diff --git a/llvm/projects/hpvm-tensor-rt/dnn_sources/include/utils.h b/llvm/projects/hpvm-tensor-rt/dnn_sources/include/utils.h
index 750891d138f1506d9ad557adc9a440708f8382b4..6cf1cc90f4f3003efc939879c9679fb3dea3e2c6 100644
--- a/llvm/projects/hpvm-tensor-rt/dnn_sources/include/utils.h
+++ b/llvm/projects/hpvm-tensor-rt/dnn_sources/include/utils.h
@@ -225,14 +225,16 @@ void* readInputTensor(const char* file_name, int data_type, int dim1_size, int d
 
 
 struct Tensor* readTrainedWeights(const char* file_name, int data_type,
-				  int dim1_size, int dim2_size,
-				  int dim3_size, int dim4_size){
+				  long int dim1_size, long int dim2_size,
+				  long int dim3_size, long int dim4_size){
 
   // FIXIT: Don't assume floating point types
   int type_size = 4; // NOTE: Assuming floating point tensors
   long int num_elems = dim1_size * dim2_size * dim3_size * dim4_size;
   long int size_in_bytes = type_size * dim1_size * dim2_size * dim3_size * dim4_size;
   float* tensor_data = (float*) malloc(sizeof(float) * num_elems);
+  printf("size_in_bytes  = %lu \n", size_in_bytes);
+  
   int file_header_size = 0;
   
   FILE* file = fopen(file_name, "rb");
@@ -578,5 +580,12 @@ float computePSNRViolation(void* gold_ptr, void* approx_ptr, float PSNR_threshol
 }
 
 
+void dumpOutput(void* output){
+
+  
+
+}
+
+
 
 #endif
diff --git a/llvm/projects/hpvm-tensor-rt/dnn_sources/src/pipeline.cc b/llvm/projects/hpvm-tensor-rt/dnn_sources/src/pipeline.cc
index c87ff87962bb0e5a3bfd34116f528ca4fb2ded80..0c74f130144488ce3901d450b24a234aefb1209c 100644
--- a/llvm/projects/hpvm-tensor-rt/dnn_sources/src/pipeline.cc
+++ b/llvm/projects/hpvm-tensor-rt/dnn_sources/src/pipeline.cc
@@ -26,16 +26,17 @@ void testLenetTanh(){
   printf("********* Pipeline: Gaussian - Outline - Motion Blur - Emboss ********** \n");
   // FIXIT: Extend this to batch of images - currently 5 images
 
-  int test_batch_size = 9145;
-  int H = 240;
-  int W = 300;
+  //long int test_batch_size = 9145;
+  long int test_batch_size = 2000;
+  long int H = 240;
+  long int W = 300;
 
   printf("Reading input\n");
-  void* input = readTrainedWeights("../pipeline/dataset/caltech101_255_float32.bin",
+  void* input = readTrainedWeights("../model_params/pipeline/dataset/caltech101_255_float32.bin",
                                         float_type,
                                         test_batch_size, 1, H, W);
   printf("Reading golden output\n");
-  void* golden_output = readTrainedWeights("../pipeline/golden_output/caltech-gaussian-outline-motionblur-emboss.bin",
+  void* golden_output = readTrainedWeights("../model_params/pipeline/golden_output/caltech-gaussian-outline-motionblur-emboss.bin",
                                         float_type,
                                         test_batch_size, 1, H, W);
 
@@ -43,19 +44,19 @@ void testLenetTanh(){
   // NOTE: Filter descriptors do NOT have batch size
   // NOTE: First two dims are output channels (configurable), input channels (MUST match input channels)
   // IMP: The output channels matches the trained model - not the Lenet arch proposed in Andrew Ng's class
-  void* gaussian_filter = readTrainedWeights("../pipeline/filters/GaussianFilter.bin",
+  void* gaussian_filter = readTrainedWeights("../model_params/pipeline/filters/GaussianFilter.bin",
 					  float_type, 1, 1, 9, 9);    
-  void* outline_filter = readTrainedWeights("../pipeline/filters/OutlineFilter.bin",
+  void* outline_filter = readTrainedWeights("../model_params/pipeline/filters/OutlineFilter.bin",
 					  float_type, 1, 1, 3, 3);    
-  void* sharpen_filter = readTrainedWeights("../pipeline/filters/SharpenFilter.bin",
+  void* sharpen_filter = readTrainedWeights("../model_params/pipeline/filters/SharpenFilter.bin",
 					  float_type, 1, 1, 3, 3);    
-  void* motionblur_filter = readTrainedWeights("../pipeline/filters/MotionblurFilter.bin",
+  void* motionblur_filter = readTrainedWeights("../model_params/pipeline/filters/MotionblurFilter.bin",
 					  float_type, 1, 1, 9, 9);    
   //void* conv1_bias = readTrainedWeights("../model_params/lenet_keras/conv1_bias.bin",
 					//float_type, 1, 32, 1, 1);  
-  void* emboss_filter = readTrainedWeights("../pipeline/filters/EmbossFilter.bin",
+  void* emboss_filter = readTrainedWeights("../model_params/pipeline/filters/EmbossFilter.bin",
 					  float_type, 1, 1, 5, 5);  
-  void* emboss_bias = readTrainedWeights("../pipeline/filters/EmbossBias.bin",
+  void* emboss_bias = readTrainedWeights("../model_params/pipeline/filters/EmbossBias.bin",
 					  float_type, 1, 1, 1, 1);  
   
   clearTensorMap();