> Just that it isn't the immediate transformer killer. What is the best/stable-ish linear alternative for transformer right now? Especially for text generatio

> Just that it isn't the immediate transformer killer. What is the best/stable-ish linear alternative for transformer right now? Especially for text generation and summarization. We have domain specific ways of over sampling and search, so we much prefer less expensive models.