Force sensors are really expensive right now, some might literally cost you as much as a brand new Mercedes. One of the reasons they cost so much is that they should have been carefully calibrated by the manufacturer. So you might want to test them out while you still have a warranty.

What follows here is one way in which the vertical force, pitch&roll torques can be tested without any need for special equipment(although you will need weights of known mass).

Please remember that force sensors are not completely accurate, check your datasheets the manufacturer probably says something about the average error(like ±5% error) .

Also, the force sensors may have a offset error so you might want to subtract these in your program.

Ok down to business, here is a force sensor mounted to the foot of the MARI-3 biped robot

Now to test the vertical forces I would simply log the sensor output on the vertical with no load and compare it with a load of known weight. In this case I can simply place them on the foot(as shown below), your robot may differ though.

In this case for example I applied 2 kg so I would expect the vertical direction to change by 2*9.8 (±5%) Newtons.

Now if the vertical force is measuring correctly then you can use the ZMP/COP(lets use Center Of Pressure for now, its a more intuative term) equation to test if the torques are correct, recall that the COP can be expressed as follows

$x_{cop} = \frac{\tau_y}{F_z}$

So this means that by applying any force to a known location and calculating the COP from the output vertical force(which has already been verified) and pitch torque,  it is possible to verify if the torques are accurate or not.

So if for example you apply a force at x=100 mm(as shown below) then the sensor data with the COP equation should also give you about 100 mm otherwise something is wrong, the magnitude and direction of the COP error will tell you how far off the torques are.

Finally you can just repeat this procedure for the roll ax.