Merge changes from github.

PiperOrigin-RevId: 190835392

3 files changed, 8 insertions, 8 deletions

diff --git a/tensorflow/contrib/seq2seq/kernels/beam_search_ops.cc b/tensorflow/contrib/seq2seq/kernels/beam_search_ops.cc
index dfa12e873a..a9a32b7b25 100644
--- a/tensorflow/contrib/seq2seq/kernels/beam_search_ops.cc
+++ b/tensorflow/contrib/seq2seq/kernels/beam_search_ops.cc
@@ -74,7 +74,7 @@ class GatherTreeOp : public OpKernel {
         ctx,
         step_ids_shape.dim_size(1) == max_sequence_lengths.shape().dim_size(0),
         errors::InvalidArgument("batch size dimensions step_ids.shape[1] and "
-                                "max_seqeuence_lengths.shape[0] must match.  "
+                                "max_sequence_lengths.shape[0] must match.  "
                                 "but shapes are: ",
                                 step_ids_shape.DebugString(), " and ",
                                 max_sequence_lengths.shape().DebugString()));
diff --git a/tensorflow/contrib/seq2seq/python/ops/attention_wrapper.py b/tensorflow/contrib/seq2seq/python/ops/attention_wrapper.py
index 9ff8a343f1..be53779826 100644
--- a/tensorflow/contrib/seq2seq/python/ops/attention_wrapper.py
+++ b/tensorflow/contrib/seq2seq/python/ops/attention_wrapper.py
@@ -736,7 +736,7 @@ class _BaseMonotonicAttentionMechanism(_BaseAttentionMechanism):
   """Base attention mechanism for monotonic attention.
 
   Simply overrides the initial_alignments function to provide a dirac
-  distribution,which is needed in order for the monotonic attention
+  distribution, which is needed in order for the monotonic attention
   distributions to have the correct behavior.
   """
 
@@ -763,7 +763,7 @@ class _BaseMonotonicAttentionMechanism(_BaseAttentionMechanism):
 class BahdanauMonotonicAttention(_BaseMonotonicAttentionMechanism):
   """Monotonic attention mechanism with Bahadanau-style energy function.
 
-  This type of attention encorces a monotonic constraint on the attention
+  This type of attention enforces a monotonic constraint on the attention
   distributions; that is once the model attends to a given point in the memory
   it can't attend to any prior points at subsequence output timesteps.  It
   achieves this by using the _monotonic_probability_fn instead of softmax to
@@ -867,7 +867,7 @@ class BahdanauMonotonicAttention(_BaseMonotonicAttentionMechanism):
 class LuongMonotonicAttention(_BaseMonotonicAttentionMechanism):
   """Monotonic attention mechanism with Luong-style energy function.
 
-  This type of attention encorces a monotonic constraint on the attention
+  This type of attention enforces a monotonic constraint on the attention
   distributions; that is once the model attends to a given point in the memory
   it can't attend to any prior points at subsequence output timesteps.  It
   achieves this by using the _monotonic_probability_fn instead of softmax to
@@ -1133,7 +1133,7 @@ class AttentionWrapper(rnn_cell_impl.RNNCell):
       output_attention: Python bool.  If `True` (default), the output at each
         time step is the attention value.  This is the behavior of Luong-style
         attention mechanisms.  If `False`, the output at each time step is
-        the output of `cell`.  This is the beahvior of Bhadanau-style
+        the output of `cell`.  This is the behavior of Bhadanau-style
         attention mechanisms.  In both cases, the `attention` tensor is
         propagated to the next time step via the state and is used there.
         This flag only controls whether the attention mechanism is propagated
diff --git a/tensorflow/contrib/seq2seq/python/ops/beam_search_decoder.py b/tensorflow/contrib/seq2seq/python/ops/beam_search_decoder.py
index a26107b0d7..184144f64a 100644
--- a/tensorflow/contrib/seq2seq/python/ops/beam_search_decoder.py
+++ b/tensorflow/contrib/seq2seq/python/ops/beam_search_decoder.py
@@ -821,9 +821,9 @@ def _get_scores(log_probs, sequence_lengths, length_penalty_weight):
   Returns:
     The scores normalized by the length_penalty.
   """
-  length_penality_ = _length_penalty(
+  length_penalty_ = _length_penalty(
       sequence_lengths=sequence_lengths, penalty_factor=length_penalty_weight)
-  return log_probs / length_penality_
+  return log_probs / length_penalty_
 
 
 def _length_penalty(sequence_lengths, penalty_factor):
@@ -860,7 +860,7 @@ def _mask_probs(probs, eos_token, finished):
   unfinished beams remain unchanged.
 
   Args:
-    probs: Log probabiltiies of shape `[batch_size, beam_width, vocab_size]`
+    probs: Log probabilities of shape `[batch_size, beam_width, vocab_size]`
     eos_token: An int32 id corresponding to the EOS token to allocate
       probability to.
     finished: A boolean tensor of shape `[batch_size, beam_width]` that