Sorry that this is confusing.
verbosity is not an argument to
fit.workflow() so that shouldn't work. Here is the help page.
To set the verbosity of the model, you can quiet the results using
control_parsnip() (as you show). Since this isn't a reproducible example, we can't tell what went wrong for your data/models, but here is an example where it does work:
set_engine("nnet", trace = TRUE) %>%
add_formula(price ~ type + sqft + beds + baths) %>%
nn_fit <- fit(nn_wflow, Sacramento)
#> # weights: 36
#> initial value 72712505768277.218750
#> final value 16007850421474.472656
model_ctrl <- control_parsnip(verbosity = 0L)
nn_fit_check <- fit(nn_wflow, Sacramento, control = control_workflow(model_ctrl))
Created on 2022-01-31 by the reprex package (v2.0.1)