-
Notifications
You must be signed in to change notification settings - Fork 0
/
evaluator.py
47 lines (40 loc) · 1.88 KB
/
evaluator.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
import os
import torch
import matplotlib.pyplot as plt
from src.algorithm.custom_ppo import CustomCombinedExtractor, CustomVecEnv, CustomPPO
from src.commons.evaluation_util import EvaluationResult, lfuEvaluate, lruEvaluate, modelEvaluate, randomEvaluate
from src.commons.road_util import RoadEnvConfig
from src.environment import RoadEnv
if __name__ == '__main__':
#threadNum = 2
threadNum = os.cpu_count()
torch.set_num_threads(threadNum)
config = RoadEnvConfig.load("./models/config.json")
config.seed = None
#config.broadBases = 0
#config.episodeDuration = 600
env = RoadEnv(config)
envWrapper = CustomVecEnv(env)
model = CustomPPO("MultiInputPolicy", envWrapper, policy_kwargs={
"features_extractor_class": CustomCombinedExtractor}, seed=env._config.seed, device="cpu")
try:
model = model.load(path="./models/model.zip", env=envWrapper)
print("use the previous model")
except:
print("use the fresh model")
evaluateTimes = 100
results: list[EvaluationResult] = []
results.append(modelEvaluate(model, envWrapper, evaluateTimes))
results.append(lruEvaluate(envWrapper, evaluateTimes))
results.append(lfuEvaluate(envWrapper, evaluateTimes))
results.append(randomEvaluate(envWrapper, evaluateTimes))
for result in results:
print(result)
envWrapper.close()
plt.subplot(121)
plt.scatter(["IPRSC", "LRU", "LFU", "Random"], [(result.directHit + result.indirectHit) / (result.directHit + result.indirectHit + result.missed) for result in results])
plt.subplot(122)
plt.scatter(["IPRSC", "LRU", "LFU", "Random"], [result.directHit / (result.directHit + result.indirectHit + result.missed) for result in results])
#plt.subplot(123)
#plt.scatter(["IPRSC", "LRU", "LFU", "Random"], [result.avgReward for result in results])
plt.savefig("./models/cache_hit_ratios.png")