EL-Attention: Memory Efficient Lossless Attention for Generation

  Yu Yan 1 Jiusheng Chen 1 Weizhen Qi * 2 Nikhil Bhendawade * 1 Yeyun Gong * 3 Nan Duan 3 Ruofei Zhang 4

             Abstract                pruning layer (Fan et al., 2019) or training a smaller stu-
                                 dent model (Shleifer & Rush, 2020). Another way is non-
    Transformer model with multi-head attention re-      autoregressive generation (Gu et al., 2018; Lee et al., 2018;
    quires caching inte ...