numberOfAttentionHeads

P130209 predicate

Indicates the number of distinct attention heads used within an attention mechanism or layer in a model.

Observed surface forms (1)

Sample triples (3)

Subject Object
BERT
surface form: BERT_BASE
12 via predicate surface "numAttentionHeads"
BERT
surface form: BERT_LARGE
16 via predicate surface "numAttentionHeads"
DistilBERT 12