SUMIN/HuggingFace_transformer
1
0
Fork 0
Code Issues Pull Requests Actions 7 Packages Projects Releases Wiki Activity

Fix TFAlbertForSequenceClassification classifier dropout probability. It was set to config.hidden_dropout_prob, but should be config.classifier_dropout_prob. (#3928)

Browse Source
This commit is contained in:
Jared T Nielsen
2020-04-23 11:18:16 -06:00
committed by GitHub
parent 8e093e5981
commit a79a9e1241
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
src/transformers/modeling_tf_albert.py
View File

@@ -781,7 +781,7 @@ class TFAlbertForSequenceClassification(TFAlbertPreTrainedModel):
self.num_labels = config.num_labels self.num_labels = config.num_labels
self.albert = TFAlbertMainLayer(config, name="albert") self.albert = TFAlbertMainLayer(config, name="albert")
self.dropout = tf.keras.layers.Dropout(config.hidden_dropout_prob) self.dropout = tf.keras.layers.Dropout(config.classifier_dropout_prob)
self.classifier = tf.keras.layers.Dense( self.classifier = tf.keras.layers.Dense(
config.num_labels, kernel_initializer=get_initializer(config.initializer_range), name="classifier" config.num_labels, kernel_initializer=get_initializer(config.initializer_range), name="classifier"
) )