#计算机科学#Implementation of Linformer for Pytorch
#计算机科学#My take on a practical implementation of Linformer for Pytorch.
Reproducing the Linear Multihead Attention introduced in Linformer paper (Linformer: Self-Attention with Linear Complexity)
We implement a encoder/decoder seq2seq model, which has the same performance (similar perplexity) as traditional transformer but also reduce time complexity from O(n^2) to O(n).
list of efficient attention modules
Implementation of faster and more efficient sequence modeling deep learning algorithms based on fairseq, e.g. Lite-Transormer, Reformer, Linformer, efficient variants of attention mechanism and so on