Update NLPModelsJuMP for MathOptInterface #37

amontoison · 2020-02-14T00:53:33Z

Interface between JuMP models (only composed of @NLobjective and @NLconstraint) and NLPModels.
#22

abelsiqueira · 2020-02-14T15:09:14Z

docs/src/tutorial.md


-```@example jumpnlp
+```julia


Does this work? @example jumpnlp was supposed to keep the previous environment.

Yes, but the code is not run. It's what I wanted because OptimizationProblems.jl is not working with JuMP 0.20.

I tested locally with an updated version of OptimizationProblems.jl and it works.
amontoison/OptimizationProblems.jl@cf02b59

I modified problems such that we have only @NLconstraint, @NLobjective and @NLexpression, it's why I didn't open a pull request.

Otherwise all models work without modifications on JuMP 0.20, we only need to replace setvalue by set_start_value to remove warnings.

abelsiqueira · 2020-02-14T15:16:42Z

src/moi_nlp_model.jl

+  return Jtv
+end
+
+# Uncomment when :JacVec becomes available in MOI.


Is this still an issue?

Yes, JacVec is still an unsupported feature...

abelsiqueira · 2020-02-14T15:32:38Z

I noticed you removed NLS models. Can they be remade?
Also, this needs the latest NLPModels, right?

amontoison · 2020-02-14T15:56:58Z

I think it can be remade but at which cost.... They completely changed how linear / quadratic terms are treated inside MOI. If we find a way to parse them, we should build ourself the jacobian , hessian...

With MPB everything was accessible through an MPB.NLPEvaluator (linear or non linear objective / constraints) and a quadratic term was a non linear term. With MOI they made a distinction between a quadratic term a non linear term and only restrictive non linear objective / constraints are accessible through MOI.NLPEvaluator.

I update the Project.toml to use the last release of NLPModels. I didn't test with previous version.

abelsiqueira · 2020-02-14T16:28:05Z

You can't get the expressions anymore? MPBNLS was essentially creating constraints for an internal model using expressions.

abelsiqueira · 2020-02-14T16:30:09Z

I update the Project.toml to use the last release of NLPModels. I didn't test with previous version.

I meant the master version, because test/problems is not in the latest released version.

amontoison · 2020-02-14T17:48:17Z

No It doesn't need the masterversion of NLPModels.
Problems are in test/problems folder for NLPModelJuMP and test folder for NLPModel.

dpo · 2020-02-14T14:13:39Z

src/NLPModelsJuMP.jl

 module NLPModelsJuMP

-include("nlsmpb_model.jl")


Let's comment this out instead of deleting it.

dpo · 2020-02-14T14:17:14Z

src/moi_nlp_model.jl

+
+  nvar = num_variables(jmodel)
+  x0 = zeros(nvar)
+  lvar = []


Maybe initialize this to a Float64 array of -Inf for consistency with x0. Same for uvar.

src/moi_nlp_model.jl

dpo · 2020-02-15T01:47:38Z

src/moi_nlp_model.jl

+
+# Uncomment when :JacVec becomes available in MOI.
+#
+# function NLPModels.jprod(nlp :: MathOptNLPModel, x :: AbstractVector, v :: AbstractVector)


Should disappear.

dpo · 2020-02-15T01:47:55Z

src/moi_nlp_model.jl

+#   return jv
+# end
+
+# function NLPModels.jtprod(nlp :: MathOptNLPModel, x :: AbstractVector, v :: AbstractVector)


src/moi_nlp_model.jl

dpo · 2020-02-15T01:49:14Z

test/problems/hs10.jl

-  setvalue(x[1], -10)
-  setvalue(x[2],  10)
+  set_start_value(x[1], -10.0)
+  set_start_value(x[2],  10.0)


At other places, you left Ints as initial value. If they are accepted, I think 10 is clearer than 10.0.

amontoison · 2020-02-15T19:56:25Z

I didn't removed jac, jac_structure and hess_structure functions because their implementations for an MathOptNLPModel are more optimized than the generic ones for an AbstractNLPModel.

The main reason is that jrows, jcols, hrows and hcols are stored inside the model. jac_structure/hess_structure don't need to call jac_structure!/hess_structure! and jac don't need to call jac_structure.

dpo · 2020-02-15T19:56:56Z

.travis.yml

-addons:
-  apt_packages:
-    - gfortran # For Ipopt
-


This shouldn't be part of this PR.

I see that you just add it in #32. It should disappear after a rebase.

dpo · 2020-02-15T19:57:11Z

appveyor.yml

@@ -2,6 +2,8 @@ environment:
  matrix:
  - julia_version: 1.0
  - julia_version: 1.1
+  - julia_version: 1.2
+  - julia_version: 1.3


This is already in #32.

amontoison · 2020-02-18T20:56:21Z

I close this pull request, #38 includes MathOptNLPModel and MathOptNLSModel.

abelsiqueira reviewed Feb 14, 2020

View reviewed changes

dpo requested changes Feb 15, 2020

View reviewed changes

dpo reviewed Feb 15, 2020

View reviewed changes

dpo added a commit that referenced this pull request Feb 15, 2020

remove tests with IPOPT in preparation for #37

dda11ec

dpo added a commit that referenced this pull request Feb 15, 2020

remove tests with IPOPT in preparation for #37

aa253cf

dpo added a commit that referenced this pull request Feb 15, 2020

remove tests with IPOPT in preparation for #37

7dead47

dpo added a commit that referenced this pull request Feb 15, 2020

remove tests with IPOPT in preparation for #37

726642d

dpo added a commit that referenced this pull request Feb 16, 2020

remove tests with IPOPT in preparation for #37

7f9c445

dpo changed the title ~~Update NLPModelJuMP for MathOptInterface~~ Update NLPModelsJuMP for MathOptInterface Feb 17, 2020

amontoison added 4 commits February 17, 2020 11:14

Update NLPModelJuMP for MathOptInterface

aebb514

Update Travis / AppVeyor

d9a4244

Use the correct consistency path

325f3c1

Add some memory improvments

fce38f6

amontoison force-pushed the master branch from e6883ce to 1350f36 Compare February 17, 2020 16:14

Add JuMP 0.21 in Project.toml

34a8b8b

amontoison force-pushed the master branch from 1350f36 to 34a8b8b Compare February 17, 2020 16:17

Improve documentation and tests -- MathOptNLPModel

8533ae4

amontoison closed this Feb 18, 2020

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Update NLPModelsJuMP for MathOptInterface #37

Update NLPModelsJuMP for MathOptInterface #37

amontoison commented Feb 14, 2020 •

edited

Loading

abelsiqueira Feb 14, 2020

amontoison Feb 14, 2020

amontoison Feb 14, 2020

abelsiqueira Feb 14, 2020

amontoison Feb 14, 2020

abelsiqueira commented Feb 14, 2020

amontoison commented Feb 14, 2020 •

edited

Loading

abelsiqueira commented Feb 14, 2020

abelsiqueira commented Feb 14, 2020

amontoison commented Feb 14, 2020

dpo Feb 14, 2020

dpo Feb 14, 2020

dpo Feb 15, 2020

dpo Feb 15, 2020

dpo Feb 15, 2020

amontoison commented Feb 15, 2020 •

edited

Loading

dpo Feb 15, 2020

amontoison Feb 15, 2020

dpo Feb 15, 2020

amontoison commented Feb 18, 2020


		```@example jumpnlp
		```julia

Update NLPModelsJuMP for MathOptInterface #37

Update NLPModelsJuMP for MathOptInterface #37

Conversation

amontoison commented Feb 14, 2020 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

abelsiqueira commented Feb 14, 2020

amontoison commented Feb 14, 2020 • edited Loading

abelsiqueira commented Feb 14, 2020

abelsiqueira commented Feb 14, 2020

amontoison commented Feb 14, 2020

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

amontoison commented Feb 15, 2020 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

amontoison commented Feb 18, 2020

amontoison commented Feb 14, 2020 •

edited

Loading

amontoison commented Feb 14, 2020 •

edited

Loading

amontoison commented Feb 15, 2020 •

edited

Loading