From 8602d75c1b8c2c4fea96f6f506f58458315f840f Mon Sep 17 00:00:00 2001 From: Sheri Gilley Date: Fri, 27 Sep 2024 13:51:39 -0500 Subject: [PATCH 1/3] Update toc.yml - fix indent --- articles/ai-studio/toc.yml | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/articles/ai-studio/toc.yml b/articles/ai-studio/toc.yml index 5d15e65834..4b4b35342e 100644 --- a/articles/ai-studio/toc.yml +++ b/articles/ai-studio/toc.yml @@ -65,7 +65,7 @@ items: - name: Create a project href: how-to/create-projects.md - name: Create and manage compute - href: how-to/create-manage-compute.md + href: how-to/create-manage-compute.md - name: Connect to services and resources items: - name: Connections overview From 68e076ca523ab38d7d5bcd0aeb34329cd71728ef Mon Sep 17 00:00:00 2001 From: Aaron Hill Date: Fri, 27 Sep 2024 15:05:06 -0500 Subject: [PATCH 2/3] updating retirement --- articles/ai-services/openai/concepts/model-retirements.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/articles/ai-services/openai/concepts/model-retirements.md b/articles/ai-services/openai/concepts/model-retirements.md index 42749e4e88..f5f63a6fd9 100644 --- a/articles/ai-services/openai/concepts/model-retirements.md +++ b/articles/ai-services/openai/concepts/model-retirements.md @@ -29,7 +29,7 @@ Azure OpenAI Service models are continually refreshed with newer and more capabl Azure OpenAI notifies customers of active Azure OpenAI Service deployments for models with upcoming retirements. We notify customers of upcoming retirements as follows for each deployment: -1. At model launch, we programmatically designate a "not sooner than" retirement date (typically six months to one year out). +1. At model launch, we programmatically designate a "not sooner than" retirement date (typically one year out). 2. At least 60 days notice before model retirement for Generally Available (GA) models. 3. At least 30 days notice before preview model version upgrades. From ef6a6b11d2e0ba40b9de21318c95e8d8a0123ea2 Mon Sep 17 00:00:00 2001 From: Frank Solomon Date: Fri, 27 Sep 2024 13:52:53 -0700 Subject: [PATCH 3/3] Freshness update for concept-what-is-managed-feature-store.md . . . --- .../concept-what-is-managed-feature-store.md | 26 +++++++++---------- 1 file changed, 13 insertions(+), 13 deletions(-) diff --git a/articles/machine-learning/concept-what-is-managed-feature-store.md b/articles/machine-learning/concept-what-is-managed-feature-store.md index 838a02de53..8c4474e88b 100644 --- a/articles/machine-learning/concept-what-is-managed-feature-store.md +++ b/articles/machine-learning/concept-what-is-managed-feature-store.md @@ -9,26 +9,26 @@ ms.service: azure-machine-learning ms.subservice: mldata ms.custom: build-2023, ignite-2023 ms.topic: conceptual -ms.date: 10/20/2023 +ms.date: 09/27/2024 --- # What is managed feature store? -In our vision for managed feature store, we want to empower machine learning professionals to independently develop and productionize features. You provide a feature set specification, and then let the system handle serving, securing, and monitoring of the features. This frees you from the overhead of underlying feature engineering pipeline set-up and management. +For managed feature store, we want machine learning professionals like you to independently develop and productionize features. You provide a feature set specification. The system handles serving, securing, and monitoring the features. This frees you from the underlying feature engineering pipeline set-up and management overhead. -Thanks to integration of our feature store across the machine learning life cycle, you can experiment and ship models faster, increase the reliability of their models, and reduce your operational costs. The redefinition of the machine learning experience provides these advantages. +Thanks to integration of our feature store across the machine learning life cycle, you can experiment and ship models faster, increase the reliability of your models, and reduce your operational costs. The redefinition of the machine learning experience provides these advantages. -For more information on top level entities in feature store, including feature set specifications, see [Understanding top-level entities in managed feature store](concept-top-level-entities-in-managed-feature-store.md). +For more information about top level entities in feature store, including feature set specifications, visit [Understanding top-level entities in managed feature store](concept-top-level-entities-in-managed-feature-store.md). ## What are features? -Features serve as the input data for your model. For data-driven use cases in an enterprise context, features often transform historical data (simple aggregates, window aggregates, row level transforms, etc.). For example, consider a customer churn machine learning model. The model inputs could include customer interaction data like `7day_transactions_sum` (number of transactions in the past seven days) or `7day_complaints_sum` (number of complaints in the past seven days). Both of these aggregate functions are computed on the previous seven-day data. +A feature serves as the input data for your model. For data-driven use cases in an enterprise context, a feature often transforms historical data (simple aggregates, window aggregates, row level transforms, etc.). For example, consider a customer churn machine learning model. The model inputs could include customer interaction data - for example, `7day_transactions_sum` (number of transactions in the past seven days) or `7day_complaints_sum` (number of complaints in the past seven days). Both of these aggregate functions are computed on the previous seven days of data. ## Problems solved by feature store -To better understand managed feature store, you should first understand the problems that feature store can solve. +To better understand managed feature store, you should first understand the problems that a feature store can solve. -- Feature store allows you to **search and reuse features created by your team, to avoid redundant work and deliver consistent predictions**. +- A feature store allows you to **search and reuse features that your team creates, to avoid redundant work and deliver consistent predictions**. -- You can create **new features with the ability for transformations**, to address feature engineering requirements in an agile, dynamic way. +- You can create **a new feature with the ability for transformations**, to address feature engineering requirements in an agile, dynamic way. - The system **operationalizes and manages the feature engineering pipelines required for transformation and materialization** to free your team from the operational aspects. @@ -36,7 +36,7 @@ To better understand managed feature store, you should first understand the prob ## Share managed feature store -:::image type="content" source="./media/concept-what-is-managed-feature-store/share-feature-store.png" alt-text="Diagram that shows how you can share a feature store among multiple users and workspaces"::: +:::image type="content" source="./media/concept-what-is-managed-feature-store/share-feature-store.png" alt-text="Diagram showing how to share feature store among multiple users and workspaces"::: Feature store is a new type of workspace that multiple project workspaces can use. You can consume features from Spark-based environments other than Azure Machine Learning, such as Azure Databricks. You can also perform local development and testing of features. @@ -78,14 +78,14 @@ Feature transformation involves dataset feature modification, to improve model p Managed feature store provides these feature transformation capabilities: -- **Support for custom transformations** - You can write a Spark transformer to develop features with custom transformations, like window-based aggregates, for example +- **Support for custom transformations** - You can write a Spark transformer to develop features with custom transformations - for example, window-based aggregates - **Support for precomputed features** - You can bring precomputed features into feature store, and serve them without writing code - **Local development and testing** - With a Spark environment, you can fully develop and test feature sets locally ### Feature materialization Materialization involves the computation of feature values for a given feature window, and persistence of those values in a materialization store. Now, feature data can be retrieved more quickly and reliably for training and inference purposes. -- **Managed feature materialization pipeline** - You declaratively specify the materialization schedule, and the system then handles the scheduling, precomputation, and materialization of the values into the materialization store. +- **Managed feature materialization pipeline** - You declaratively specify the materialization schedule, and the system then handles the scheduling, precomputation, and materialization of the values into the materialization store - **Backfill support** - You can perform on-demand materialization of feature sets for a given feature window - **Managed Spark support for materialization** - Azure Machine Learning managed Spark (in serverless compute instances) runs the materialization jobs. It frees you from set-up and management of the Spark infrastructure. @@ -106,12 +106,12 @@ Managed feature store provides these feature retrieval capabilities: Managed feature store provides the following monitoring capabilities: -- **Status of materialization jobs** - You can view status of materialization jobs using the UI, CLI or SDK +- **Status of materialization jobs** - You can view the status of materialization jobs using the UI, CLI, or SDK - **Notification on materialization jobs** - You can set up email notifications on the different statuses of the materialization jobs ### Security -Managed feature store provides the following security capabilities: +Managed feature store provides these security capabilities: - **RBAC** - Role based access control for feature store, feature set and entities. - **Query across feature stores** - You can create multiple feature stores with different access permissions for users, but allow querying (for example, generate training data) from across multiple feature stores