FlashFFTConv can be definitely be implemented on Mamba, right?

[nam-drun]

[DanFu09]

Mamba does not have a convolutional form, so there isn't an exact mapping. For Mamba you'll have to use the scan formulation as documented in the paper.

[nam-drun]

For Mamba you'll have to use the scan formulation as documented in the paper. I get what you mean (algorithm 2 in figure 2). I was confused a bit because figure 3 showing mamba architecture also has Conv layer???