Searched refs:attention_scores (Results 1 – 2 of 2) sorted by relevance
| /external/tensorflow/tensorflow/python/keras/layers/ |
| D | multi_head_attention.py | 417 def _masked_softmax(self, attention_scores, attention_mask=None): argument
424 for _ in range(len(attention_scores.shape) - len(attention_mask.shape)):
427 return self._softmax(attention_scores, attention_mask)
461 attention_scores = special_math_ops.einsum(self._dot_product_equation, key,
464 attention_scores = self._masked_softmax(attention_scores, attention_mask)
469 attention_scores, training=training)
474 return attention_output, attention_scores
499 attention_output, attention_scores = self._compute_attention(
504 return attention_output, attention_scores
|
| D | dense_attention.py | 174 result, attention_scores = self._apply_scores(
181 return result, attention_scores
|