Fix logging in transformer dygraph (#4827)

dygraph/transformer/train.py

@@ -29,6 +29,10 @@ from utils.check import check_gpu, check_version
 import reader
 from model import Transformer, CrossEntropyCriterion, NoamDecay
 
+FORMAT = '%(asctime)s-%(levelname)s: %(message)s'
+logging.basicConfig(level=logging.INFO, format=FORMAT)
+logger = logging.getLogger(__name__)
+
 
 def do_train(args):
     if args.use_cuda:
@@ -180,7 +184,7 @@ def do_train(args):
                     total_avg_cost = avg_cost.numpy() * trainer_count
 
                     if step_idx == 0:
-                        logging.info(
+                        logger.info(
                             "step_idx: %d, epoch: %d, batch: %d, avg loss: %f, "
                             "normalized loss: %f, ppl: %f" %
                             (step_idx, pass_id, batch_id, total_avg_cost,
@@ -189,7 +193,7 @@ def do_train(args):
                     else:
                         train_avg_batch_cost = args.print_step / (
                             time.time() - batch_start)
-                        logging.info(
+                        logger.info(
                             "step_idx: %d, epoch: %d, batch: %d, avg loss: %f, "
                             "normalized loss: %f, ppl: %f, avg_speed: %.2f step/s"
                             % (step_idx, pass_id, batch_id, total_avg_cost,
@@ -216,11 +220,11 @@ def do_train(args):
                             total_sum_cost += sum_cost.numpy()
                             total_token_num += token_num.numpy()
                             total_avg_cost = total_sum_cost / total_token_num
-                        logging.info("validation, step_idx: %d, avg loss: %f, "
-                                     "normalized loss: %f, ppl: %f" %
-                                     (step_idx, total_avg_cost,
-                                      total_avg_cost - loss_normalizer,
-                                      np.exp([min(total_avg_cost, 100)])))
+                        logger.info("validation, step_idx: %d, avg loss: %f, "
+                                    "normalized loss: %f, ppl: %f" %
+                                    (step_idx, total_avg_cost,
+                                     total_avg_cost - loss_normalizer,
+                                     np.exp([min(total_avg_cost, 100)])))
                         transformer.train()
 
                     if args.save_model and (
@@ -242,8 +246,8 @@ def do_train(args):
 
             train_epoch_cost = time.time() - epoch_start
             ce_time.append(train_epoch_cost)
-            logging.info("train epoch: %d, epoch_cost: %.5f s" %
-                         (pass_id, train_epoch_cost))
+            logger.info("train epoch: %d, epoch_cost: %.5f s" %
+                        (pass_id, train_epoch_cost))
 
         if args.save_model:
             model_dir = os.path.join(args.save_model, "step_final")