evaluation/grader.py

def clamp(score):
    return max(0.01, min(0.99, score))


def grade_easy(env, episode):
    rewards = episode["rewards"]
    score = sum(rewards) / max(1, len(rewards))
    return clamp(score)


def grade_medium(env, episode):
    rewards = episode["rewards"]
    score = (sum(rewards) / max(1, len(rewards))) + 0.1
    return clamp(score)


def grade_hard(env, episode):
    rewards = episode["rewards"]
    score = (sum(rewards) / max(1, len(rewards))) - 0.2
    return clamp(score)