Merge pull request google-research#25 from eric-haibin-lin/patch-1

Minor documentation fix
mangojiez · Nov 2, 2018 · 39ed6c5 · 39ed6c5
2 parents 9c8798f + 4cf3d02
commit 39ed6c5
Showing 1 changed file with 2 additions and 1 deletion.
diff --git a/modeling.py b/modeling.py
@@ -611,7 +611,8 @@ def attention_layer(from_tensor,
  query_act: (optional) Activation function for the query transform.
  key_act: (optional) Activation function for the key transform.
  value_act: (optional) Activation function for the value transform.
- attention_probs_dropout_prob:
+ attention_probs_dropout_prob: (optional) float. Dropout probability of the attention
+ probabilities.
  initializer_range: float. Range of the weight initializer.
  do_return_2d_tensor: bool. If True, the output will be of shape [batch_size
  * from_seq_length, num_attention_heads * size_per_head]. If False, the