Skip to content

Releases: mdolab/pyoptsparse

v2.0.2

05 Mar 21:54
32a61fe
Compare
Choose a tag to compare

pyOptSparse v2.0.2 release notes

This is a bugfix release to address #80 related to hotstarts when saving the history file.

v2.0.1

26 Feb 22:52
907b558
Compare
Choose a tag to compare

pyOptSparse v2.0.1 release notes

This is a bugfix release. We have removed the git commit hash from the history file due to

  1. unless the code is pip-installed in place (via -e), the approach will not work in extracting the hash
  2. there's a possibility that this causes some segfaults under MPI with PETSc. We're not sure if this is the root cause, but since we haven't fully debugged this we're removing it as a precaution.

v2.0

19 Feb 18:51
e772e84
Compare
Choose a tag to compare

pyOptSparse v2.0 release notes

Features

  • Support for SNOPT v7.2 has been removed. From this version forward the code will not be compatible with SNOPT v7.2. The currently tested version of SNOPT is v7.7. This is due to pySNOPT switching to a different entry point for SNOPT (snkerc instead of snoptc) which allows for a user-specified snstop function. This allows us to extract optimizer information at major iterations.
  • (#39) Major iteration counting has been implemented for SNOPT. All callCounters will now contain the key isMajor which will be True if the point evaluated was for a major step. Note that this is only supported for SNOPT, and isMajor will be set to True for all call counters for other optimizers.
  • (#39) Additional information for the major iterations can now be stored when using SNOPT. The information stored can be selected via the option Save major iteration variables option. The choices are:
    • step
    • merit
    • feasibility
    • optimality
    • penalty
    • Hessian
    • slack
    • lambda
    • condZHZ.
      The first five values are stored at every major iteration by default. Be careful with the Hessian option as the full Hessian matrix can be rather large, and unnecessarily bloat the history file size.
  • (#77) A new metadata entry is now stored in the history file. This encodes the following information:
    • optName: name of the optimization
    • optOptions: all the optimization options used
    • nprocs: number of procs used
    • startTime: starting time as a string
    • endTime: end time as a string
    • optTime: optimization time in seconds
    • optimizer: name of the optimizer
    • version: the pyOptSparse version
    • gitHash: the git hash of the code used (as discussed in #42)

Bugfixes

v1.2

05 Feb 23:16
91909c4
Compare
Choose a tag to compare

This is another minor release. Notable changes are:

  • sqlitedict is now an external dependency, no longer provided by pyoptsparse itself
  • a requirements.txt file is provided to facilitate installation of Python dependencies
  • IPOPT wrapper has been updated to work with Python 3 (we have not updated the version of IPOPT supported)
  • various minor documentation updates
  • pyoptsparse is now continuously tested on Travis-CI

v1.1

05 Dec 18:42
bb90d52
Compare
Choose a tag to compare

This release mainly provides the feature in PR #63, which allows for triggering of user-requested termination in SNOPT. See POEM 002 for the rationale and additional information.

Additional bug fixes:

v1.0

23 Oct 02:25
6d2ae0a
Compare
Choose a tag to compare

This is a release of the current codebase as we transition to using software versioning. As such, there are no major changes to the code, we are simply releasing the existing version as v1.0. The release_notes.md file highlights some recent changes.