diff --git a/paddle/framework/lod_tensor_test.cu b/paddle/framework/lod_tensor_test.cu
index 97e69cdb2e5e1e64031c899f5e04020665485ba8..647d07536dd070bc37137fc01f683ec07ba7d6f4 100644
--- a/paddle/framework/lod_tensor_test.cu
+++ b/paddle/framework/lod_tensor_test.cu
@@ -36,8 +36,8 @@ TEST(LoDTensor, LoDInGPU) {
   lod_tensor.mutable_data<float>(place);
 
   lod_tensor.set_lod(src_lod);
-  CHECK_EQ(lod_tensor.lod_element(0, 2), 4);
-  CHECK_EQ(lod_tensor.lod_element(0, 4), 8);
+  CHECK_EQ(lod_tensor.lod_element(0, 2), 4UL);
+  CHECK_EQ(lod_tensor.lod_element(0, 4), 8UL);
 
   auto lod = lod_tensor.lod();
 
diff --git a/paddle/gserver/layers/MKLDNNConvLayer.cpp b/paddle/gserver/layers/MKLDNNConvLayer.cpp
index 9a0abd291ae8fae43b0e95c7371f3ce35d1261ec..0d6742e909635c1097b4fe21bbb304f8a71af5cb 100644
--- a/paddle/gserver/layers/MKLDNNConvLayer.cpp
+++ b/paddle/gserver/layers/MKLDNNConvLayer.cpp
@@ -28,7 +28,7 @@ bool MKLDNNConvLayer::init(const LayerMap& layerMap,
   if (!MKLDNNLayer::init(layerMap, parameterMap)) {
     return false;
   }
-  CHECK_EQ(inputLayers_.size(), 1) << "Only support one input layer yet";
+  CHECK_EQ(inputLayers_.size(), 1UL) << "Only support one input layer yet";
   CHECK_EQ(inputLayers_.size(), parameters_.size());
   CHECK(config_.shared_biases()) << "Only support shared biases yet";
 
diff --git a/paddle/gserver/layers/MKLDNNFcLayer.cpp b/paddle/gserver/layers/MKLDNNFcLayer.cpp
index 8cbfbd0d2b9f2149f7c959aec5c4ae1de952f903..e829456d6afd7cc844f752d4571cd9f90c73997f 100644
--- a/paddle/gserver/layers/MKLDNNFcLayer.cpp
+++ b/paddle/gserver/layers/MKLDNNFcLayer.cpp
@@ -28,7 +28,7 @@ bool MKLDNNFcLayer::init(const LayerMap& layerMap,
     return false;
   }
 
-  CHECK_EQ(inputLayers_.size(), 1) << "Only support one input layer yet";
+  CHECK_EQ(inputLayers_.size(), 1UL) << "Only support one input layer yet";
   CHECK_EQ(inputLayers_.size(), parameters_.size());
   CHECK(!parameters_[0]->isSparse()) << "Do not support sparse yet";
 
diff --git a/paddle/gserver/tests/test_CrossEntropyOverBeamGrad.cpp b/paddle/gserver/tests/test_CrossEntropyOverBeamGrad.cpp
index 538d18cdc3d262df0ddb031d9e6b38a3fea57606..c922237d33da5de0ece61df732334bee5592249d 100644
--- a/paddle/gserver/tests/test_CrossEntropyOverBeamGrad.cpp
+++ b/paddle/gserver/tests/test_CrossEntropyOverBeamGrad.cpp
@@ -228,7 +228,7 @@ void genGroundTruth(vector<SingleBeamExpansion>& beamExpansions,
         curBeam.groundTruth[j] = *(start + n);
         curBeam.inBeam[j] = 1;
       } else {
-        CHECK_LE(curBeam.rowIdxInBeam[j] + 1,
+        CHECK_LE((size_t)curBeam.rowIdxInBeam[j] + 1,
                  curBeam.subSeqStartPos.size() - 1);
         int start = curBeam.subSeqStartPos[curBeam.rowIdxInBeam[j]];
         int end = curBeam.subSeqStartPos[curBeam.rowIdxInBeam[j] + 1];
diff --git a/paddle/operators/multiplex_op.cu b/paddle/operators/multiplex_op.cu
index 70e46815fc9148a2530d437d20c14f5d40baa1a4..505776612e7119e568493506b113661a839e5bd1 100644
--- a/paddle/operators/multiplex_op.cu
+++ b/paddle/operators/multiplex_op.cu
@@ -42,7 +42,7 @@ class MultiplexGPUKernel : public framework::OpKernel {
     for (auto i = 0; i < rows; i++) {
       int32_t k = index[i];
       PADDLE_ENFORCE_GE(k, 0, "index must be nonnegative.");
-      PADDLE_ENFORCE_LT(k, ins.size(),
+      PADDLE_ENFORCE_LT((size_t)k, ins.size(),
                         "index exceeds the number of candidate tensors.");
       memory::Copy(place, out->data<T>() + i * cols, place,
                    ins[k]->data<T>() + i * cols, cols * sizeof(T), stream);