Hawk and Griffin: Efficient RNN Models Redefining AI Performance | HackerNoon
Briefly

Hawk represents a significant advance in recurrent model architecture, utilizing the RG-LRU layer to effectively address the challenges of scaling and efficiency in training.
The introduction of Griffin, which combines RG-LRU with local attention, shows promising results in improving both training efficiency and inference speed compared to traditional models.
Read at Hackernoon
[
|
]