diff --git a/src/operator/quantization/mkldnn/mkldnn_quantized_fully_connected.cc b/src/operator/quantization/mkldnn/mkldnn_quantized_fully_connected.cc
index 71daf2ec2c16..cf3d789e2882 100644
--- a/src/operator/quantization/mkldnn/mkldnn_quantized_fully_connected.cc
+++ b/src/operator/quantization/mkldnn/mkldnn_quantized_fully_connected.cc
@@ -74,7 +74,7 @@ void MKLDNNQuantizedFullyConnectedForward(const nnvm::NodeAttrs &attrs,
     int32_t *quantized_bias_ptr = quantized_bias.data().dptr<int32_t>();
     size_t bias_size = bias.shape().Size();
     #pragma omp parallel for num_threads(engine::OpenMP::Get()->GetRecommendedOMPThreadCount())
-    for (size_t i = 0; i < bias_size; ++i) {
+    for (index_t i = 0; i < static_cast<index_t>(bias_size); ++i) {
       quantized_bias_ptr[i] = bias_ptr[i] * bias_int32_rescale;
     }
   }
diff --git a/src/operator/subgraph/mkldnn/mkldnn_fc.cc b/src/operator/subgraph/mkldnn/mkldnn_fc.cc
index 0ec05a2af087..857a27d9a134 100644
--- a/src/operator/subgraph/mkldnn/mkldnn_fc.cc
+++ b/src/operator/subgraph/mkldnn/mkldnn_fc.cc
@@ -156,7 +156,7 @@ void SgMKLDNNFCOp::Forward(const OpContext &ctx,
         int32_t *quantized_bias_ptr = cached_bias_.data().dptr<int32_t>();
         size_t bias_size = bias.shape().Size();
         #pragma omp parallel for num_threads(engine::OpenMP::Get()->GetRecommendedOMPThreadCount())
-        for (size_t i = 0; i < bias_size; ++i) {
+        for (index_t i = 0; i < static_cast<index_t>(bias_size); ++i) {
           quantized_bias_ptr[i] = bias_ptr[i] * bias_int32_rescale;
         }
       }