Losses for parameter estimation of Neural SDE

@jlmaccal Hi, I was curious: how did you solve it in the end?