Releases: ansys/pysimai
Releases · ansys/pysimai
v0.2.4
New:
- Auth tokens are now cached in file system and get re-authenticated in a parallel fashion.
- Invalid refresh token now triggers a reauth instead of crashing.
build_preset
option in classModelConfiguration
can now be one ofdebug
,1_day
,2_days
,7_days
.- Model Evaluation Report data (csv file) can now be downloaded with
Workspace.download_mer_data
. - Typing improvements; introducing
JSON
type is introduced andAPIResponse
type is updated to includeJSON
type. - New property
raw_confidence_score
is added to classPrediction
, which returns the raw confidence score.
Fix:
- Fixed the error where
GlobalCoefficients.data
was not in coordinance with the BE response.GlobalCoefficients.data
now runs without errors.
v0.2.3
New:
- Added :py:class:
PostProcessInput<ansys.simai.core.data.model_configuration.PostProcessInput>
class to define post processing input fields. - Added support for NaN and Inf for Global Coefficients and Post Processings.
Fixes:
- Removed compute argument from :py:meth:
TrainingData.upload_folder()<ansys.simai.core.data.training_data.TrainingData.upload_folder>
- Fixed Model Configuration to raise a ProcessingError when volume field is missing from a sample specifying volume output.
- Removed wakepy error mode success (deprecated) during optimization.
- Renamed TrainingData method compute() to :py:meth:
TrainingData.extract_data()<ansys.simai.core.data.training_data.TrainingData.extract_data>
. - Updated documentation of :py:meth:
GeometryDirectory.upload()<ansys.simai.core.data.geometries.GeometryDirectory.upload>
: theworkspace_id
argument was moved toworkspace
but never updated.
v0.2.2
New:
- Added support for the postprocessing of custom volume of point cloud. Use :py:meth:
Geometry.upload_point_cloud<ansys.simai.core.data.geometries.Geometry.upload_point_cloud>
to upload a point cloud file on a geometry and run the post processing through :py:meth:Prediction.post.custom_volume_point_cloud<ansys.simai.core.data.post_processings.PredictionPostProcessings.custom_volume_point_cloud>
to run the postprocessing.
Fix:
- Remove internal uses of deprecated
workspace.model
v0.2.1
Fixes:
- Fixed bug that was crashing method
ModelConfiguration.compute_global_coefficient
. The result of the Global Coefficient formula can now be retrieved.
v0.2.0
New:
- Model configuration can now be created from scratch and be used in training requests.
- Training-data subsets can now be assigned to
None
. OptionsIgnored
andValidation
are retired.
Fixes:
- Fixed bug when uploading large files. Large files can now be uploaded.
- Fixed bug when listing prediction without current_workspace being set.
v0.1.7
New:
- Added :py:class:
DomainOfAnalysis<ansys.simai.core.data.models.DomainOfAnalysis>
class to help set the domain of analysis on a new model. - Add workspace option where we previously relied only on the global workspace
- Add prediction.post.list()
Fixes:
- Reestablish python 3.9 compatibility.
- Bump wakepy lib to fix errors when not able to prevent sleep during optimization.
v0.1.6
New:
- Added new method :py:meth:
TrainingData.assign_subset()<ansys.simai.core.data.training_data.TrainingData.assign_subset>
that allows you to assign a Train, Validation, or Test subset to your data.
Fixes:
- The method :py:meth:
Optimization.run()<ansys.simai.core.data.optimizations.OptimizationDirectory.run>
now raises an exception if no workspace is provided and none is configured. - Fix RecursionError on authentication refresh
Full Changelog: v0.1.5...v0.1.6
v0.1.5
- Training can now be launched using the most recent model configuration from a project.
- Enabled non-interactive mode capability, allowing for automation or operations without manual inputs.
- Added new validation
Project.is_trainable()
to verify if the project meets all minimum requirements for training. - Added new method
Project.get_variables()
to get all available variables used for a model's inputs and outputs. - Fixed bug where a subset of training data could not be pulled. A subset of training data is now correctly retrieved.
- Fixed erroneous call to a private function during the optimization run.
New Contributors
- @kliment-slice made their first contribution in #36
Full Changelog: v0.1.4...v0.1.5
v0.1.4
What's Changed
- bump: ruff 0.2.x by @tmpbeing in #27
- Less verbose SSE disconnects by @awoimbee in #28
- fix: Double description of client config vars by @tmpbeing in #29
- build(deps): bump codecov/codecov-action from 3 to 4 by @dependabot in #25
- Fix ModelManifest available-post-processings by @awoimbee in #31
- Direction for Windows-developer mode by @yias in #30
- fix monitor_callback interface not respected in upload_file_with_presigned_post by @awoimbee in #32
- fix: weird main README spacings by @RobPasMue in #33
- Release 0.1.4 by @awoimbee in #34
Full Changelog: v0.1.3...v0.1.4
v0.1.3
What's Changed
- dev: add codecov action by @tmpbeing in #20
- fix: training data upload_folder method by @tmpbeing in #21
- build(deps): bump pdm-project/setup-pdm from 3 to 4 by @dependabot in #23
- build(deps): bump peter-evans/create-or-update-comment from 2 to 4 by @dependabot in #22
- Merge release branch into main by @awoimbee in #24
- Fix arguments ignored if no config file found by @awoimbee in #26
Full Changelog: v0.1.2...v0.1.3