Dakota reports an error when building a radial_basis model

[ffttkx]

I use dakota6.21 to build the surrogate model. When I use other models like # polynomial quadratic # neural_network # gaussian_process surfpack trend quadratic # mars # experimental_gaussian_process # moving_least_squares # function_train # experimental_polynomial, they can be successfully built, but when I use radial_basis, an error message will be prompted: ">>>>> Building global_radial_basis approximations.
Constructing global approximations with no anchor, 0 DACE samples, and 10 reused points.
dakota: /home/dakota-jenkins/jenkins/workspace/public_rhel8_builder/source/packages/surfpack/src/surfaces/RadialBasisFunctionModel.cpp:116: VecRbf makeRbfs(const SurfData&, const SurfData&): Assertion `generators.size() == radii.size()' failed.
Aborted (core dumped)". After trying to add ’bases = 1‘, it can run successfully but the result is wrong (the number of model parameters is 0). The terminal output is:
" Building global_radial_basis approximations.
Constructing global approximations with no anchor, 0 DACE samples, and 10 reused points.

Surrogate quality metrics at build (training) points for height:
sum_squared 20331.4
mean_squared 2033.14
root_mean_squared 45.0903
sum_abs 412.611
mean_abs 41.2611
max_abs 66.7301
rsquared 5.14877

global_radial_basis approximation builds completed.
Beginning Approximate Fn Evaluations...
Function evaluation summary (APPROX_INTERFACE_1): 40 total (40 new, 0 duplicate)
Statistics based on 40 samples:

Sample moment statistics for each response function:
Mean Std Dev Skewness Kurtosis
height 0.0000000000e+00 0.0000000000e+00 0.0000000000e+00 -3.0000000000e+00

95% confidence intervals for each response function:
LowerCI_Mean UpperCI_Mean LowerCI_StdDev UpperCI_StdDev
height 0.0000000000e+00 0.0000000000e+00 0.0000000000e+00 0.0000000000e+00

At least one correlation coefficient is nan or inf. This commonly occurs when
discrete variables (including histogram variables) are present, a response is
completely insensitive to variables (response variance equal to 0), there are
fewer samples than variables, or some samples are approximately collinear.

Simple Correlation Matrix among all inputs and outputs:
weir_h height
weir_h 1.00000e+00
height -nan -nan

Partial Correlation Matrix between input and output:
height
weir_h-nan

Simple Rank Correlation Matrix among all inputs and outputs:
weir_h height
weir_h 1.00000e+00
height -nan -nan

Partial Rank Correlation Matrix between input and output:
height
weir_h -nan". How on earth should I solve this problem and build the radial_basis surrogate model? Here is my .in file:
"environment,
tabular_data
tabular_data_file = 'surrogate_values_ex2.dat'
custom_annotated header eval_id

variables,
uniform_uncertain=1
lower_bounds = 0.0
upper_bounds = 70.0
descriptors = 'weir_h'

method
id_method = 'surrogate_global'
model_pointer = 'surrogate_neural'

sampling
sample_type lhs
samples=40
seed=6347

model
id_model = 'surrogate_neural'
responses_pointer = 'response_height'
surrogate global
radial_basis
metrics =
'sum_squared'
'mean_squared'
'root_mean_squared'
'sum_abs'
'mean_abs'
'max_abs'
'rsquared'
import_build_points_file = 'model_values_ex2.dat'
custom_annotated header eval_id
responses
id_responses = 'response_height'
response_functions = 1
descriptors = 'height'
no_gradients no_hessians
"