Analyzing Multi-Head Self-Attention: Specialized Heads Do the Heavy Lifting, the Rest Can Be Pruned
39385Knowledge Triples
| Source Entity | Relationship | Target Entity |
|---|---|---|
|
Қырқу әдісі
Method |
evaluatedIn |
WMT деректер жинағы
Material |
|
WMT деректер жинағы
Material |
measuredBy |
BLEU
Metric |