AWAC Advantage Weights

Problem

Compute AWAC policy weights as exp(advantage / lam), capped at max_weight to bound the update. Return one weight per advantage.

Hint

The exponential tilt makes good actions much more likely to be imitated.

case 1

in: [[0,1],1,10]

expected: [1,2.718281828459045]

case 2

in: [[0,5],1,10]

expected: [1,10]

case 3

in: [[-1],0.5,100]

expected: [0.1353352832366127]

Starter codePython

import math
def awac_advantage_weights(advantages, lam, max_weight):
    # Your code here
    pass

3 test cases · run your code against them in the editor