Input x is shape (B, d_model, L)

u: (B H L) if self.transposed else (B L H) state: (H N) never needed unless you know what you're doing Returns: same shape a

Input x is shape (B, d_model, H, W)

X: (batch, dim, lengths...)

evaluate how well we could estimate the phase

Must have called self._setup_step() and created state with self.default_state() before calling this