Yes, YALMIP passes initial values if asked to do so through the 'usex0' options
x = intvar(...)
...
assign(x,feasiblevalue);
solvesdp(Constraints,Objective,sdpsettings('usex0',1))
Note though, this will only work if you actually know all the integer variables. In complex models, YALMIP will introduce a lot of internal binary and integer variables which you have no control of.
Also, knowing reasonable solutions is in many cases not of much help, since the main effort for the solver is not to find the optimal solution, but to close the gap to the relaxation.