by zjn
1.克隆一下源码
1
| git clone https://github.com/FeiLiu36/EOH/
|
2.进入源码目录，创建虚拟环境
我家里的电脑没有装conda，所以我用的是Python自带的venv，如果你有conda，可以用conda创建虚拟环境。
然后激活虚拟环境:
1
| source eoh_env/Scripts/activate
|
记得虚拟环境的Python版本要>3.10
可以确定一下：
3.安装依赖
1
2
|
pip install numpy numba joblib
|
在这个目录下，直接
1
| pip install -r requirements.txt
|
也可以
4.配置llm参数
按照官网的指示，我选择了
示例1：旅行商问题的构造算法
1
2
3
| cd examples/tsp_construct
python runEoH.py
|
但是需要先去买api
DeepSeek 开放平台
在runEoH里面改一下参数
理论上就可以跑
但是，可能会遇到timeout
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
| PS C:\Users\zny\Desktop\eoh\EoH-main\eoh> & C:\Users\zny\Desktop\eoh\eoh_env\Scripts\python.exe c:/Users/zny/Desktop/eoh/EoH-main/examples/tsp_construct/runEoH.py
-----------------------------------------
--- Start EoH ---
-----------------------------------------
- output folder created -
- parameters loaded -
- Prob tsp_construct loaded
- EoH parameters loaded -
- Evolution Start -
- check LLM API
remote llm api is used ...
creating initial population:
Parallel time out .
Parallel time out .
Pop initial:
initial population has been created!
OP: e1, [1 / 4] | Obj: None| Obj: None| Obj: None| Obj: None|
OP: e2, [2 / 4] | Obj: None| Obj: None| Obj: None| Obj: None|
OP: m1, [3 / 4] | Obj: None| Obj: None| Obj: None| Obj: None|
OP: m2, [4 / 4] | Obj: None| Obj: None| Obj: None| Obj: None|
Traceback (most recent call last):
File "c:\Users\zny\Desktop\eoh\EoH-main\examples\tsp_construct\runEoH.py", line 22, in <module>
evolution.run()
File "C:\Users\zny\Desktop\eoh\EoH-main\eoh\src\eoh\eoh.py", line 42, in run
method.run()
File "C:\Users\zny\Desktop\eoh\EoH-main\eoh\src\eoh\methods\eoh\eoh.py", line 177, in run
json.dump(population[0], f, indent=5)
~~~~~~~~~~^^^
IndexError: list index out of range
|
找到utils getParas
timeout设的高一点，但没有完全解决问题，推测可能是r1模型不是很聪明，因为花了大几万token只找到一个解
或许也可以降低tsp问题的复杂度
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
| (eoh_env) PS C:\Users\zny\Desktop\eoh\EoH-main\eoh> & C:\Users\zny\Desktop\eoh\eoh_env\Scripts\python.exe c:/Users/zny/Desktop/eoh/EoH-main/examples/tsp_construct/runEoH.py
-----------------------------------------
--- Start EoH ---
-----------------------------------------
- output folder created -
- parameters loaded -
- Prob tsp_construct loaded
- EoH parameters loaded -
- Evolution Start -
- check LLM API
remote llm api is used ...
creating initial population:
Pop initial:
Obj: 7.03416| Obj: 7.50039| Obj: 11.58909|
initial population has been created!
OP: e1, [1 / 4] | Obj: 7.65304| Obj: 7.59725| Obj: 9.37639| Obj: 7.45505|
OP: e2, [2 / 4] | Obj: 7.6436| Obj: 8.47352| Obj: 8.59245| Obj: 8.47352|
OP: m1, [3 / 4] | Obj: None| Obj: None| Obj: None| Obj: None|
OP: m2, [4 / 4] | Obj: None| Obj: None| Obj: None| Obj: None|
--- 1 of 4 populations finished. Time Cost: 30.6 m
Pop Objs: 7.03416 7.45505 7.50039 7.59725
OP: e1, [1 / 4] | Obj: None| Obj: None| Obj: None| Obj: None|
OP: e2, [2 / 4] | Obj: None| Obj: None| Obj: None| Obj: None|
OP: m1, [3 / 4] | Obj: None| Obj: None| Obj: None| Obj: None|
OP: m2, [4 / 4] | Obj: None| Obj: None| Obj: None| Obj: None|
--- 2 of 4 populations finished. Time Cost: 36.9 m
Pop Objs: 7.03416 7.45505 7.50039 7.59725
OP: e1, [1 / 4] | Obj: None| Obj: None| Obj: None| Obj: None|
OP: e2, [2 / 4] | Obj: 7.45505| Obj: 10.23619| Obj: 7.57794| Obj: 8.05993|
OP: m1, [3 / 4] | Obj: None| Obj: None| Obj: None| Obj: None|
OP: m2, [4 / 4] | Obj: None| Obj: None| Obj: None| Obj: None|
--- 3 of 4 populations finished. Time Cost: 50.8 m
Pop Objs: 7.03416 7.45505 7.50039 7.57794
OP: e1, [1 / 4] | Obj: None| Obj: None| Obj: None| Obj: None|
OP: e2, [2 / 4] | Obj: None| Obj: None| Obj: None| Obj: None|
OP: m1, [3 / 4] | Obj: None| Obj: None| Obj: None| Obj: None|
OP: m2, [4 / 4] | Obj: None| Obj: None| Obj: None| Obj: None|
--- 4 of 4 populations finished. Time Cost: 57.1 m
Pop Objs: 7.03416 7.45505 7.50039 7.57794
> End of Evolution!
-----------------------------------------
--- EoH successfully finished ! ---
|
虽然成功finish
但是m1和m2的obj都是None
说明修改策略都超时了，因为e1和e2都简单一些所以成功了