Batch-Latency Tradeoff

Problem

Given measured latency per batch size, compute throughput = batch / latency * 1000 (items per second) and return [best_batch, best_throughput] for the batch with the highest throughput, breaking ties toward the smaller batch.

The signature is batch_latency_tradeoff(batch_sizes, latencies).

Hint

Throughput usually saturates while latency keeps growing with batch size.

Test cases

case 1

in: [[1,4,16],[10,20,100]]

expected: [4,200]

case 2

in: [[1,2],[5,5]]

expected: [2,400]

case 3

in: [[1],[2]]

expected: [1,500]

case 4

in: [[8,16],[40,120]]

expected: [8,200]

Starter codePython

def batch_latency_tradeoff(batch_sizes, latencies):
    # Your code here
    pass

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