Syncing #1
Merged
Syncing #1
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Fixes #982 * different Config files for train and test * solves problem of long running time * train benchmarks contain only one iteration as it gives more idea on how the users will use. (with no warmup iteration) * predict config is the original version
* Add analayzer to nuget * Generalize target file to enable easier analayzer inclusions in future
* Transform wrappers and a reference implementation for tokenizers * Added pigsty extensions * Added pigsty test * Fixed most important PR comments * PR comments * Converted all text transforms into transformers/estimators. * Addressed reviewers's comments. * Addressed reviewers' comments. * Converted LdaTransform into Transformer/Estimator. * Fixed LdaNative issue and addressed reviewers' comments. * Fixed issue with test. * Diabled end-to-end LdaTransform test due to incosistency of outputs.
* Microsoft.ML.Data.StaticPipe to Microsoft.ML.StaticPipe. * Columns now in StaticPipe as opposed to StaticPipe.Runtime.
…ssion) to estimator (#1002)
* Stop loading assemblies in ComponentCatalog. Write the AssemblyName into the model, and use it to register the assembly during model load. * Move ComponentCatalog from a static class to a member of IHostEnvironment. * Update tests for ComponentCatalog refactoring. * minor cleanup * Add AssemblyName to all model VersionInfo instances. Also fix a couple more tests. * Load and register all assemblies in the Maml directory. Ensure all loaded assemblies are registered in Experiment to maintain compability. Fix tests to not use ComponentCatalog but direct instantiation instead. * Sync up with latest code. * Fix newly added test * Clean up some test changes. * Fix up for latest code * Add path filtering back to LoadAssembliesInDir * Update TestAutoInference to use the correct Environment. * Respond to PR feedback. * Make all AutoInference tests use LocalEnvironment.
…ot frozen (#853) * building transform from ground up * dummy transform works after fixing the getters * SavedModel format works for Train, but fails for Save&Predict * remove dummy transform * remove dummy unit test * Works with non-frozen models * building transform from ground up * dummy transform works after fixing the getters * SavedModel format works for Train, but fails for Save&Predict * remove dummy transform * remove dummy unit test * fix compilation issues; verify existing tests work fine * works locally; need to refactor code * refactored code; keeping only 1 version of the convenience API * added class for directory structure * using latest nuget package (0.0.3) for Microsoft.ML.TensorFlow.TestModels * delete temporary files used when loading/saving models * delete local models; the updated nuget version (0.0.3) for Microsoft.ML.TensorFlow.TestModels contains these models * modified logic for load/restore of models * modified logic for load&restore of unfrozen models * model version update to support non-frozen models * based on the code review comments, we now infer if the provided model is frozen or not * simplify the logic in Save() related to loading of SavedModel. * trying Eric's suggestion * revert back to previous changes * attempt to use stream copy approach instead of in-memory * deleting some commented out code * Ensure we only copy the file segment & cleanup path logic * added finalizer that closes the session (if it isn't closed) and deletes the temporary directory * cleanup + misc review comments * trying to create temp dir with proper ACLs for high priviledge users * create temp dir with proper ACLs for high-privilege processes * fix build after merge with latest master * taking care of review comments related to model versioning of TFTransform * remove IDisposable from the TensorFlowTransform; renaming some methods * refactor code so we have only 1 constructor for the TensorFlowTransform (as suggested in review comment) * fix issues with nuget packaging; refactored the code + added comments * add checks in code to make sure that the input is not a variable length vector * fix typo in name of package * (1) added SavedModel test for MNIST model (2) added try/finally for deleting temp folder (3) deleted test using Legacy Learning API * remove and sort usings in file TrainSaveModelAndPredict.cs * using spaces in nupkgproj * error checking for passed in IHostEnvironment * fix TargetFramework version (netcore 2.0) of DnnAnalyzer to match that of Microsoft.ML.TensorFlow
…o estimators (#957) * derived from trainerestimatorbase * cleaned up * sorting namespaces * fixed review comments, still some more features to add * pr comments on tests * updated the tests * fixed review comments * fixing review comments * fixed bugs * refactored code * fixed review comments and cleaned up * cleaned up code and fixed documentation text * fixed review comments * fixed review comments * fixed review comments * fized review comment * fixed more review comments
adding PcaTrainer as estimator and tidying the PredictionTransformer class.
…on (#1009) * FastTree classification and regression xtensions
…ML.Scoring/Sonoma Library (#942)
Adds a benchmark test to measure performance of doing many single predictions with PredictionEngine. Closes #1013
* Remove the error tracing when assembly loading fails for Maml. Also adding our native assemblies to the list to skip, so they aren't attempted to be loaded. Fix #1034
Helps the user to relate to the macOS version faster.
* Multiclass logistic Regression tests enabled * threshold providing in tests * defining tolerance as a constant in baseTestBaseline Class * upper case camel for constant and _ for large decimal numbers
* Add a workaround for the tests hanging while loading MKL. The workaround is to ensure the MKL library is loaded very early in the test process, so it doesn't cause the deadlock. Workaround #1073 Another deadlock also occurs when running TestAutoInference and TestPipelineSweeper in parallel. Marking these tests to not run in parallel anymore. Workaround #1095 Moving back to the Azure Hosted VS2017 pool to run the tests now that we've narrowed the deadlocks down.
…owUtils.GetModelNodes (#1093)
* Port of time series.
) * Static pipelines now handle types with PipelineColumn properties. * Update the internal infrastructure to accomodate these types, * Update the Roslyn analyzer to accomodate these types. * Update the tests so that they exercise this capability. * Opportunistically fix some problems with the Roslyn analyzer brought up in this work.
* turned string separators into char array separators * fixed review comments * allowed the old api to still work through the arguments object * added command line test * fixed test, and added visibility field to arguments * fixing review comments
…tructors (#1135) * Remove ComponentCatalog from EntryPointGraph's and GraphRunner's constructors * Remove catalog temp variable, use directly in call to ValidateNodes * Remove catalog temp variable from GraphRunner constructor
* add .NET Core 3.0 support for the benchmarks * code review fixes: keep it simple
* Adding the Samples.StaticPipe project. * Adding a sample for SDCA Regression
* Remove explicit ComponentCatalog ValidateNodes and EntryPointNode now use the ComponentCatalog property of IHostEnvironment.
) (#1141) * Updating the CopyColumnsEstimator and Transform to use common code (#706) This builds on the Estimator conversion for the CopyColumnsTransform. This change is mainly refactoring as common code has moved to base level classes. This change is the following: - CopyColumnTransform now derives from OneToOneTransformerBase - CopyColumnEstimator now derives from TrivialEstimator - CopyColumnTransform::Mapper now derives from MapperBase - Removed code that was no longer needed due to these changes * - Moved CopyColumnsTransform into Microsoft.ML.Transforms namespace, updated namespace usage and entrypoints due to this change. - Save now uses the SaveColumns from the base class - Other various changes based upon feedback.
* conversion of multiclass naive bayes classifier to estimator * added pigstension and related test * added public methods to access label and feature histograms in the predictor * fixed review comments on new access functions * moved test to main file
See #1013 for the benchmark results
…#1145) * Fix MatchNumberWithTolerance to better compare floating-point values * Updating CheckEqualityFromPathsCore to allow a tolerance match on Windows
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
We are excited to review your PR.
So we can do the best job, please check:
Fixes #nnnnin your description to cause GitHub to automatically close the issue(s) when your PR is merged.