Dakota complains analytic gradients when specified

[k-ingles]

Using Dakota v6.18

I have this very simple problem that I wish to run:
The Dakota input file is:

environment
  tabular_data
    tabular_data_file = 'dakota_parabola_summary.dat'

method
  id_method 'searcher'
  optpp_q_newton
    gradient_tolerance       = 1.e-3
    max_iteration            = 100
    convergence_tolerance    = 1.e-3
    max_function_evaluations = 100

model
  single

variables
  continuous_design = 1
    descriptors      'x'

interface
  fork
    analysis_driver = 'python3 parabola.py'
    parameters_file = 'parabola.param'
    results_file    = 'parabola.obj_func'

response
  objective_functions = 1
  analytic_gradients 
  no_hessians

which runs the following python script

from sys import argv

def parabola(x: float) -> float:
    return (x - 2.2) ** 2 + 1.6, 2 * ( x - 2.2)

if __name__ == "__main__":
    input_file = argv[1]
    output_file = argv[2]

    with open(input_file, 'r') as fin:
        num_vars = int(fin.readline().split()[0])
        input_vars = {}
        for _ in range(num_vars):
            value, name = fin.readline().split()
            input_vars[name] = float(value)

    with open(output_file, 'w') as fout:
        output = parabola(**input_vars)
        fout.write(f"{output[0]} objective_function_1\n")
        fout.write(f"[ {output[1]} ]")

but then running dakota parabola.in gives

Using Dakota input file 'parabola.in'
Writing new restart file 'dakota.rst'.

>>>>> Executing environment.

>>>>> Running optpp_q_newton iterator.

---------------------
Begin Evaluation    1
---------------------
Parameters for evaluation 1:
                      0.0000000000e+00 x

blocking fork: python3 parabola.py parabola.param parabola.obj_func

Active response data for evaluation 1:
Active set vector = { 3 } Deriv vars vector = { 1 }
                      6.4400000000e+00 obj_fn
 [ -4.4000000000e+00 ] obj_fn gradient



---------------------
Begin Evaluation    2
---------------------
Parameters for evaluation 2:
                      9.9999997020e-01 x

blocking fork: python3 parabola.py parabola.param parabola.obj_func
Error(s) encountered reading results file parabola.obj_func for Evaluation 2:
Expected 0 gradients but found 1.

why is Dakota expecting zero gradients?

[jadamstephens]

Dakota has an optimization called the active set vector. It asks your driver only for the information that the method (optpp_q_newton) asks for. It would be any combination of the function value, the gradient, and the hessian of each response.

See here for more information.

You may want to disable the active set vector using the interface keywords deactivate active_set_vector, which will cause Dakota to request and expect everything you said your driver can return - function values and analytic gradients - for every evaluation.

[k-ingles]

Okay, I was suspected this was the issue. I saw that in the dakota.interfacing module you can query the asv, which alerted me to existing.
Thank you!