View publication

This paper was accepted at WMT conference at EMNLP.

The Transformer architecture has two main non-embedding components: Attention and the Feed Forward Network (FFN). Attention captures interdependencies between words regardless of their position, while the FFN non-linearly transforms each input token independently. In this work, we explore the role of FFN and find that despite, and find that despite taking up a significant fraction of the model's parameters, it is highly redundant. Concretely, we are able to substantially reduce the number of parameters with only a modest drop in accuracy by removing the FFN on the decoder layers and sharing a single FFN across the encoder. Finally, we scale this architecture back to its original size by increasing the hidden dimension of the shared FFN, achieving substantial gains in both accuracy and latency with respect to the original Transformer Big.

Related readings and updates.

Deploying Attention-Based Vision Transformers to Apple Neural Engine

Motivated by the effective implementation of transformer architectures in natural language processing, machine learning researchers introduced the concept of a vision transformer (ViT) in 2021. This innovative approach serves as an alternative to convolutional neural networks (CNNs) for computer vision applications, as detailed in the paper, An Image Is Worth 16x16 Words: Transformers for Image Recognition at Scale.

See highlight details

Stabilizing Transformer Training by Preventing Attention Entropy Collapse

m*= Equal Contributors Training stability is of great importance to Transformers. In this work, we investigate the training dynamics of Transformers by examining the evolution of the attention layers. In particular, we track the attention entropy for each attention head during the course of training, which is a proxy for model sharpness. We identify a common pattern across different architectures and tasks, where low attention entropy is…
See paper details