Question

我正在进行CPLEX 12.5 v / s CPLEX 12.8的性能比较。我正在使用CPLEX Python（v2.7.15）API来读取<script src="https://ajax.googleapis.com/ajax/libs/jquery/2.1.1/jquery.min.js"></script> <div class="main-content"> <section id="jumbotron"> <div class="container">  </div> </section> <section id="about"> <div class="container">  </div> </section> <section id="projects"> <div class="container">  </div> </section> <section id="contacts"> <div class="container">  </div> </section> </div>个文件。我首先对大约32个.sav文件进行了试运行，其中以下是总解决时间（以秒为单位）的摘要：

.sav

32个CPLEX 12.5: mean = 252.4; median = 127.0; standard deviation = 360.0 CPLEX 12.8: mean = 256.7; median = 129.5; standard deviation = 351.8文件中只有4个文件CPLEX 12.8花费的时间少于CPLEX 12.5。对于这4个文件，CPLEX 12.8平均减少了4.2秒，而其余28个文件平均花费了5.5秒。正如您在下面的代码中看到的，我只更改了四个参数设置：.sav以在1小时停止运行时间，timelim_cb.timelimit = 3600将工作内存设置为64 MB，c.parameters.workmem.set(64)为压缩并将节点文件存储到磁盘，并c.parameters.mip.strategy.file.set(3)更改工作目录。

我希望CPLEX 12.8花费更少的时间来解决，但这并没有发生。如果有人可以指出我应该探索哪些额外的参数/设置/想法以便找到我得到的结果的解释，那将是非常有帮助的。我知道CPLEX 12.5的默认c.parameters.workdir.set(r'C:\Users\username\folder')参数设置为128 MB，CPLEX 12.8的默认workmem参数设置为2048 MB。我为此设置为64 MB，因为之前对于CPLEX 12.5，我收到了错误：Warning: MIP starts not constructed because of out-of-memory status. Consider decreasing WorkMem parameter value to reduce memory usage. CPLEX Error 1001: Out of memory.

我的笔记本电脑的功能包括：Windows 10，Intel Core i7-6600U @ 2.60GHz，8 GB RAM，64位。

代码：

    import cplex
    import sys
    from cplex.callbacks import MIPInfoCallback

    class TimeLimitCallback(MIPInfoCallback):

        def __call__(self):
            if not self.aborted and self.has_incumbent():
                gap = 100.0 * self.get_MIP_relative_gap()
                timeused = self.get_time() - self.starttime
                if timeused > self.timelimit:
                    print("Good enough solution at", timeused, "sec., gap =",
                          gap, "%, quitting.")
                    self.aborted = True
                    self.abort()

    def solveCplex(filename):
        c = cplex.Cplex(filename)
        timelim_cb = c.register_callback(TimeLimitCallback)
        timelim_cb.starttime = c.get_time()
        timelim_cb.timelimit = 3600
        timelim_cb.aborted = False

        c.parameters.workmem.set(64)
        c.parameters.mip.strategy.file.set(3)
        c.parameters.workdir.set(r'C:\Users\username\folder')
        start_time = timelim_cb.starttime
        c.solve()
        end_time = c.get_time()
        print '\n'+'printing solutions'+'\n'

        obj_val = c.solution.get_objective_value()
        best_obj_val = c.solution.MIP.get_best_objective()
        abs_gap = best_obj_val - obj_val
        rel_gap = 100 * c.solution.MIP.get_mip_relative_gap()
        sol_time = end_time - start_time
        return [obj_val, abs_gap, rel_gap, sol_time]

    solveCplex('filename.sav')

CPLEX Python API：为什么CPLEX 12.8在（主要）默认参数设置上花费的时间比CPLEX 12.5要多？

0 个答案: