batch size > 1

include/tkDNN/Network.h

@@ -62,6 +62,7 @@ public:
     dataDim_t getOutputDim();
 
     bool fp16, dla, int8;
+    int maxBatchSize;
     bool dontLoadWeights;
     std::string fileImgList;
     std::string fileLabelList;

include/tkDNN/NetworkRT.h

@@ -74,11 +74,18 @@ public:
     NetworkRT(Network *net, const char *name);
     virtual ~NetworkRT();
 
+    int getMaxBatchSize() {
+        if(engineRT != nullptr)
+            return engineRT->getMaxBatchSize();
+        else
+            return 0;
+    }
+
     /**
         Do inferece
     */
     dnnType* infer(dataDim_t &dim, dnnType* data);
-    void enqueue();    
+    void enqueue(int batchSize = 1);    
 
     nvinfer1::ILayer* convert_layer(nvinfer1::ITensor *input, Layer *l);
     nvinfer1::ILayer* convert_layer(nvinfer1::ITensor *input, Conv2d *l);

include/tkDNN/pluginsRT/MaxPoolingFixedSizeRT.h

@@ -42,7 +42,7 @@ public:
 
 	virtual int enqueue(int batchSize, const void*const * inputs, void** outputs, void* workspace, cudaStream_t stream) override {
 
-		std::cout<<this->n<<"  "<<this->c<<"  "<<this->h<<"  "<<this->w<<"  "<<this->stride_H<<"  "<<this->stride_W<<"  "<<this->winSize<<"  "<<this->padding<<std::endl;
+		//std::cout<<this->n<<"  "<<this->c<<"  "<<this->h<<"  "<<this->w<<"  "<<this->stride_H<<"  "<<this->stride_W<<"  "<<this->winSize<<"  "<<this->padding<<std::endl;
 		dnnType *srcData = (dnnType*)reinterpret_cast<const dnnType*>(inputs[0]);
 		dnnType *dstData = reinterpret_cast<dnnType*>(outputs[0]);
 		MaxPoolingForward(srcData, dstData, this->n, this->c, this->h, this->w, this->stride_H, this->stride_W, this->winSize, this->padding);