You cannot. Your SW potential is simulating point particles. Those don’t have an explicit diameter, only some kind of “effective radius” that is implied by the potential parameters.
Let me repeat, your atoms do not have an explicit radius. Technically, you can stop the error from happening, by changing the atom style, but such modification of you input would have no effect, since your pair styles do not use the per-atom diameter property.
What I wrote about SW applies to LJ as well. Both simulate point particles which have no explicit diameter so the corresponding per-atom property would be ignored. The only way you can change the effective radius on the fly with fix adapt would be by changing the potential parameters. This should be possible for LJ (-> sigma) and it is not possible for SW. You have to refer to the documentation for details.