Hi all,

I have been working with CPLEX for some years for a new platform we are developing. We bought some license 12.10 to use CPLEX in a distributed parallel optimization (here the logic)

Then IBM released the new CPLEX 20.X (now 22) and talking with them I understand that now it should be possible to use Cplex as a Service inside Watson Studio. So I am imagining to use Cplex not calling my remote VM but some online machines (provided by IBM) to resolve my problems.

I created a free account for Watson Studio but I am really struggling with this because I have looked for the mechanism but I did not understand if it works this way and how it should work if it is possible.

I have found this example.

The questions: is it possible to create a Cplex problem locally (pyomo + python) and send it to a Cplex as a service to receive the results like I did with the distributed parallel optimization? Is it the right way or am I missing something?

Thank you all for your time.

Best Regards

Riccardo

Hi,

in watsonstudio the python API for CPLEX is docplex so what you can do is generate your model with pyomo and then import it through docplex like in https://github.com/AlexFleischerParis/zoodocplex/blob/master/zoopyomogenerateanddocplexsolve.py

Or you can generate your .sav file with pyomo and then send this .sav file to the IBM Cloud

See https://medium.com/@AlainChabrier/cplex-and-cpo-file-formats-with-do-experiments-in-cloud-pak-for-data-a2abf93c9c18

Hi all,

I have been working with CPLEX for some years for a new platform we are developing. We bought some license 12.10 to use CPLEX in a distributed parallel optimization (here the logic)

Then IBM released the new CPLEX 20.X (now 22) and talking with them I understand that now it should be possible to use Cplex as a Service inside Watson Studio. So I am imagining to use Cplex not calling my remote VM but some online machines (provided by IBM) to resolve my problems.

I created a free account for Watson Studio but I am really struggling with this because I have looked for the mechanism but I did not understand if it works this way and how it should work if it is possible.

I have found this example.

The questions: is it possible to create a Cplex problem locally (pyomo + python) and send it to a Cplex as a service to receive the results like I did with the distributed parallel optimization? Is it the right way or am I missing something?

Thank you all for your time.

Best Regards

Riccardo

Hi Alex,

I really appreciate the help. I managed to create a mlp file, load it manually on cloud park and solve it. I would like to ask you something else: is it possible to load the model automatically through a Rest API on Cloud Park or Watson? We are running multiple optimisation every 15 min or every hour and, of course, the manual loading is not a possibility.

Thank you

Riccardo

Hi,

in watsonstudio the python API for CPLEX is docplex so what you can do is generate your model with pyomo and then import it through docplex like in https://github.com/AlexFleischerParis/zoodocplex/blob/master/zoopyomogenerateanddocplexsolve.py

Or you can generate your .sav file with pyomo and then send this .sav file to the IBM Cloud

See https://medium.com/@AlainChabrier/cplex-and-cpo-file-formats-with-do-experiments-in-cloud-pak-for-data-a2abf93c9c18

Hi all,

I have been working with CPLEX for some years for a new platform we are developing. We bought some license 12.10 to use CPLEX in a distributed parallel optimization (here the logic)

Then IBM released the new CPLEX 20.X (now 22) and talking with them I understand that now it should be possible to use Cplex as a Service inside Watson Studio. So I am imagining to use Cplex not calling my remote VM but some online machines (provided by IBM) to resolve my problems.

I created a free account for Watson Studio but I am really struggling with this because I have looked for the mechanism but I did not understand if it works this way and how it should work if it is possible.

I have found this example.

The questions: is it possible to create a Cplex problem locally (pyomo + python) and send it to a Cplex as a service to receive the results like I did with the distributed parallel optimization? Is it the right way or am I missing something?

Thank you all for your time.

Best Regards

Riccardo

Hi,

indeed you can do the same with APIs

https://medium.com/@AlainChabrier/solve-lp-models-with-do-for-wml-5299da95c5ad

regards

Hi Alex,

I really appreciate the help. I managed to create a mlp file, load it manually on cloud park and solve it. I would like to ask you something else: is it possible to load the model automatically through a Rest API on Cloud Park or Watson? We are running multiple optimisation every 15 min or every hour and, of course, the manual loading is not a possibility.

Thank you

Riccardo

Hi,

in watsonstudio the python API for CPLEX is docplex so what you can do is generate your model with pyomo and then import it through docplex like in https://github.com/AlexFleischerParis/zoodocplex/blob/master/zoopyomogenerateanddocplexsolve.py

Or you can generate your .sav file with pyomo and then send this .sav file to the IBM Cloud

See https://medium.com/@AlainChabrier/cplex-and-cpo-file-formats-with-do-experiments-in-cloud-pak-for-data-a2abf93c9c18

Hi all,

I have been working with CPLEX for some years for a new platform we are developing. We bought some license 12.10 to use CPLEX in a distributed parallel optimization (here the logic)

Then IBM released the new CPLEX 20.X (now 22) and talking with them I understand that now it should be possible to use Cplex as a Service inside Watson Studio. So I am imagining to use Cplex not calling my remote VM but some online machines (provided by IBM) to resolve my problems.

I created a free account for Watson Studio but I am really struggling with this because I have looked for the mechanism but I did not understand if it works this way and how it should work if it is possible.

I have found this example.

The questions: is it possible to create a Cplex problem locally (pyomo + python) and send it to a Cplex as a service to receive the results like I did with the distributed parallel optimization? Is it the right way or am I missing something?

Thank you all for your time.

Best Regards

Riccardo

Hi Alex,

we managed to run a job on Watson in an automatic way! thank you again

{'CPLEXSolution': {'@version': '1.2', 'header': {'@writeLevel': '1', '@solutionName': 'incumbent', '@solutionTypeString': 'primal', '@problemName': 'model_opt.lp', '@solutionStatusString': 'integer optimal, tolerance', '@solutionStatusValue': '102', '@dualFeasible': '0', '@primalFeasible': '1', '@solutionTypeValue': '3', '@solutionMethodString': 'mip', '@solutionIndex': '-1', '@MIPNodes': '0', '@MIPIterations': '1266', '@objectiveValue': '98.846203769477228'}, 'quality': {'@epRHS': '9.9999999999999995e-07', '@epInt': '1.0000000000000001e-05', '@maxX': '9000', '@maxPrimalInfeas': '4.0971871302361387e-08', '@maxSlack': '10000000', '@maxIntInfeas': '6.0825600000001143e-09'}, 'linearConstraints': {'constraint': [{'@slack': '1.4210854715202004e-14', '@name': 'c_e_opex_calc_', '@index': '0'}, {'@slack': '0', '@name': 'c_e_penalty_withdrawn_', '@index': '1'}, {'@slack': '0', '@name': 'c_e_penalty_slack_calc_', '@index': '2'}, {'@slack': '0', '@name': 'c_e_imbalance_penalty_calc_', '@index': '...52'}, {'@slack': '0.52590000000054715', '@name': 'c_u_EX_b(ESSA_EX_0001)_max_withdrawn_calc(49)_', '@index': '53'}, {'@slack': '0.52590000000054715', '@name': 'c_u_EX_b(ESSA_EX_0001)_max_withdrawn_calc(50)_', '@index': '54'}, {'@slack': '0.52590000000054715', '@name': 'c_u_EX_b(ESSA_EX_0001)_max_withdrawn_calc(51)_', '@index': '55'}, {'@slack': '0.52590000000054715', '@name': 'c_u_EX_b(ESSA_EX_0001)_max_withdrawn_calc(52)_', '@index': '56'}, {'@slack': '8059.7299999999996', '@name': 'c_u_EX_b(ESSA_EX_0001)_max_withdrawn_calc(53)_', '@index': '57'}, {'@slack': '8059.7299999999996', '@name': 'c_u_EX_b(ESSA_EX_0001)_max_wi ...

this is a json view of the results.

May I ask you if there is a simple way to load it in the Pyomo ConcreteModel in order to use the same code we are using with the distributed cplex?

Usually we do this and then we parse the pyomo instance to get the results

opt = pyopt.SolverFactory("cplex_direct", solver_io='python')

self.instance_results = opt.solve(self.instance_ems_model, symbolic_solver_labels=True, tee=True)

thank you

Riccardo

Hi,

indeed you can do the same with APIs

https://medium.com/@AlainChabrier/solve-lp-models-with-do-for-wml-5299da95c5ad

regards

Hi Alex,

I really appreciate the help. I managed to create a mlp file, load it manually on cloud park and solve it. I would like to ask you something else: is it possible to load the model automatically through a Rest API on Cloud Park or Watson? We are running multiple optimisation every 15 min or every hour and, of course, the manual loading is not a possibility.

Thank you

Riccardo

Hi,

in watsonstudio the python API for CPLEX is docplex so what you can do is generate your model with pyomo and then import it through docplex like in https://github.com/AlexFleischerParis/zoodocplex/blob/master/zoopyomogenerateanddocplexsolve.py

Or you can generate your .sav file with pyomo and then send this .sav file to the IBM Cloud

See https://medium.com/@AlainChabrier/cplex-and-cpo-file-formats-with-do-experiments-in-cloud-pak-for-data-a2abf93c9c18

Hi all,

I have been working with CPLEX for some years for a new platform we are developing. We bought some license 12.10 to use CPLEX in a distributed parallel optimization (here the logic)

Then IBM released the new CPLEX 20.X (now 22) and talking with them I understand that now it should be possible to use Cplex as a Service inside Watson Studio. So I am imagining to use Cplex not calling my remote VM but some online machines (provided by IBM) to resolve my problems.

I created a free account for Watson Studio but I am really struggling with this because I have looked for the mechanism but I did not understand if it works this way and how it should work if it is possible.

I have found this example.

The questions: is it possible to create a Cplex problem locally (pyomo + python) and send it to a Cplex as a service to receive the results like I did with the distributed parallel optimization? Is it the right way or am I missing something?

Thank you all for your time.

Best Regards

Riccardo