We are happy to announce the availability of MLflow 1.21.0!
MLflow 1.21.0 includes several major features and improvements:
Features:
- [UI] Add a diff-only toggle to the runs table for filtering out columns with constant values (#4862, @marijncv)
- [UI] Add a duration column to the runs table (#4840, @marijncv)
- [UI] Display the default column sorting order in the runs table (#4847, @marijncv)
- [UI] Add start_time and duration information to exported runs CSV (#4851, @marijncv)
- [UI] Add lifecycle stage information to the run page (#4848, @marijncv)
- [UI] Collapse run page sections by default for space efficiency, limit artifact previews to 50MB (#4917, @dbczumar)
- [Tracking] Introduce autologging capabilities for PaddlePaddle model training (#4751, @jinminhao)
- [Tracking] Add an optional tags field to the CreateExperiment API (#4788, @dbczumar; #4795, @apurva-koti)
- [Tracking] Add support for deleting artifacts from SFTP stores via the mlflow gc CLI (#4670, @afaul)
- [Tracking] Support AzureDefaultCredential for authenticating with Azure artifact storage backends (#4002, @marijncv)
- [Models] Upgrade the fastai model flavor to support fastai V2 (>=2.4.1) (#4715, @jinzhang21)
- [Models] Introduce an mlflow.prophet model flavor for Prophet time series models (#4773, @BenWilson2)
- [Models] Introduce a CLI for publishing MLflow Models to the SageMaker Model Registry (#4669, @jinnig)
- [Models] Print a warning when inferred model dependencies are not available on PyPI (#4891, @dbczumar)
- [Models, Projects] Add MLFLOW_CONDA_CREATE_ENV_CMD for customizing Conda environment creation (#4746, @giacomov)
Bug fixes and documentation updates:
- [UI] Fix an issue where column selections made in the runs table were persisted across experiments (#4926, @sunishsheth2009)
- [UI] Fix an issue where the text null was displayed in the runs table column ordering dropdown (#4924, @harupy)
- [UI] Fix a bug causing the metric plot view to display NaN values upon click (#4858, @arpitjasa-db)
- [Tracking] Fix a model load failure for paths containing spaces or special characters on UNIX systems (#4890, @BenWilson2)
- [Tracking] Correct a migration issue that impacted usage of MLflow Tracking with SQL Server (#4880, @marijncv)
- [Tracking] Spark datasource autologging tags now respect the maximum allowable size for MLflow Tracking (#4809, @dbczumar)
- [Model Registry] Add previously-missing certificate sources for Model Registry REST API requests (#4731, @ericgosno91)
- [Model Registry] Throw an exception when users supply invalid Model Registry URIs for Databricks (#4877, @yunpark93)
- [Scoring] Fix a schema enforcement error that incorrectly cast date-like strings to datetime objects (#4902, @wentinghu)
- [Docs] Expand the documentation for the MLflow Skinny Client (#4113, @eedeleon)
For a comprehensive list of changes, see the release change log, and check out the latest documentation on mlflow.org.
We are happy to announce the availability of MLflow 1.20.2!
MLflow 1.20.2 is a patch release containing the following features and bug fixes:
Features:
- Enabled auto dependency inference in spark flavor in autologging (#4759, @harupy)
Bug fixes and documentation updates:
For a comprehensive list of changes, see the release change log, and check out the latest documentation on mlflow.org.
We are happy to announce the availability of MLflow 1.20.0!
Note: The MLflow R package for 1.20.0 is not yet available but will be in a week because CRAN's submission system will be offline until September 1st.
In addition to bug and documentation fixes, MLflow 1.20.0 includes the following features and improvements:
- Autologging for scikit-learn now records post training metrics when scikit-learn evaluation APIs, such as
sklearn.metrics.mean_squared_error
, are called (#4491, #4628 #4638, @WeichenXu123)
- Autologging for PySpark ML now records post training metrics when model evaluation APIs, such as
Evaluator.evaluate()
, are called (#4686, @WeichenXu123)
- Add
pip_requirements
and extra_pip_requirements
to mlflow.*.log_model
and mlflow.*.save_model
for directly specifying the pip requirements of the model to log / save (#4519, #4577, #4602, @harupy)
- Added
stdMetrics
entries to the training metrics recorded during PySpark CrossValidator autologging (#4672, @WeichenXu123)
- MLflow UI updates:
- Improved scalability of the parallel coordinates plot for run performance comparison,
- Added support for filtering runs based on their start time on the experiment page,
- Added a dropdown for runs table column sorting on the experiment page,
- Upgraded the AG Grid plugin, which is used for runs table loading on the experiment page, to version 25.0.0,
- Fixed a bug on the experiment page that caused the metrics section of the runs table to collapse when selecting columns from other table sections (#4712, @dbczumar)
- Added support for distributed execution to autologging for PyTorch Lightning (#4717, @dbczumar)
- Expanded R support for Model Registry functionality (#4527, @bramrodenburg)
- Added model scoring server support for defining custom prediction response wrappers (#4611, @Ark-kun)
mlflow.*.log_model
and mlflow.*.save_model
now automatically infer the pip requirements of the model to log / save based on the current software environment (#4518, @harupy)
- Introduced support for running Sagemaker Batch Transform jobs with MLflow Models (#4410, #4589, @YQ-Wang)
For a comprehensive list of changes, see the release change log, and check out the latest documentation on mlflow.org.
We are happy to announce the availability of MLflow 1.20.1!
MLflow 1.20.1 is a patch release containing the following bug fixes:
- Avoid calling
importlib_metadata.packages_distributions
upon mlflow.utils.requirements_utils
import (#4741, @dbczumar)
- Avoid depending on
importlib_metadata==4.7.0
(#4740, @dbczumar)
For a comprehensive list of changes, see the release change log, and check out the latest documentation on mlflow.org.
We are happy to announce the availability of MLflow 1.19.0!
In addition to bug and documentation fixes, MLflow 1.19.0 includes the following features and improvements:
- Add support for plotting per-class feature importance computed on linear boosters in XGBoost autologging (#4523, @dbczumar)
- Add
mlflow_create_registered_model
and mlflow_delete_registered_model
for R to create/delete registered models.
- Add support for setting tags while resuming a run (#4497, @dbczumar)
- MLflow UI updates (#4490, @sunishsheth2009)
- Add framework for internationalization support.
- Move metric columns before parameter and tag columns in the runs table.
- Change the display format of run start time to elapsed time (e.g. 3 minutes ago) from timestamp (e.g. 2021-07-14 14:02:10) in the runs table.
For a comprehensive list of changes, see the release change log, and check out the latest documentation on mlflow.org.
We are happy to announce the availability of MLflow 1.18.0!
In addition to bug and documentation fixes, MLflow 1.18.0 includes the following features and improvements:
- Autologging performance improvements for XGBoost, LightGBM, and scikit-learn (#4416, #4473, @dbczumar)
- Add new PaddlePaddle flavor to MLflow Models (#4406, #4439, @jinminhao)
- Introduce paginated ListExperiments API (#3881, @wamartin-aml)
- Include Runtime version for MLflow Models logged on Databricks (#4421, @stevenchen-db)
- MLflow Models now log dependencies in pip requirements.txt format, in addition to existing conda format (#4409, #4422, @stevenchen-db)
- Add support for limiting the number child runs created by autologging for scikit-learn hyperparameter search models (#4382, @mohamad-arabi)
- Improve artifact upload / download performance on Databricks (#4260, @dbczumar)
- Migrate all model dependencies from conda to "pip" section (#4393, @WeichenXu123)
For a comprehensive list of changes, see the release change log, and check out the latest documentation on mlflow.org.
We are happy to announce the availability of MLflow 1.17.0!
In addition to bug and documentation fixes, MLflow 1.17.0 includes the following features and improvements:
- Add support for hyperparameter-tuning models to
mlflow.pyspark.ml.autolog()
(#4270, @WeichenXu123)
For a comprehensive list of changes, see the release change log, and check out the latest documentation on mlflow.org.
We are happy to announce the availability of MLflow 1.16.0!
In addition to bug and documentation fixes, MLflow 1.16.0 includes the following features and improvements:
- Add
mlflow.pyspark.ml.autolog()
API for autologging of pyspark.ml
estimators (#4228, @WeichenXu123)
- Add
mlflow.catboost.log_model
, mlflow.catboost.save_model
, mlflow.catboost.load_model
APIs for CatBoost model persistence (#2417, @harupy)
- Enable
mlflow.pyfunc.spark_udf
to use column names from model signature by default (#4236, @Loquats)
- Add
datetime
data type for model signatures (#4241, @vperiyasamy)
- Add
mlflow.sklearn.eval_and_log_metrics
API that computes and logs metrics for the given scikit-learn model and labeled dataset. (#4218, @alkispoly-db)
For a comprehensive list of changes, see the release change log, and check out the latest documentation on mlflow.org.
We are happy to announce the availability of MLflow 1.15.0!
In addition to bug and documentation fixes, MLflow 1.15.0 includes the following features and improvements:
- Add
silent=False
option to all autologging APIs, to allow suppressing MLflow warnings and logging statements during autologging setup and training (#4173, @dbczumar)
- Add
disable_for_unsupported_versions=False
option to all autologging APIs, to disable autologging for versions of ML frameworks that have not been explicitly tested against the current version of the MLflow client (#4119, @WeichenXu123)
For a comprehensive list of changes, see the release change log, and check out the latest documentation on mlflow.org.
We are happy to announce the availability of MLflow 1.14.1!
MLflow 1.14.1 is a patch release containing the following bug fix:
- Fix issues in handling flexible numpy datatypes in TensorSpec (#4147, @arjundc-db)