Open
Description
Forward sensitivities tend to give solver errors if they're fed analytic or forward diff jacobians, but work fine with jacobians not specified. I plan to make that adjustment shortly. The mode forward sensitivities operate in when jacobians aren't specified is discussed here: https://diffeq.sciml.ai/stable/extras/sensitivity_math/. I'm unsure if using the analytic jacobians should be faster than that mode currently, but once we have fast parameter jacobians I would think it should be faster to use the analytic jacobians.