I think your expression for the kinetic energy is incorrect.

Disregarding physics and considering what you have done only programming-wise

1. Boundary conditions need to be used with equal (==) rather then Set (i.e, a single = sign)
2. There is a problem with the differentiation (the first part)
3. You have two variables but only one differential equation, so after correcting the mistakes, you are still left with insoluble problem since NDSolve (or any other solution solution method, be it computational or analytical) will not be able to solve this.

Here is the corrected code, you may continue from there

l1 = 1;
l2 = 1g = 9.81;
m1 = 1;
m2 = 1;
x1 = l1*Sin[\[Theta]1[t]];
x1' = D[x1, t];
y1 = -l1*Cos[\[Theta]1[t]];
y1' = D[y1, t];
x2 = l2*Sin[\[Theta]2[t]] + x1;
x2' = D[x2, t];
y2 = -l2*Cos[\[Theta]2[t]] + y1;
y2' = D[y2, t];

V = m1*g*y1 + m2*g*y2;
T = m1/2*(x1' + y1')^2 + m2*(x2' + y2')^2;
Lagrange = T - V

eqs = D[D[Lagrange, \[Theta]'[t]], t] - D[Lagrange, \[Theta]];
sol = NDSolve[{Lagrange == 0, \[Theta]1[0] == Pi, \[Theta]2[0] == 
    Pi, \[Theta]1'[0] == 0, \[Theta]2'[0] == 
    0}, {\[Theta]1, \[Theta]2}, {t, 0, 10}]

HTH

yehuda