diff --git a/deepspeech/exps/deepspeech2/model.py b/deepspeech/exps/deepspeech2/model.py
index b1ff5c591404152f7c96357e564ca700a2d73c74..c1fe822509a830b52a03f7e8fa696e2e376d241d 100644
--- a/deepspeech/exps/deepspeech2/model.py
+++ b/deepspeech/exps/deepspeech2/model.py
@@ -323,6 +323,8 @@ class DeepSpeech2Tester(DeepSpeech2Trainer):
         config.data.manifest = config.data.test_manifest
         config.data.keep_transcription_text = True
         config.data.augmentation_config = ""
+        # filter test examples, will cause less examples, but no mismatch with training
+        # and can use large batch size , save training time, so filter test egs now.
         # config.data.min_input_len = 0.0  # second
         # config.data.max_input_len = float('inf')  # second
         # config.data.min_output_len = 0.0  # tokens
diff --git a/deepspeech/exps/u2/model.py b/deepspeech/exps/u2/model.py
index 0bcd1d22421b01d2f9df6e7a208b0557256465ac..00f4f5ec4d53e96e7eba22e7a6565d76a2f8eae3 100644
--- a/deepspeech/exps/u2/model.py
+++ b/deepspeech/exps/u2/model.py
@@ -254,6 +254,8 @@ class U2Trainer(Trainer):
         config.data.manifest = config.data.test_manifest
         config.data.keep_transcription_text = True
         config.data.augmentation_config = ""
+        # filter test examples, will cause less examples, but no mismatch with training
+        # and can use large batch size , save training time, so filter test egs now.
         # config.data.min_input_len = 0.0  # second
         # config.data.max_input_len = float('inf')  # second
         # config.data.min_output_len = 0.0  # tokens