Search Results for author:
Found 1 papers, 1 papers with code
Relaxed Attention for Transformer Models
The powerful modeling capabilities of all-attention-based transformer architectures often cause overfitting and - for natural language processing tasks - lead to an implicitly learned internal language model in the autoregressive transformer decoder complicating the integration of external language models.
Ranked #3 on
Lipreading
on LRS3-TED
(using extra training data)
Cannot find the paper you are looking for? You can
Submit a new open access paper.
