Merge pull request #205923 from Kmumuda/main

AzureML Python SDK 1.44 Release Note update

Author: v-regandowner

articles/machine-learning/azure-machine-learning-release-notes.md

@@ -9,7 +9,7 @@ ms.custom: event-tier1-build-2022
 ms.topic: reference
 ms.author: larryfr
 author: BlackMist
-ms.date: 06/27/2022
+ms.date: 08/03/2022
 ---
 
 # Azure Machine Learning Python SDK release notes
@@ -19,6 +19,34 @@ In this article, learn about Azure Machine Learning Python SDK releases.  For th
 __RSS feed__: Get notified when this page is updated by copying and pasting the following URL into your feed reader:
 `https://docs.microsoft.com/api/search/rss?search=%22Azure+machine+learning+release+notes%22&locale=en-us`
 
+## 2022-08-01
+
+### Azure Machine Learning SDK for Python v1.44.0  
+  
+  + **azureml-automl-dnn-nlp** 
+    + Weighted accuracy and Matthews correlation coefficient (MCC) will no longer be a metric displayed on calculated metrics for NLP Multilabel classification.
+  + **azureml-automl-dnn-vision**   
+    + Raise user error when invalid annotation format is provided
+  + **azureml-cli-common**
+    + Updated the v1 CLI description
+  + **azureml-contrib-automl-dnn-forecasting**   
+    + Fixed the "Failed to calculate TCN metrics." issues caused for TCNForecaster when different timeseries in the validation dataset have different lengths.
+    + Added auto timeseries ID detection for DNN forecasting models like TCNForecaster.
+    + Fixed a bug with the Forecast TCN model where validation data could be corrupted in some circumstances when the user provided the validation set. 
+  + **azureml-core**
+    + Allow setting a timeout_seconds parameter when downloading artifacts from a Run
+    + Warning message added - Azure ML CLI v1 is getting retired on 30 Sep 2025. Users are recommended to adopt CLI v2.
+    + Fix submission to non-AmlComputes throwing exceptions.
+    + Added docker context support for environments
+  + **azureml-interpret**
+    + Increase numpy version for AutoML packages
+  + **azureml-pipeline-core**
+    + Fix regenerate_outputs=True not taking effect when submit pipeline.
+  + **azureml-train-automl-runtime**
+    + Increase numpy version for AutoML packages
+    + Enable code generation for vision and nlp
+    + Original columns on which grains are created are added as part of predictions.csv
+
 ## 2022-07-21
 
 ### Announcing end of support for Python 3.6 in AzureML SDK v1 packages
@@ -37,7 +65,7 @@ __RSS feed__: Get notified when this page is updated by copying and pasting the
   + **azureml-automl-dnn-nlp**
     + Remove duplicate labels column from multi-label predictions
   + **azureml-contrib-automl-pipeline-steps**
-    + Many Models now provides the capability to generate prediction output in csv format as well. - Many Models prediction will now include column names in the output file in case of **csv** file format.
+    + Many Models now provide the capability to generate prediction output in csv format as well. - Many Models prediction will now include column names in the output file in case of **csv** file format.
   + **azureml-core**
     + ADAL authentication is now deprecated and all authentication classes now use MSAL authentication. Please install azure-cli>=2.30.0 to utilize MSAL based authentication when using AzureCliAuthentication class.
     + Added fix to force environment registration when `Environment.build(workspace)`. The fix solves confusion of the latest environment built instead of the asked one when environment is cloned or inherited from another instance.
@@ -54,10 +82,10 @@ __RSS feed__: Get notified when this page is updated by copying and pasting the
     + Now OutputDatasetConfig is supported as the input of the MM/HTS pipeline builder. The mappings are: 1) OutputTabularDatasetConfig -> treated as unpartitioned tabular dataset. 2) OutputFileDatasetConfig -> treated as filed dataset.
   + **azureml-train-automl-runtime**
     + Added data validation that requires the number of minority class samples in the dataset to be at least as much as the number of CV folds requested.    
-    + Automatic cross-validation parameter configuration is now available for automl forecasting tasks. Users can now specify "auto" for n_cross_validations and cv_step_size or leave them empty, and automl will provide those configurations base on your data. However, currently this feature is not supported when TCN is enabled.
+    + Automatic cross-validation parameter configuration is now available for AutoML forecasting tasks. Users can now specify "auto" for n_cross_validations and cv_step_size or leave them empty, and AutoML will provide those configurations base on your data. However, currently this feature is not supported when TCN is enabled.
     + Forecasting Parameters in Many Models and Hierarchical Time Series can now be passed via object rather than using individual parameters in dictionary.
-    + Enabled forecasting model endpoints with quantiles support to be consumed in PowerBI.
-    + Updated automl scipy dependency upper bound to 1.5.3 from 1.5.2
+    + Enabled forecasting model endpoints with quantiles support to be consumed in Power BI.
+    + Updated AutoML scipy dependency upper bound to 1.5.3 from 1.5.2
 
 ## 2022-04-25
 
@@ -94,7 +122,7 @@ This breaking change comes from the June release of `azureml-inference-server-ht
   + **azureml-core**
     + * Return logs for runs that went through our new runtime when calling any of the get logs function on the run object, including `run.get_details`, `run.get_all_logs`, etc.
     + Added experimental method Datastore.register_onpremises_hdfs to allow users to create datastores pointing to on-premises HDFS resources.
-    + Updating the cli documentation in the help command
+    + Updating the CLI documentation in the help command
   + **azureml-interpret**
     + For azureml-interpret package, remove shap pin with packaging update. Remove numba and numpy pin after CE env update.
   + **azureml-mlflow**
@@ -118,7 +146,7 @@ This breaking change comes from the June release of `azureml-inference-server-ht
     +  Adding min-label-classes check for both classification tasks (multi-class and multi-label). It will throw an error for the customer's run if the unique number of classes in the input training dataset is fewer than 2. It is meaningless to run classification on fewer than two classes.
   + **azureml-automl-runtime**
     +  Converting decimal type y-test into float to allow for metrics computation to proceed without errors.   
-    +  Automl training now supports numpy version 1.8.    
+    +  AutoML training now supports numpy version 1.8.    
   + **azureml-contrib-automl-dnn-forecasting**
     +  Fixed a bug in the TCNForecaster model where not all training data would be used when cross-validation settings were provided.
     +  TCNForecaster wrapper's forecast method that was corrupting inference-time predictions. Also fixed an issue where the forecast method would not use the most recent context data in train-valid scenarios.
@@ -130,9 +158,9 @@ This breaking change comes from the June release of `azureml-inference-server-ht
     +  Fix the issue that magic widget is disappeared.
   + **azureml-train-automl-runtime**
     +  Updating AutoML dependencies to support Python 3.8. This change will break compatibility with models trained with SDK 1.37 or below due to newer Pandas interfaces being saved in the model.
-    +  Automl training now supports numpy version 1.19
-    +  Fix automl reset index logic for ensemble models in automl_setup_model_explanations API
-    +  In automl, use lightgbm surrogate model instead of linear surrogate model for sparse case after latest lightgbm version upgrade
+    +  AutoML training now supports numpy version 1.19
+    +  Fix AutoML reset index logic for ensemble models in automl_setup_model_explanations API
+    +  In AutoML, use lightgbm surrogate model instead of linear surrogate model for sparse case after latest lightgbm version upgrade
     +  All internal intermediate artifacts that are produced by AutoML are now stored transparently on the parent run (instead of being sent to the default workspace blob store). Users should be able to see the artifacts that AutoML generates under the 'outputs/` directory on the parent run.
 
 
@@ -555,7 +583,7 @@ The `ml` extension to the Azure CLI is the next-generation interface for Azure M
 
 
 ## 2021-03-31
-### Azure Machine Learning Studio Notebooks Experience (March Update)
+### Azure Machine Learning studio Notebooks Experience (March Update)
 + **New features**
   + Render CSV/TSV. Users will be able to render and TSV/CSV file in a grid format for easier data analysis. 
   + SSO Authentication for Compute Instance. Users can now easily authenticate any new compute instances directly in the Notebook UI, making it easier to authenticate and use Azure SDKs directly in AzureML. 
@@ -600,7 +628,7 @@ The `ml` extension to the Azure CLI is the next-generation interface for Azure M
     +  Fixed show_output=False to return control to the user when running using spark.
 
 ## 2021-02-28
-### Azure Machine Learning Studio Notebooks Experience (February Update)
+### Azure Machine Learning studio Notebooks Experience (February Update)
 + **New features**
   + [Native Terminal (GA)](./how-to-access-terminal.md). Users will now have access to an integrated terminal as well as Git operation via the integrated terminal.
   + Notebook Snippets (preview). Common Azure ML code excerpts are now available at your fingertips. Navigate to the code snippets panel, accessible via the toolbar, or activate the in-code snippets menu using Ctrl + Space.  
@@ -623,7 +651,7 @@ The `ml` extension to the Azure CLI is the next-generation interface for Azure M
 ### Azure Machine Learning SDK for Python v1.23.0
 + **Bug fixes and improvements**
   + **azureml-core**
-    + [Experimental feature] Add support to link synapse workspace into AML as an linked service
+    + [Experimental feature] Add support to link synapse workspace into AML as a linked service
     + [Experimental feature] Add support to attach synapse spark pool into AML as a compute
     + [Experimental feature] Add support for identity based data access. Users can register datastore or datasets without providing credentials. In such case, users' Azure AD token or managed identity of compute target will be used for authentication. To learn more, see [Connect to storage by using identity-based data access](./how-to-identity-based-data-access.md).
   + **azureml-pipeline-steps**
@@ -678,7 +706,7 @@ The `ml` extension to the Azure CLI is the next-generation interface for Azure M
 
 
  ## 2021-01-31
-### Azure Machine Learning Studio Notebooks Experience (January Update)
+### Azure Machine Learning studio Notebooks Experience (January Update)
 + **New features**
   + Native Markdown Editor in AzureML. Users can now render and edit markdown files natively in AzureML Studio.
   + [Run Button for Scripts (.py, .R and .sh)](./how-to-run-jupyter-notebooks.md#run-a-notebook-or-python-script). Users can easily now run Python, R and Bash script in AzureML
@@ -742,7 +770,7 @@ The `ml` extension to the Azure CLI is the next-generation interface for Azure M
 
 
 ## 2020-12-31
-### Azure Machine Learning Studio Notebooks Experience (December Update)
+### Azure Machine Learning studio Notebooks Experience (December Update)
 + **New features**
   + User Filename search. Users are now able to search all the files saved in a workspace.
   + Markdown Side by Side support per Notebook Cell. In a notebook cell, users can now have the option to view rendered markdown and markdown syntax side-by-side.
@@ -818,7 +846,7 @@ The `ml` extension to the Azure CLI is the next-generation interface for Azure M
     + Deprecated the use of Mpi as a valid type of input for Estimator classes in favor of using MpiConfiguration with ScriptRunConfig.
 
 ## 2020-11-30
-### Azure Machine Learning Studio Notebooks Experience (November Update)
+### Azure Machine Learning studio Notebooks Experience (November Update)
 + **New features**
    + Native Terminal. Users will now have access to an integrated terminal as well as Git operation via the [integrated terminal.](./how-to-access-terminal.md)
   + Duplicate Folder 
@@ -850,7 +878,7 @@ The `ml` extension to the Azure CLI is the next-generation interface for Azure M
     + Improved error message to include potential fixes when a dataset is incorrectly passed to an experiment (e.g. ScriptRunConfig). 
     + Improved documentation for `OutputDatasetConfig.register_on_complete` to include the behavior of what will happen when the name already exists.
     + Specifying dataset input and output names that have the potential to collide with common environment variables will now result in a warning
-    + Repurposed `grant_workspace_access` parameter when registering datastores. Set it to `True` to access data behind virtual network from Machine Learning Studio.
+    + Repurposed `grant_workspace_access` parameter when registering datastores. Set it to `True` to access data behind virtual network from Machine Learning studio.
       [Learn more](./how-to-enable-studio-virtual-network.md)
     + Linked service API is refined. Instead of providing resource ID, we have 3 separate parameters sub_id, rg, and name defined in configuration.
     + In order to enable customers to self-resolve token corruption issues, enable workspace token synchronization to be a public method.
@@ -908,7 +936,7 @@ Learn more about [image instance segmentation labeling](how-to-label-data.md).
   + **azureml-train-automl-client**
     + Fixed an issue where get_output may raise an XGBoostError.
 
-### Azure Machine Learning Studio Notebooks Experience (October Update)
+### Azure Machine Learning studio Notebooks Experience (October Update)
 + **New features**
   + [Full virtual network support](./how-to-enable-studio-virtual-network.md)
   + [Focus Mode](./how-to-run-jupyter-notebooks.md#focus-mode)
@@ -931,14 +959,14 @@ Learn more about [image instance segmentation labeling](how-to-label-data.md).
     + Updated run.log_table to allow individual rows to be logged.
     + Added static method `Run.get(workspace, run_id)` to retrieve a run only using a workspace 
     + Added instance method `Workspace.get_run(run_id)` to retrieve a run within the workspace
-    + Introducing command property in run configuration which will enables users to submit command instead of script & arguments.
+    + Introducing command property in run configuration which will enable users to submit command instead of script & arguments.
   + **azureml-interpret**
     + fixed explanation client is_raw flag behavior in azureml-interpret
   + **azureml-sdk**
     + `azureml-sdk` officially support Python 3.8.
   + **azureml-train-core**
     + Adding TensorFlow 2.3 curated environment
-    + Introducing command property in run configuration which will enables users to submit command instead of script & arguments.
+    + Introducing command property in run configuration which will enable users to submit command instead of script & arguments.
   + **azureml-widgets**
     + Redesigned interface for script run widget.
 
@@ -1072,7 +1100,7 @@ Learn more about [image instance segmentation labeling](how-to-label-data.md).
     + Reverting PyTorch Default Version to 1.4.
     + Adding PyTorch 1.6 & TensorFlow 2.2 images and curated environment.
 
-### Azure Machine Learning Studio Notebooks Experience (August Update)
+### Azure Machine Learning studio Notebooks Experience (August Update)
 + **New features**
   + New Getting started landing Page 
 
@@ -1340,7 +1368,7 @@ Learn more about [image instance segmentation labeling](how-to-label-data.md).
     + Fix the run failures happening when the lookback features are enabled and the data contain short grains.
     + Fixed the issue with duplicated time index error message when lags or rolling windows were set to 'auto'.
     + Fixed the issue with Prophet and Arima models on data sets, containing the lookback features.
-    + Added support of dates before 1677-09-21 or after 2262-04-11 in columns other then date time in the forecasting tasks. Improved error messages.
+    + Added support of dates before 1677-09-21 or after 2262-04-11 in columns other than date time in the forecasting tasks. Improved error messages.
     + The forecasting parameters documentation was improved. The lag_length parameter was deprecated.
     + Better exception message on featurization step fit_transform() due to custom transformer parameters.
     + Add support for multiple languages for deep learning transformer models such as BERT in automated ML.