HDDM issue : 'race_no_bias_3' (and not with 'race_no_bias_4')

[luzuzek]

Hello there,

I am working with the latest version of the HDDM 0.98RC, using this virtual machine https://github.com/hcp4715/dockerHDDM_Guide which has everything installed.

I'm trying to use the 'race_no_bias_3'
So I first followed the instructions from your tutorial,

1) I generate the data

Metadata
nmcmc = 1000
model = 'race_no_bias_3'
n_samples = 1000

includes = hddm.model_config.model_config[model]['hddm_include']

from hddm.simulators.hddm_dataset_generators import simulator_h_c
data, full_parameter_dict = simulator_h_c(n_subjects = 1,
n_trials_per_subject = n_samples,
model = model,
p_outlier = 0.00,
conditions = None,
depends_on = None,
regression_models = None,
regression_covariates = None, # need this to make initial covariate matrix from which to use dmatrix (patsy)
group_only_regressors = False,
group_only = None,
fixed_at_default = None)

and it works perfectly.

2) Then, I try to run the model to obtain the estimated parameters of the data

hddmnn_model = hddm.HDDMnn(data,
informative = False,
include = includes,
p_outlier = 0.01,
w_outlier = 0.1,
model = model,)

And I obtain this message:

Couldn't execute load_torch_mlp()...
Option 1: pytorch not installed or version older than 1.7?
Option 2: pytorch model for your model string is not yet available

And this is super weird because when I try the same procedure with model = 'race_no_bias_4' (instead or 3) it works perfectly! I don't get any weird messages about Pytorch and I'm able to estimate the parameters. And the difference between one model and the other is just one more option as a response, right?

Does anyone encounter the same problem?