Coin Cbc Warm Starts in Pulp

[cswor]

I've been working on some MIP problems, some which run days. I decided to log solutions and try use them to warm start the solver. I'm first to admit I'm probably doing something dumb, but seems like it ignores the warmStart=True flag. A sample of my code is:

create initial binary vars

for j in range(facs) :
y[j] = LpVariable('Y_'+str(j),0,1,LpBinary)

for i in range(cuss) :
    for j in range(facs) :
        x[i,j] = LpVariable('X_'+str(i)+'_'+str(j),0,1,LpBinary)

I later read prev solve results (usually quit before optimality due to relOptGap or timeLimit. I set the vars as follows:

make sure all are initially 0

    for j in range(facs) :
        y[j].setInitialValue(0.0)
    for i in range(cuss) :
        for j in range(facs) :
            x[i,j].setInitialValue(0.0)

set variable values to those from prev solve

     for i in range(cuss) :
        j = vals[i] # values from a previous solve vals[i] is the loc# cust i assigned
        y[j].setInitialValue(1.0)
        x[i,j].setInitialValue(1.0)

my solve statement

   prob.solve(PULP_CBC_CMD(msg=True, warmStart = True, 
         timeLimit=timeLim))

Cbc seems to start from a fresh solve with or without the warmStart flag. The new values I set are feasible and the result of a previous solve of this same problem which terminated early, so it was not optimal, but it is feasible. The solve instead goes thru a lot of effort (same in both cases) to reduce the suminf but there should be none with the warmstart data.

If someone has the time to help, it would be much appreciated; even a pointer to a good tutorial/example in Pulp that would allow me to help myself. I modelled my code off an example from Lehigh(?) . I'm stuck for the moment.

Thanks in advance.

[jjhforrest]

What does the output look like? I downloaded pulp and ran two examples wedding.py and then wedding_initial.py and got the following - Problem MODEL has 18 rows, 3213 columns and 15062 elements Coin0008I MODEL read with 0 errors will open mipstart file /tmp/b2156f6c98a345c68fbc91f4343a02a3-pulp.mst. mipstart values read for 3213 variables. and Cbc0045I mipstart provided solution with cost 12 What did you get on your run? John Forrest

…

On 05/03/2021 01:32, cswor wrote: I've been working on some MIP problems, some which run days. I decided to log solutions and try use them to warm start the solver. I'm first to admit I'm probably doing something dumb, but seems like it ignores the warmStart=True flag. A sample of my code is: create initial binary vars for j in range(facs) : y[j] = LpVariable('Y_'+str(j),0,1,LpBinary) |for i in range(cuss) : for j in range(facs) : x[i,j] = LpVariable('X_'+str(i)+'_'+str(j),0,1,LpBinary) | I later read prev solve results (usually quit before optimality due to relOptGap or timeLimit. I set the vars as follows: make sure all are initially 0 |for j in range(facs) : y[j].setInitialValue(0.0) for i in range(cuss) : for j in range(facs) : x[i,j].setInitialValue(0.0) | set variable values to those from prev solve |for i in range(cuss) : j = vals[i] # values from a previous solve vals[i] is the loc# cust i assigned y[j].setInitialValue(1.0) x[i,j].setInitialValue(1.0) | my solve statement |prob.solve(PULP_CBC_CMD(msg=True, warmStart = True, timeLimit=timeLim)) | Cbc seems to start from a fresh solve with or without the warmStart flag. The new values I set are feasible and the result of a previous solve of this same problem which terminated early, so it was not optimal, but it is feasible. The solve instead goes thru a lot of effort (same in both cases) to reduce the suminf but there should be none with the warmstart data. If someone has the time to help, it would be much appreciated; even a pointer to a good tutorial/example in Pulp that would allow me to help myself. I modelled my code off an example from Lehigh(?) . I'm stuck for the moment. Thanks in advance. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#369>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABWJYHFBQ2OAJDCAJ5FFL3DTCAYEJANCNFSM4YUKF3NQ>.

[cswor]

Thanks for the prompt reply. Given your response, I assume that the warmstart should be working. Is that the case?

I have rerun the seating problem for the wedding and with warmStart=True, get the following and perhaps my problem is in the first 3 lines below (it's old circa 2015):

Welcome to the CBC MILP Solver
Version: 2.9.0
Build Date: Feb 12 2015

command line - C:\Users\cswil\anaconda3\lib\site-packages\pulp\apis..\solverdir\cbc\win\64\cbc.exe C:\Users\cswil\AppData\Local\Temp\ac8cd07d396549569959f3efa58d6bc9-pulp.mps mips C:\Users\cswil\AppData\Local\Temp\ac8cd07d396549569959f3efa58d6bc9-pulp.mst branch printingOptions all solution C:\Users\cswil\AppData\Local\Temp\ac8cd07d396549569959f3efa58d6bc9-pulp.sol (default strategy 1)
At line 2 NAME MODEL
At line 3 ROWS
At line 23 COLUMNS
At line 24708 RHS
At line 24727 BOUNDS
At line 27941 ENDATA
Problem MODEL has 18 rows, 3213 columns and 15062 elements
Coin0008I MODEL read with 0 errors
will open mipstart file .\C:\Users\cswil\AppData\Local\Temp\ac8cd07d396549569959f3efa58d6bc9-pulp.mst.
Continuous objective value is 12 - 0.01 seconds
Cgl0004I processed model has 18 rows, 3213 columns (3213 integer (3213 of which binary)) and 15062 elements
Cutoff increment increased from 1e-005 to 0.9999
Cbc0038I Initial state - 0 integers unsatisfied sum - 0
Cbc0038I Solution found of 12
Cbc0038I Before mini branch and bound, 3213 integers at bound fixed and 0 continuous
Cbc0038I Mini branch and bound did not improve solution (0.07 seconds)
Cbc0038I After 0.07 seconds - Feasibility pump exiting with objective of 12 - took 0.01 seconds
Cbc0012I Integer solution of 12 found by feasibility pump after 0 iterations and 0 nodes (0.07 seconds)
Cbc0001I Search completed - best objective 12, took 0 iterations and 0 nodes (0.07 seconds)
Cbc0035I Maximum depth 0, 0 variables fixed on reduced cost
Cuts at root node changed objective from 12 to 12
Probing was tried 0 times and created 0 cuts of which 0 were active after adding rounds of cuts (0.000 seconds)
Gomory was tried 0 times and created 0 cuts of which 0 were active after adding rounds of cuts (0.000 seconds)
Knapsack was tried 0 times and created 0 cuts of which 0 were active after adding rounds of cuts (0.000 seconds)Clique was tried 0 times and created 0 cuts of which 0 were active after adding rounds of cuts (0.000 seconds)
MixedIntegerRounding2 was tried 0 times and created 0 cuts of which 0 were active after adding rounds of cuts (0.000 seconds)
FlowCover was tried 0 times and created 0 cuts of which 0 were active after adding rounds of cuts (0.000 seconds)
TwoMirCuts was tried 0 times and created 0 cuts of which 0 were active after adding rounds of cuts (0.000 seconds)

Result - Optimal solution found

Objective value: 12.00000000

I have also redone the problem solve that is causing me the problem and even though I am certain it is a feasible solution, it still starts from scratch, disregarding the advanced starting point, and runs thru the attempt to eliminate a number of infeasibilities.

Maybe my problem is simply an outdated version of CBC delivered with the fresh Pulp install. The attached file has both without and with warmstart on the seating problem.

warmstart_results.txt

[jjhforrest]

Yes - that does seem to be an old version - but that may not be the problem. warmstart just ignores the file if it can't be opened - should put out a message (I can fix that). On my Linux run I got - mipstart values read for 3213 variables. while on your Windows run you did not - so probably it could not open file. Maybe some Pulp expert (or Cbc expert) who runs on Windows could check out that example. John Forrest

…

On 05/03/2021 22:50, cswor wrote: Thanks for the prompt reply. Given your response, I assume that the warmstart should be working. Is that the case? I have rerun the seating problem for the wedding and with warmStart=True, get the following and perhaps my problem is in the first 3 lines below (it's old circa 2015): Welcome to the CBC MILP Solver Version: 2.9.0 Build Date: Feb 12 2015 command line - C:\Users\cswil\anaconda3\lib\site-packages\pulp\apis..\solverdir\cbc\win\64\cbc.exe C:\Users\cswil\AppData\Local\Temp\ac8cd07d396549569959f3efa58d6bc9-pulp.mps mips C:\Users\cswil\AppData\Local\Temp\ac8cd07d396549569959f3efa58d6bc9-pulp.mst branch printingOptions all solution C:\Users\cswil\AppData\Local\Temp\ac8cd07d396549569959f3efa58d6bc9-pulp.sol (default strategy 1) At line 2 NAME MODEL At line 3 ROWS At line 23 COLUMNS At line 24708 RHS At line 24727 BOUNDS At line 27941 ENDATA Problem MODEL has 18 rows, 3213 columns and 15062 elements Coin0008I MODEL read with 0 errors will open mipstart file .\C:\Users\cswil\AppData\Local\Temp\ac8cd07d396549569959f3efa58d6bc9-pulp.mst. Continuous objective value is 12 - 0.01 seconds Cgl0004I processed model has 18 rows, 3213 columns (3213 integer (3213 of which binary)) and 15062 elements Cutoff increment increased from 1e-005 to 0.9999 Cbc0038I Initial state - 0 integers unsatisfied sum - 0 Cbc0038I Solution found of 12 Cbc0038I Before mini branch and bound, 3213 integers at bound fixed and 0 continuous Cbc0038I Mini branch and bound did not improve solution (0.07 seconds) Cbc0038I After 0.07 seconds - Feasibility pump exiting with objective of 12 - took 0.01 seconds Cbc0012I Integer solution of 12 found by feasibility pump after 0 iterations and 0 nodes (0.07 seconds) Cbc0001I Search completed - best objective 12, took 0 iterations and 0 nodes (0.07 seconds) Cbc0035I Maximum depth 0, 0 variables fixed on reduced cost Cuts at root node changed objective from 12 to 12 Probing was tried 0 times and created 0 cuts of which 0 were active after adding rounds of cuts (0.000 seconds) Gomory was tried 0 times and created 0 cuts of which 0 were active after adding rounds of cuts (0.000 seconds) Knapsack was tried 0 times and created 0 cuts of which 0 were active after adding rounds of cuts (0.000 seconds)Clique was tried 0 times and created 0 cuts of which 0 were active after adding rounds of cuts (0.000 seconds) MixedIntegerRounding2 was tried 0 times and created 0 cuts of which 0 were active after adding rounds of cuts (0.000 seconds) FlowCover was tried 0 times and created 0 cuts of which 0 were active after adding rounds of cuts (0.000 seconds) TwoMirCuts was tried 0 times and created 0 cuts of which 0 were active after adding rounds of cuts (0.000 seconds) Result - Optimal solution found Objective value: 12.00000000 I have also redone the problem solve that is causing me the problem and even though I am certain it is a feasible solution, it still starts from scratch, disregarding the advanced starting point, and runs thru the attempt to eliminate a number of infeasibilities. Maybe my problem is simply an outdated version of CBC delivered with the fresh Pulp install. The attached file has both without and with warmstart on the seating problem. warmstart_results.txt <https://github.com/coin-or/Cbc/files/6093941/warmstart_results.txt> — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#369 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABWJYHBT7SXBAL5MDQ5GKRTTCFN2BANCNFSM4YUKF3NQ>.

[U-Coding]

Hello,

I had the same problem with pulp (version 2.4, cbc version 2.10.3) and experimented a little bit with the wedding example.

It did work when I set "keepFiles = True": PULP_CBC_CMD(msg=True, warmStart = True, keepFiles = True)

So there seems to be some problem with opening/finding the files. I will post this in pulp discussions.

Thank you and best regards

[tkralphs]

Just to note, looks like this might be due to #32 (see coin-or/pulp#411).