Mixed Precision Layer Assignment

Problem

Assign layers to higher precision under a byte budget. Upgrading layer i costs extra = layer_bytes[i] * (high_multiplier - 1) and yields sensitivity[i]; greedily pick upgrades by sensitivity per extra byte (ties by lower index) while the budget allows. Return the chosen indices in ascending order.

The signature is mixed_precision_assignment(layer_bytes, sensitivities, budget_bytes, high_multiplier=2.0).

Hint

This is a greedy knapsack: prioritize sensitivity gained per extra byte spent.

Test cases

case 1

in: [[10,20,30],[5,4,3],25]

expected: [0]

case 2

in: [[10,20,30],[5,4,3],35]

expected: [0,1]

case 3

in: [[10,20,30],[5,4,3],0]

expected: []

case 4

in: [[10,20,30],[5,4,3],1000]

expected: [0,1,2]

case 5

in: [[10,10],[5,5],10]

expected: [0]

Starter codePython

def mixed_precision_assignment(layer_bytes, sensitivities, budget_bytes, high_multiplier=2.0):
    # Your code here
    pass

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