Yes and no. There's an dual connection between ssms and convolutional models if certain constraints are met. Training convolutionally and inferring sequentially

Yes and no. There's an dual connection between ssms and convolutional models if certain constraints are met. Training convolutionally and inferring sequentially seeks a compromise between the two sides. I think we're about to find out the degree to which those constraints impact "easily".