fix

src/Network.cpp

@@ -1,5 +1,4 @@
 #include <iostream>
-#include "NvInfer.h"
 
 #include "tkdnn.h"
 #include "Network.h"
@@ -12,10 +11,9 @@ Network::Network(dataDim_t input_dim) {
 
     float tk_ver = float(tkDNN::getVersion())/1000;
     float cu_ver = float(cudnnGetVersion())/1000;
-    float rt_ver = float(NV_TENSORRT_MAJOR) + float(NV_TENSORRT_MINOR)/10 + float(NV_TENSORRT_PATCH)/100;
 
     std::cout<<"New NETWORK (tkDNN v"<<tk_ver
-             <<", CUDNN v"<<cu_ver<<", TensorRT v"<<rt_ver<<")\n";
+             <<", CUDNN v"<<cu_ver<<")\n";
     dataType = CUDNN_DATA_FLOAT;
     tensorFormat = CUDNN_TENSOR_NCHW;
 
@@ -57,4 +55,4 @@ dataDim_t Network::getOutputDim() {
             return layers[num_layers-1]->output_dim;
 }
 
-}
+}

src/NetworkRT.cpp

@@ -10,12 +10,10 @@ using namespace nvinfer1;
 #include "pluginsRT/RegionRT.cpp"
 
 // Logger for info/warning/errors
-class Logger : public ILogger			
-{
-	void log(Severity severity, const char* msg) override
-	{
-		std::cout <<"TENSORRT: "<< msg << std::endl;
-	}
+class Logger : public ILogger {
+    void log(Severity severity, const char* msg) override {
+        std::cout <<"TENSORRT LOG: "<< msg << std::endl;
+    }
 } loggerRT;
 
 namespace tkDNN {
@@ -24,6 +22,11 @@ std::map<Layer*, nvinfer1::ITensor*>tensors;
 
 NetworkRT::NetworkRT(Network *net) {
 
+    float rt_ver = float(NV_TENSORRT_MAJOR) + 
+                   float(NV_TENSORRT_MINOR)/10 + 
+                   float(NV_TENSORRT_PATCH)/100;
+    std::cout<<"New NetworkRT (TensorRT v"<<rt_ver<<")\n";
+
     builderRT = createInferBuilder(loggerRT);
     networkRT = builderRT->createNetwork();
     dtRT = DataType::kFLOAT;
@@ -250,4 +253,4 @@ ITensor* NetworkRT::convert_layer(ITensor *input, Region *l) {
     return lRT->getOutput(0);
 }
 
-}
+}

src/utils.cpp

@@ -23,9 +23,7 @@ void readBinaryFile(const char* fname, int size, value_type** data_h, value_type
     }
     
     checkCuda( cudaMalloc(data_d, size_b) );
-    checkCuda( cudaMemcpy(*data_d, *data_h,
-                                size_b,
-                                cudaMemcpyHostToDevice) );
+    checkCuda( cudaMemcpy(*data_d, *data_h, size_b, cudaMemcpyHostToDevice) );
 }
 
 void printDeviceVector(int size, value_type* vec_d, bool device)
@@ -39,8 +37,7 @@ void printDeviceVector(int size, value_type* vec_d, bool device)
         vec = vec_d;
     }
     
-    for (int i = 0; i < size; i++)
-    {
+    for (int i = 0; i < size; i++) {
         std::cout << vec[i] << " ";
     }
     std::cout << std::endl;
@@ -52,6 +49,7 @@ void printDeviceVector(int size, value_type* vec_d, bool device)
 int checkResult(int size, value_type *data_d, value_type *correct_d, bool device) {
 
     value_type *data_h, *correct_h;
+    const float eps = 0.0001f;
 
     if(device) {
         data_h = new value_type[size];
@@ -67,10 +65,12 @@ int checkResult(int size, value_type *data_d, value_type *correct_d, bool device
 
     int diffs = 0;
     for(int i=0; i<size; i++) {
-        if(fabs(data_h[i] - correct_h[i]) > 0.0001) {
+        if(fabs(data_h[i] - correct_h[i]) > eps) {
             diffs += 1;
+            if(diffs == 1)
+                std::cout<<"\n";
             if(diffs < 10)
-                printf("%d: %f %f\n", i, data_h[i], correct_h[i]);
+                std::cout<<" | [ "<<i<<" ]: "<<data_h[i]<<" "<<correct_h[i]<<"\n";
         }
     }
 
@@ -79,6 +79,13 @@ int checkResult(int size, value_type *data_d, value_type *correct_d, bool device
         delete [] correct_h;
     }
 
+    std::cout<<" | ";
+    if(diffs == 0)
+        std::cout<<COL_GREENB<<"OK";
+    else
+        std::cout<<COL_REDB<<"Wrongs: "<<diffs;
+
+    std::cout<<COL_END<<" ~"<<eps<<"\n";
     return diffs;
 }
 
@@ -108,4 +115,4 @@ void matrixMulAdd(  cublasHandle_t handle, value_type* srcData, value_type* dstD
     value_type alpha = mul;
     checkERROR( cublasSaxpy(handle, dim, &alpha, srcData, 1, dstData, 1));
 
-}
+}