more accurate to early stop train sparse model

paddle/trainer/Tester.cpp

@@ -46,6 +46,12 @@ Tester::Tester(const std::shared_ptr<TrainerConfigHelper>& config,
       gradientMachine_(gradientMachine),
       parameterUpdater_(parameterUpdater),
       testDataProvider_(testDataProvider) {
+  if (config_->getOptConfig().use_sparse_remote_updater()) {
+    LOG(FATAL) << "It's prohibited to set sparse_remote_update "
+               << "in some layers if testing will be under going "
+               << "in the middle of training. You can do testing "
+               << "within separate process.";
+  }
   testEvaluator_.reset(gradientMachine_->makeEvaluator());
   if (intconfig_->distributeTest) {
     testParameterClient_.reset(new ParameterClient2(true));

paddle/trainer/Trainer.cpp

@@ -222,12 +222,6 @@ void Trainer::init(const std::shared_ptr<TrainerConfigHelper>& config,
         DataProvider::create(config_->getTestDataConfig(), *config_, gpuData));
   }
   if (testDataProvider_) {
-    if (config_->getOptConfig().use_sparse_remote_updater()) {
-      LOG(FATAL) << "It's prohibited to set sparse_remote_update "
-                 << "in some layers if testing will be under going "
-                 << "in the middle of training. You can do testing "
-                 << "within separate process.";
-    }
     createTester();
   }