Hi Miljena,
first of all, you need to specify
what you mean by "particle size":
(i)
effective hydrodynamic size, i.e. friction coefficient
(ii) effective
size that matters for excluded-volume interactions
Depending on the physics of your system, you may need (i), (ii) or both.
(i) you need to set gamma per particle as you describe, and you need Espresso compiled with LANGEVIN_PER_PARTICLE activated, as pointed out by Hender Lopez. Also, note that when dealing with diffusion properties, Langevin thermostat is not the best choice as it screens hydrodynamics.
(ii) you need to modify the parameters of the short-range interaction potential that you use, e.g. in the LJ potential you would vary sigma or add some cutoff.