clean the code

source/network/XBackwardMath.cpp

@@ -843,7 +843,6 @@ void XMathGrad::GradMultiplyBroadcast(XTensor * node, bool isEfficient)
 
     XTensor * a = income.tails[0];
     XTensor * b = income.tails[1];
-    DTYPE beta = income.GetParam(0);
 
     XNoder::MakeGrad(a);
     _MultiplyBroadcast(node->grad, b, a->grad, 1.0F);
@@ -1311,7 +1310,7 @@ void XMathGrad::GradSumBroadcast(XTensor * node, bool isEfficient)
 
     XTensor * a = income.tails[0];
     XTensor * b = income.tails[1];
-    DTYPE beta = income.GetParam(0);
+    //DTYPE beta = income.GetParam(0);
 
     XNoder::MakeGrad(a);
     _Sum(a->grad, node->grad, a->grad);

source/sample/transformer/T2TTrainer.cpp

@@ -150,7 +150,6 @@ void T2TTrainer::Train(const char * fn, const char * validFN, const char * model
     int wc = 0;
     int ws =0;
     int wordCount = 0;
-    int totalW;
     int wordCountTotal = 0;
     int wordCountBatch = 0;
     bool isEnd = false;
@@ -590,7 +589,6 @@ int T2TTrainer::LoadBuf(FILE * file, bool isSorted, int step)
         /* distribute samples into buckets. In each bucket, sequences have
            similar a length */
         if (bucketSize > 0) {
-            int bucketCount = 0;
             int low = 0;
             int high = low + bucketSize;
             int n = count - 1;
@@ -947,7 +945,6 @@ int T2TTrainer::LoadBatchMT(FILE * file,
 
     int sCount = sc/2;
     int seqSize = 0;
-    int dimsDec[3] = {sCount, maxDec, vsDec};
 
     InitTensor2D(batchEnc, sCount, maxEnc, X_INT, devID, mem);
     InitTensor2D(paddingEnc, sCount, maxEnc, X_FLOAT, devID, mem);
@@ -966,7 +963,6 @@ int T2TTrainer::LoadBatchMT(FILE * file,
     int wCountEnc = 0;
     int wCountDec = 0;
     int wCountPad = 0;
-    int wGold = 0;
     wCount = 0;
 
     int * batchEncValues = new int[batchEnc->unitNum];

source/sample/transformer/Transformer.cpp

@@ -25,6 +25,7 @@
 #include "T2TModel.h"
 #include "T2TUtility.h"
 #include "T2TTrainer.h"
+#include "T2TSearcher.h"
 #include "../../tensor/XDevice.h"
 #include "../../tensor/XUtility.h"
 #include "../../tensor/XGlobal.h"

source/tensor/core/getandset/OnehotAndIndex.cpp

@@ -111,10 +111,9 @@ void _IndexToOnehot(XTensor * index, XTensor * onehot, int size, float labelSmoo
 
     onehot->SetZeroAll();
 
+#ifdef USE_CUDA
     float confidence = 1 - labelSmoothingP;
     float lowconfidence = labelSmoothingP / size;
-
-#ifdef USE_CUDA
     if(onehot->devID >= 0 && index->devID >= 0) {
         _CudaIndexToOnehot(index, onehot, size, confidence, lowconfidence);
         return;

source/tensor/core/reduce/ReduceSumAll.cpp

@@ -62,7 +62,7 @@ DTYPE _ReduceSumAll(const XTensor * source)
     
     return r;
 
-    int order = source->order;
+    /*int order = source->order;
     DTYPE summation;
 
     XTensor * big = NewTensor(source);
@@ -90,7 +90,7 @@ DTYPE _ReduceSumAll(const XTensor * source)
     summation = big->Get1D(0);
     delete big;
 
-    return summation;
+    return summation;*/
 }
 
 /*