revert scale loss to False in Embeddingintent

dakshvar22 · dakshvar22 · commit b8fcae1aa15e · 2020-04-27T15:06:35.000+05:30
diff --git a/changelog/5717.improvement.rst b/changelog/5717.improvement.rst
@@ -1,3 +1,3 @@
 Change default values for hyper-parameters in ``EmbeddingIntentClassifier`` and ``DIETClassifier``
 
-Use ``scale_loss=False`` in both classifiers. Additionally, reduce the number of dense dimensions for sparse features of text from 512 to 256 in ``EmbeddingIntentClassifier``.
+Use ``scale_loss=False`` in ``DIETClassifier``. Reduce the number of dense dimensions for sparse features of text from 512 to 256 in ``EmbeddingIntentClassifier``.
diff --git a/docs/nlu/components.rst b/docs/nlu/components.rst
@@ -908,7 +908,7 @@ EmbeddingIntentClassifier
          |                                 |                  | over incorrect intent labels, used only if 'loss_type' is    |
          |                                 |                  | set to 'margin'.                                             |
          +---------------------------------+------------------+--------------------------------------------------------------+
-         | scale_loss                      | False            | Scale loss inverse proportionally to confidence of correct   |
+         | scale_loss                      | True             | Scale loss inverse proportionally to confidence of correct   |
          |                                 |                  | prediction.                                                  |
          +---------------------------------+------------------+--------------------------------------------------------------+
          | regularization_constant         | 0.002            | The scale of regularization.                                 |
diff --git a/rasa/nlu/classifiers/embedding_intent_classifier.py b/rasa/nlu/classifiers/embedding_intent_classifier.py
@@ -115,7 +115,7 @@ def required_components(cls) -> List[Type[Component]]:
         # incorrect intent labels, used only if 'loss_type' is set to 'margin'.
         USE_MAX_NEG_SIM: True,
         # Scale loss inverse proportionally to confidence of correct prediction
-        SCALE_LOSS: False,
+        SCALE_LOSS: True,
         # ## Regularization parameters
         # The scale of regularization
         REGULARIZATION_CONSTANT: 0.001,

Original file line number	Diff line number	Diff line change
`@@ -1,3 +1,3 @@`
`1`	`1`	Change default values for hyper-parameters in ``EmbeddingIntentClassifier`` and ``DIETClassifier``
`2`	`2`
`3`		-Use ``scale_loss=False`` in both classifiers. Additionally, reduce the number of dense dimensions for sparse features of text from 512 to 256 in ``EmbeddingIntentClassifier``.
	`3`	+Use ``scale_loss=False`` in ``DIETClassifier``. Reduce the number of dense dimensions for sparse features of text from 512 to 256 in ``EmbeddingIntentClassifier``.