Request for Implementation of Memory-Efficient Attention in Conformer

[N-damo]

Hello Conformer developers and contributors,

I have been using the Conformer model for my research and I'm very impressed with its performance. However, I noticed that the current implementation does not include memory-efficient attention, which is crucial for handling long sequences and reducing memory usage during training and inference.
Best regards,
LianLin