Inspect values of derived (deterministic, un-transformed) variables at point?

I’m debugging a miss-specified model and trying to figure out why the logp of my observations is inf.
Unfortunately for me, one of the variables is transformed, I believe because it’s a TruncatedNormal. Is there any way to recover the un-transformed value from a point?
Here’s me in the debugger:

ipdb> p point
{'βtemp1': array(1.11700581), 'βtemp2': array(0.93765761), 'βod1': array(0.03401939), 'βod2': array(0.63373602),
 'AND Output_interval__': array([ 0.2386266 ,  0.13873085, -0.53425323, -0.16257509]), 
   'medium influences': array([[-0.69232133,  0.76863831,  0.71339414, -0.31977171],
       [ 0.55893651, -0.30711682,  0.10664718,  0.68078547],
       [-0.29005906,  0.37676899,  0.62135856, -0.14252686]]), 
   'err_sd_log__': array([-0.07765381, -1.20334943, -0.99376342, ...,  0.09904621,
       -0.8738467 , -0.15598646])}

Is there any way I can find the values of "AND Output" from the values of "AND Output__interval__"? If I can do that, I think I will be able to tell almost instantly why my model is breaking.
In general, if there was a way to interpret the un-transformed, named variables of a model from a point, that would be great.
There’s probably a function or method to do this, I just don’t know where to look for it.

OK, this is kind of terrifying, but I believe that this is how I do it:

p model['AND Output'].transformation.backward(np.array([ 0.2386266 ,  0.13873085, -0.53425323, -0.16257509])).eval()

Where the argument to backward comes out of the debugger output above.

That does for transformed variables, but doesn’t get me to the deterministic ones, which would really make the debugging easier.

There is this helper function you might find helpful:

I think if you do update_start_vals(testpoint, testpoint, model) it should output a dict with parameters that in the domain that the logp actually takes.

Thank you very much! This looks like just what I need!