diff --git a/deepspeech/models/lm/transformer.py b/deepspeech/models/lm/transformer.py
index 9392a15ed2176924fb0ef759dedcc1c36d045599..dcae4ea0f110c484b26a8b24afa8cbee2ebde03a 100644
--- a/deepspeech/models/lm/transformer.py
+++ b/deepspeech/models/lm/transformer.py
@@ -45,8 +45,7 @@ class TransformerLM(nn.Layer, LMInterface, BatchScorerInterface):
         if pos_enc == "sinusoidal":
             pos_enc_layer_type = "abs_pos"
         elif pos_enc is None:
-            #TODO
-            pos_enc_layer_type = "None"
+            pos_enc_layer_type = "no_pos"
         else:
             raise ValueError(f"unknown pos-enc option: {pos_enc}")
 
diff --git a/deepspeech/modules/encoder.py b/deepspeech/modules/encoder.py
index bbe822af67df978836295f9e0fee8f75cf092aa3..a9b5e8a6a27971f7b7f9c220ac3c0366c4a2ab04 100644
--- a/deepspeech/modules/encoder.py
+++ b/deepspeech/modules/encoder.py
@@ -384,7 +384,7 @@ class TransformerEncoder(BaseEncoder):
 
         Args:
             xs (paddle.Tensor): Input tensor. (B, T, D)
-            masks (paddle.Tensor): Mask tensor. (B, 1, T)
+            masks (paddle.Tensor): Mask tensor. (B, T, T)
             cache (List[paddle.Tensor]): List of cache tensors.
 
         Returns: