opencv\lib` directory. If you build the plugin using approach #2 or #3 you need to install OpenCV or [build it from source].  
 Let's try to run [Object Detection for SSD sample].  
 #### Model preparation
 1. Prepare model `vehicle-license-plate-detection-barrier-0123` using Model Preparation precedure described in Open Model Zoo demos section. 
 #### Model inference on ARM
 1. Copy OpenVINO™ and ARM plugin artefacts to ARM platform. If you build the plugin using approach #1, all artefacts are packed into `OV_ARM_package.tar.gz`.
-2. Go to `deployment_tools/inference_engine/bin` directory:
+2. Go to Inference Engine bin directory:
 ```
-cd deployment_tools/inference_engine/bin
+cd deployment_tools/inference_engine/bin/<platform_type>
 ```
 3. Download a vehicle image, for instance, [this image]:
 ```
@@ -208,7 +208,7 @@ wget https://raw.githubusercontent.com/openvinotoolkit/openvino/master/scripts/d
 
 5. Run object detection sample on ARM platform:
 ```
-export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$PWD/lib
+export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:<package_dir>/opencv/lib/:<package_dir>/deployment_tools/inference_engine/lib/<platform_type>/
 ./object_detection_sample_ssd -m vehicle-license-plate-detection-barrier-0123.xml -i car_1.bmp -d CPU
 ```
 

diff --git a/modules/arm_plugin/armplg_build.sh b/modules/arm_plugin/armplg_build.sh
@@ -29,14 +29,27 @@ fail()
     exit $retval
 }
 
+cloneSrcTree()
+{
+    DESTDIR=$1
+    shift
+    SRCURL=$1
+    shift
+    while [ $# -gt 0 ]; do
+        git lfs clone --recurse-submodules --shallow-submodules --depth 1 --branch=$1 $SRCURL $DESTDIR && return 0
+        shift
+    done
+    return 1
+}
+
 checkSrcTree()
 {
     [ $# -lt 3 ] && fail
 
     if ! [ -d $1 ]; then
         echo "Unable to detect $1"
         echo "Cloning $2..."
-        git lfs clone --recurse-submodules --shallow-submodules --depth 1 --branch=$3 $2 $1 || fail 3 "Failed to clone $2. Stopping"
+        cloneSrcTree $@ || fail 3 "Failed to clone $2. Stopping"
     else
         echo "Detected $1"
         echo "Considering it as source directory"
@@ -45,7 +58,7 @@ checkSrcTree()
             echo "Removing existing sources..."
             rm -rf $1 || fail 1 "Failed to remove. Stopping"
             echo "Cloning $2..."
-            git lfs clone --recurse-submodules --shallow-submodules --depth 1 --branch=$3 $2 $1 || fail 3 "Failed to clone $2. Stopping"
+            cloneSrcTree $@ || fail 3 "Failed to clone $2. Stopping"
         elif [ -d $1/build ]; then
             echo "Build directory detected at $1"
             if [ "$UPDATE_SOURCES" = "clean" ]; then

diff --git a/modules/arm_plugin/src/transformations/arm_optimizations.cpp b/modules/arm_plugin/src/transformations/arm_optimizations.cpp
@@ -110,6 +110,7 @@ bool ArmPlugin::pass::ArmOptimizations::run_on_function(std::shared_ptr<ngraph::
     manager.register_pass<pass::ConvertConv1D>();
     manager.register_pass<pass::ConvertGroupConv1D>();
     manager.register_pass<pass::ConvertGroupConvolution>();
+    manager.register_pass<ngraph::pass::ConstantFolding>();
     manager.register_pass<pass::ConvBiasActivationFusion>();
     manager.register_pass<pass::ConvertMatMulToFC>();
     manager.register_pass<pass::ConvertEltwise>();

diff --git a/modules/arm_plugin/src/transformations/conv_bias_activ_fusion.cpp b/modules/arm_plugin/src/transformations/conv_bias_activ_fusion.cpp
@@ -134,7 +134,7 @@ ngraph::matcher_pass_callback ArmPlugin::pass::ConvBiasFusionBase::fuse_conv_wit
         }
 
         if (!std::dynamic_pointer_cast<opset::Constant>(eltwise->input_value(1 - conv_idx).get_node_shared_ptr())) {
-            IE_THROW() << "Unsupported Convolution with inconstant weights.";
+            return false; // Unsupported Convolution with inconstant bias
         }
 
         auto bias = eltwise->input_value(1 - conv_idx);
@@ -160,10 +160,6 @@ ngraph::matcher_pass_callback ArmPlugin::pass::ConvBiasFusionBase::fuse_conv_wit
                     opset::Constant::create(ngraph::element::i64, ngraph::Shape{1}, {channel_dim}), true);
         }
 
-        if (m_conv->output(0).get_target_inputs().size() != 1) {
-            return false;
-        }
-
         if (m_conv->inputs().size() == 3) {
             new_bias = std::make_shared<opset::Add>(new_bias, m_conv->input_value(Inputs::Bias));
         }
@@ -194,6 +190,10 @@ ngraph::matcher_pass_callback ArmPlugin::pass::ConvertConvBase::convert_conv_to_
             return false;
         }
 
+        if (!std::dynamic_pointer_cast<opset::Constant>(m_conv->input_value(Inputs::Weights).get_node_shared_ptr())) {
+            IE_THROW() << "Unsupported Convolution with inconstant weights.";
+        }
+
         auto conv_arm = std::make_shared<ArmConv>(
                     m_conv->input_value(Inputs::Data),
                     m_conv->input_value(Inputs::Weights),
@@ -233,17 +233,17 @@ ArmPlugin::pass::ConvertGroupConvolutionToArm::ConvertGroupConvolutionToArm() {
 
 ArmPlugin::pass::ConvBiasFusion::ConvBiasFusion() {
     auto m = std::make_shared<ngraph::pattern::Matcher>(
-            ngraph::pattern::wrap_type<opset::ArmConvolution>({ngraph::pattern::any_input(ngraph::pattern::has_static_shape()),
-                                                               ngraph::pattern::any_input(ngraph::pattern::has_static_shape())},
-                                                               ngraph::pattern::has_static_shape()), "ConvBiasFusion");
+            ngraph::pattern::wrap_type<opset::Add>({ngraph::pattern::any_input(ngraph::pattern::has_static_shape()),
+                                                    ngraph::pattern::any_input(ngraph::pattern::has_static_shape())},
+                                                    ngraph::pattern::has_static_shape()), "ConvBiasFusion");
     register_matcher(m, fuse_conv_with_bias<opset::ArmConvolution>());
 }
 
 ArmPlugin::pass::GroupConvBiasFusion::GroupConvBiasFusion() {
     auto m = std::make_shared<ngraph::pattern::Matcher>(
-                ngraph::pattern::wrap_type<opset::ArmGroupConvolution>({ngraph::pattern::any_input(ngraph::pattern::has_static_shape()),
-                                                                        ngraph::pattern::any_input(ngraph::pattern::has_static_shape())},
-                                                                        ngraph::pattern::has_static_shape()), "GroupConvBiasFusion");
+                ngraph::pattern::wrap_type<opset::Add>({ngraph::pattern::any_input(ngraph::pattern::has_static_shape()),
+                                                        ngraph::pattern::any_input(ngraph::pattern::has_static_shape())},
+                                                        ngraph::pattern::has_static_shape()), "GroupConvBiasFusion");
     register_matcher(m, fuse_conv_with_bias<opset::ArmGroupConvolution>());
 }
 

diff --git a/modules/java_api/cpp/blob.cpp b/modules/java_api/cpp/blob.cpp
@@ -162,14 +162,15 @@ JNIEXPORT jlong JNICALL Java_org_intel_openvino_Blob_BlobLong(JNIEnv *env, jobje
 JNIEXPORT jlong JNICALL Java_org_intel_openvino_Blob_BlobCArray(JNIEnv *env, jobject obj, jlong tensorDescAddr, jlong matDataAddr)
 {
     static const char method_name[] = "BlobCArray";
+    Blob::Ptr *blob = nullptr;
     try
     {
         TensorDesc *tDesc = (TensorDesc *)tensorDescAddr;
 
         auto precision = tDesc->getPrecision();
 
         std::vector<size_t> dims = tDesc->getDims();
-        Blob::Ptr *blob = new Blob::Ptr();
+        blob = new Blob::Ptr();
 
         switch (precision) {
             case Precision::FP32:
@@ -210,18 +211,26 @@ JNIEXPORT jlong JNICALL Java_org_intel_openvino_Blob_BlobCArray(JNIEnv *env, job
                 *blob = make_shared_blob<short>((*tDesc), data);
                 break;
             }
-            default:
+            default: {
+                delete blob;
                 throw std::runtime_error("Unsupported precision value!");
+            }
         }
 
         return (jlong)blob;
     }
     catch (const std::exception &e)
     {
+        if (blob) {
+            delete blob;
+        }
         throwJavaException(env, &e, method_name);
     }
     catch (...)
     {
+        if (blob) {
+            delete blob;
+        }
         throwJavaException(env, 0, method_name);
     }
 
@@ -258,8 +267,13 @@ JNIEXPORT jlong JNICALL Java_org_intel_openvino_Blob_rmap(JNIEnv *env, jobject o
         Blob::Ptr *output = reinterpret_cast<Blob::Ptr *>(addr);
 
         if ((*output)->is<MemoryBlob>()) {
-            LockedMemory<const void> *lmem = new LockedMemory<const void> (as<MemoryBlob>(*output)->rmap());
-            return (jlong)lmem;
+            auto mBlob = as<MemoryBlob>(*output);
+            if (mBlob) {
+                LockedMemory<const void> *lmem = new LockedMemory<const void> (mBlob->rmap());
+                return (jlong)lmem;
+            } else {
+                throw std::runtime_error("Target Blob cannot be cast to the MemoryBlob!");
+            }
         } else {
             throw std::runtime_error("Target Blob cannot be cast to the MemoryBlob!");
         }

diff --git a/modules/java_api/cpp/tensor_desc.cpp b/modules/java_api/cpp/tensor_desc.cpp
@@ -45,6 +45,9 @@ JNIEXPORT jintArray JNICALL Java_org_intel_openvino_TensorDesc_GetDims(JNIEnv *e
         std::vector<size_t> size_t_dims = tDesc->getDims();
 
         jintArray result = env->NewIntArray(size_t_dims.size());
+        if (!result) {
+            throw std::runtime_error("Out of memory!");
+        }
         jint *arr = env->GetIntArrayElements(result, nullptr);
 
         for (int i = 0; i < size_t_dims.size(); ++i)

diff --git a/modules/java_api/org/intel/openvino/IEWrapper.java b/modules/java_api/org/intel/openvino/IEWrapper.java
@@ -14,6 +14,7 @@ protected long getNativeObjAddr() {
     @Override
     protected void finalize() throws Throwable {
         delete(nativeObj);
+        super.finalize();
     }
 
     /*----------------------------------- native methods -----------------------------------*/

diff --git a/modules/java_api/samples/ArgumentParser.java b/modules/java_api/samples/ArgumentParser.java
@@ -25,24 +25,24 @@ private void printHelp() {
     }
 
     public void parseArgs(String[] args) {
-        try {
-            for (int i = 0; i < args.length; i++) {
-                String arg = args[i];
-                if (arg.equals("--help") | arg.equals("-h")) {
-                    printHelp();
-                    System.exit(0);
-                } else {
-                    if (description.containsKey(arg)) {
+        for (int i = 0; i < args.length; i++) {
+            String arg = args[i];
+            if (arg.equals("--help") || arg.equals("-h")) {
+                printHelp();
+                System.exit(0);
+            } else {
+                if (description.containsKey(arg)) {
+                    if (i < args.length - 1) {
                         input.put(arg, args[++i]);
                     } else {
-                        System.out.println("Non-existent key: '" + arg + "'");
+                        System.out.println("Missed value for key: '" + arg + "'");
                         System.exit(0);
                     }
+                } else {
+                    System.out.println("Non-existent key: '" + arg + "'");
+                    System.exit(0);
                 }
             }
-        } catch (ArrayIndexOutOfBoundsException e) {
-            System.out.println("Error: Incorrect number of arguments");
-            System.exit(0);
         }
     }
 

diff --git a/modules/java_api/samples/benchmark_app/Main.java b/modules/java_api/samples/benchmark_app/Main.java
@@ -1,10 +1,10 @@
 import org.intel.openvino.*;
 
+import java.security.SecureRandom;
 import java.util.ArrayList;
 import java.util.Arrays;
 import java.util.HashMap;
 import java.util.Map;
-import java.util.Random;
 import java.util.Vector;
 
 public class Main {
@@ -25,6 +25,9 @@ static boolean adjustShapesBatch(
             } else if (layout == Layout.CN) {
                 batchIndex = 1;
             }
+            if (!shapes.containsKey(entry.getKey())) {
+                continue;
+            }
             if ((batchIndex != -1) && (shapes.get(entry.getKey())[batchIndex] != batchSize)) {
                 shapes.get(entry.getKey())[batchIndex] = batchSize;
                 updated = true;
@@ -91,7 +94,7 @@ static Blob blobRandomByte(TensorDesc tDesc) {
         }
 
         byte[] buff = new byte[size];
-        Random rand = new Random();
+        SecureRandom rand = new SecureRandom();
         rand.nextBytes(buff);
 
         return new Blob(tDesc, buff);
@@ -108,7 +111,7 @@ static double getMedianValue(Vector<Double> vec) {
         else return (double) arr[arr.length / 2];
     }
 
-    static boolean getApiBoolean(String api) throws RuntimeException {
+    static boolean getApiBoolean(String api) throws Exception {
         if (api.equals("sync")) return false;
         else if (api.equals("async")) return true;
         else throw new RuntimeException("Incorrect argument: '-api'");
@@ -193,7 +196,7 @@ public static void main(String[] args) {
 
         try {
             isAsync = getApiBoolean(api);
-        } catch (RuntimeException e) {
+        } catch (Exception e) {
             System.out.println(e.getMessage());
             return;
         }
@@ -382,6 +385,10 @@ public static void main(String[] args) {
                 || (durationMs != 0L && execTime < durationMs)
                 || (isAsync && iteration % nireq != 0)) {
             inferRequest = inferRequestsQueue.getIdleRequest();
+            if (inferRequest == null) {
+                System.out.println("No idle Infer Requests!");
+                return;
+            }
 
             if (isAsync) {
                 // As the inference request is currently idle, the wait() adds no additional

diff --git a/modules/mo_pytorch/test/requirements.txt b/modules/mo_pytorch/test/requirements.txt
@@ -1,6 +1,3 @@
-defusedxml>=0.5.0
-networkx>=1.11
-test-generator==0.1.1
 torch==1.7.0
 torchvision==0.8.1
 opencv-python==4.4.0.46