New GAMS solver interface, writer, and solution loader. #3683
Conversation
…ittest and coverage test
|
@AnhTran01 - Please make sure to run |
…riting process. Added WIP unittest for the new solver interface
… and writing to gms.
| @@ -518,6 +518,11 @@ def write(self, model): | |||
| con_list[label] = declaration + definition + bounds | |||
| self.var_symbol_map.addSymbol(obj, label) | |||
|
|
|||
| # Write the GAMS model | |||
| ostream.write("$offlisting\n") | |||
| # $offdigit ignores extra precise digits instead of erroring | |||
There was a problem hiding this comment.
what was the error that you were seeing that prompted this addition? maybe something was being set above 1e300?
There was a problem hiding this comment.
in the previous commit test (b0c3450), the listing file output error message is "Too many digits in number $offdigit can be used to ignore trailing digits"
There was a problem hiding this comment.
but do you know what triggered this? which test was causing this error?
There was a problem hiding this comment.
I went back to check the failed test. The following tests that triggered the "Too many digits error" are:
- TestSolvers.test_equality
- TestSolvers.test_linear_expression
- TestSolvers.test_mutable_param_with_range
- TestSolvers.test_param_changes
- TestSolvers.test_no_objective
- TestSolvers.test_time_limit
- TestLegacySolverInterface.test_param_updates
By testing locally, I see there are trailing decimal places and not the cases of 1e300 or large numbers.
… constraint). Fixed bug of load_solution that contains constraint id in addition to variable id
…arsing solver_options
pyomo/contrib/solver/solvers/gams.py
Outdated
| self.writer_config: ConfigDict = self.declare( | ||
| 'writer_config', GAMSWriter.CONFIG() | ||
| ) | ||
| # Share the same config as the writer, passes at the function call write |
There was a problem hiding this comment.
Here and below: this should just insert a copy of the CONFIG from the writer. See Ipopt here for an example.
pyomo/contrib/solver/solvers/gams.py
Outdated
| n, | ||
| ) | ||
|
|
||
| def version(self, config=None): |
There was a problem hiding this comment.
In our review of the new solver interfaces, we changed the API to:
def version(self, rehash: bool=False) -> Tuple:Much like available(), this should leverage a cache (class dict attribute). You can look at the ipopt implementation for an idea (although that hasn't been updated to the new API).
There was a problem hiding this comment.
Could you clarify the purpose of using rehash and whether the user can call rehash from the solve statement?
I found the documentation here: https://pyomo.readthedocs.io/en/latest/api/pyomo.common.fileutils.PathManager.html
Would the implementation be Executable(config.executable.path()).rehash()?
pyomo/contrib/solver/solvers/gams.py
Outdated
| if config.logfile is not None: | ||
| config.logfile = os.path.abspath(config.logfile) | ||
|
|
||
| config.writer_config.put_results_format = 'gdx' if gdxcc_available else 'dat' |
There was a problem hiding this comment.
This should probably only overwrite the configuration if it is still None (that way users can set / override the interface)
…lable() and version()
…are no longer valid.
…etting default to gdx that caused failed test. Implemented raise infeasible
… rehash handling with Executable.rehash()
…age for infeasible model when calling load_var, dual, reduced_cost
…ded handling solvestat and modelstat when solver is interrupted
|
@mrmundt is the |
Fixes NA
Summary/Motivation:
This PR introduced a new GAMS solver interface that uses the
LinearRepnVisitorto enable coefficient gathering when writing out the scalar GAMS model ingmsformat.Changes proposed in this PR:
pyomo.contrib.solver.solverspyomo.repn.pluginspyomo.contrib.solver.solversLegal Acknowledgement
By contributing to this software project, I have read the contribution guide and agree to the following terms and conditions for my contribution: