hinter solution feasible at presolve, but infeasible at preload

[YINXIONG CHEN]

I am using cp.SAT 9.13, cpModel and cpSolver for a scheduling problem. Basically is to schedule jobs to workers. There are 6 objectives, their weighted sum is the final objective. 

To speed up, I firstly optimized 5 objectives. Get a solution from minimizing the weighted sum of the 5 objectives. The solution becomes the hinter to the next solver.

Secondly, optimize weighted sum of 6 objectives together. Models are exactly the same. The additional objective in the 2nd step is the sum(work start of every jobs), try to schedule the work earlier in the span.

Why adding additional objective will cause infeasible? It is not additional constraint.

Log as below, 

Starting presolve at 0.01s
The solution hint is complete and is feasible. Its objective value is 1.20001067e+10.

...

Preloading model.
#Bound   0.58s best:inf   next:[-4800,4.80002712e+10] initial_domain
The solution hint is complete, but it is infeasible! we will try to repair it.
#Model   0.58s var:1708/1708 constraints:2650/2650

Starting search at 0.58s with 8 workers.
6 full problem subsolvers: [core, default_lp, max_lp_sym, no_lp, quick_restart, reduced_costs]
2 first solution subsolvers: [fj, fs_random_no_lp]
12 interleaved subsolvers: [feasibility_pump, graph_arc_lns, graph_cst_lns, graph_dec_lns, graph_var_lns, ls, rins/rens, rnd_cst_lns, rnd_var_lns, scheduling_intervals_lns, scheduling_precedences_lns, scheduling_time_window_lns]
3 helper subsolvers: [neighborhood_helper, synchronization_agent, update_gap_integral]
1 ignored subsolver: [fixed]

#1       0.72s best:3.60000267e+10 next:[-4800,3.60000267e+10] fj_restart_decay_perturb(batch:1 lin{mvs:404 evals:2'206} gen{mvs:559 evals:1'770} #w_updates:167 #perturb:0)

[YINXIONG CHEN]

I forgot to mention one more question. The next solver with hinter solution never start with hinter solution, it starts from scratch. 

I print the value of every objective in each search in the log. So I can see the unscheduled jobs objective always starts from the largest number. But I try to minimize it.

Why is that?

[Laurent Perron]

so try with 9.14

Laurent Perron | Operations Research | lpe...@google.com | (33) 1 42 68 53 00

--
You received this message because you are subscribed to the Google Groups "or-tools-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to or-tools-discu...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/or-tools-discuss/54f44a56-818c-4de6-ac3e-fe7ecd1cbfc3n%40googlegroups.com.

[Ruihua Zhang]

good job

YINXIONG CHEN <yinxio...@gmail.com> 于2025年8月22日周五 16:24写道：

--
You received this message because you are subscribed to the Google Groups "or-tools-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to or-tools-discu...@googlegroups.com.

To view this discussion visit https://groups.google.com/d/msgid/or-tools-discuss/4fcb5ba2-a4cc-4756-9239-54672eae8494n%40googlegroups.com.

[YINXIONG CHEN]

It works perfectly with 9.14. Thanks, Laurent.

I set

solver.parameters.keep_all_feasible_solutions_in_presolve = True

solver.parameters.hint_conflict_limit = 20
solver.parameters.repair_hint = True

Are there any other setting I need to pay attention to?

[Laurent Perron]

Do you need any of these with 9.14 ?

--Laurent

To view this discussion visit https://groups.google.com/d/msgid/or-tools-discuss/9a81a714-6157-4a8c-9909-09aa420c5d84n%40googlegroups.com.

[YINXIONG CHEN]

Hi Laurent

If I don't set solver.parameters.keep_all_feasible_solutions_in_presolve = True, hunter solution will not pass to the next solver run with 9.14. I tested. 

The other two is for any infeasible hunter solution, just in case.

You received this message because you are subscribed to a topic in the Google Groups "or-tools-discuss" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/or-tools-discuss/pPMcmritG3s/unsubscribe.
To unsubscribe from this group and all its topics, send an email to or-tools-discu...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/or-tools-discuss/CAPyyUTv%2B-2cGfHQwZXtT%2BJq712byFzTPX%2BUhL%2BD0Jzt7X6h_9A%40mail.gmail.com.

[Laurent Perron]

can you send me the proto of the model ?

Laurent Perron | Operations Research | lpe...@google.com | (33) 1 42 68 53 00

To view this discussion visit https://groups.google.com/d/msgid/or-tools-discuss/CAB3obBYZxPvKVAU5ZYdvQVZ%2BPVXYChCCMk6TU8-PgfTPwsdhRA%40mail.gmail.com.

[YINXIONG CHEN]

Hi Laurent

I am sorry I cannot share the model because I built it for a project in my company. :(

To view this discussion visit https://groups.google.com/d/msgid/or-tools-discuss/CABcmEebppeWieJR5p3GeuV198KBfZxaUvK-d0zSpBwCCLKQLCg%40mail.gmail.com.