Correct posteriors but many divergences

Questions v5
8

So after testing pytensor.gradient.verify_grad I see that if I set the tolerance to 1e-7 I get:

GradientError: GradientError: numeric gradient and analytic gradient exceed tolerance:
        At position 0 of argument 5 with shape (),
            val1 = 22120.189872      ,  val2 = 22120.179274
            abs. error = 0.010598,  abs. tolerance = 0.000000
            rel. error = 0.000000,  rel. tolerance = 0.000000
Exception args:

As far as prior predictive checks, I believe this has already been done in Octofitter.

Does this mean I should make a custom gradient?