finetune ce and refine gnn (#4089)

PaddleCV/image_classification/utils/utility.py

@@ -489,7 +489,7 @@ def print_ce(device_num, metrics, time_info):
     print("kpis\ttrain_cost_card{}\t{}".format(device_num, train_loss))
     print("kpis\ttrain_acc1_card{}\t{}".format(device_num, train_acc1))
     print("kpis\ttrain_acc5_card{}\t{}".format(device_num, train_acc5))
-    print("kpis\ttest_loss_card{}\t{}".format(device_num, test_loss))
+    print("kpis\ttest_cost_card{}\t{}".format(device_num, test_loss))
     print("kpis\ttest_acc1_card{}\t{}".format(device_num, test_acc1))
     print("kpis\ttest_acc5_card{}\t{}".format(device_num, test_acc5))
     print("kpis\ttrain_speed_card{}\t{}".format(device_num, train_speed))

PaddleCV/metric_learning/train_elem.py

@@ -274,8 +274,8 @@ def train_async(args):
     # This is for continuous evaluation only
     if args.enable_ce:
         # Use the mean cost/acc for training
-        print("kpis train_cost      %s" % (avg_loss))
-        print("kpis test_recall     %s" % (recall))
+        print("kpis\ttrain_cost\t{}".format(avg_loss))
+        print("kpis\ttest_recall\t{}".format(recall))
 
 
 def initlogging():

PaddleRec/gnn/train.py

@@ -101,10 +101,8 @@ def train():
     feed_list = [e.name for e in feed_datas]
 
     if use_parallel:
-        exec_strategy = fluid.ExecutionStrategy()
-        exec_strategy.num_threads = 1 if os.name == 'nt' else 0
         train_exe = fluid.ParallelExecutor(
-            use_cuda=use_cuda, loss_name=loss.name, exec_strategy=exec_strategy)
+            use_cuda=use_cuda, loss_name=loss.name)
     else:
         train_exe = exe