Obtaining predicted values using margins postcommand in a model log-transformed outcome variable

[Nader Mehri]

Using the below fake data, I am trying to fit a nbreg model to predict my outcome variable of interest in both original count format (totalparticipants) as well as log-transformed format (log10_totpart). Using margins postcommand, I am also trying to obtain the predicted value for my outcome variables by the serv_partner variable. I wonder to know how can I convert the coefficients and margins for model 2 (i.e., with the log-transformed outcome variable) to its original non-transformed values.
Thanks,
Nader

clear

set obs 3750
gen serv_partner = runiformint(0,1)
replace serv_partner = 1 if serv_partner >0.5
label define serv_partner_lbl 0 "no" 1 "yes"
label values serv_partner serv_partner_lbl

gen svi = rnormal(0,1)
gen Zip_pop = rnormal(50000,10000)
gen totalparticipants = rnormal(32.51378, 154.2246)
drop if totalparticipants<0
gen log10_totpart = log10(totalparticipants)
drop if log10_totpart<0
replace log10_totpart=log10_totpart+3
label var serv_partner "Partner"
label var svi "SVI"
label var Zip_pop "Zip code population"
label var totalparticipants "Total participants"
label var log10_totpart "Log10 total participants"

*model 1
nbreg totalparticipants i.serv_partner##c.svi c.Zip_pop, nolog
*model 2: with log-transforemd totalparticipants  
nbreg log10_totpart i.serv_partner##c.svi c.Zip_pop, nolog