Gradient Steps Per Env Step

Problem

Compute the number of gradient steps per environment step: epochs * (buffer_size / batch_size) / env_steps. A non-positive batch size or env step count returns 0.0.

Hint

This is just the replay ratio expressed through epochs over the buffer.

Test cases

case 1

in: [2,1000,100,100]

expected: 0.2

case 2

in: [1,500,50,100]

expected: 0.1

case 3

in: [4,64,32,8]

expected: 1

case 4

in: [1,100,0,10]

expected: 0

Starter codePython

def gradient_steps_per_env_step(epochs, buffer_size, batch_size, env_steps):
    # Your code here
    pass

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