diff --git a/paddle/optimizer/sgd_optmizer.cc b/paddle/optimizer/sgd_optmizer.cc
index f4fa7756eabbf6a313cae3eddf1f3db49a83152e..a222672815bfd9a84138dd3db4218a12632aecd7 100644
--- a/paddle/optimizer/sgd_optmizer.cc
+++ b/paddle/optimizer/sgd_optmizer.cc
@@ -5,7 +5,6 @@ namespace paddle {
 namespace optimizer {
 
 void SGDOptimizer::set_weight(Tensor *p) {
-  //  ParameterOptimizer::set_weight(p);
   parameter_ = p;
   size_t size = p->size();
   // TODO: fix it with align aware allocator bind to Tensor
@@ -44,8 +43,6 @@ const char *SGDOptimizer::SerializeState() {
   state.set_version(kOptimizerVersion);
   TensorToProto(*parameter_, state.add_data());
   TensorToProto(*momentums_, state.add_data());
-  // state.add_data(param_proto);
-  // state.add_data(momentum_proto);
   state.add_hyperparam(momentum_);
   return state.SerializeAsString().c_str();
 }
diff --git a/proto/OptimizerConfig.proto b/proto/OptimizerConfig.proto
index f492364a5aadf1e352756ad706da6344799e4d53..16b041a9b257d11cbaa2e07c265053ecdedcaa71 100644
--- a/proto/OptimizerConfig.proto
+++ b/proto/OptimizerConfig.proto
@@ -87,7 +87,6 @@ message OptimizerState {
 
 message OptimizerConfig {
   // common config of optimizer
-  // algorithm config, type : string
   enum Optimizer {
    SGD = 1;
    Adadelta = 2;
@@ -100,7 +99,6 @@ message OptimizerConfig {
   optional AdagradConfig adagrad = 5;
   optional AdamConfig adam = 6;
 
-  // learning rate runtime policy config
   enum LrPolicy {
    ConstLr = 0;
    LinearLr = 1;