Merge pull request #231258 from lgayhardt/amlacpt0323

Azure Container for PyTorch

Author: ttorble

articles/machine-learning/how-to-azure-container-for-pytorch-environment.md

@@ -0,0 +1,72 @@
+---
+title: How to create Azure Container for PyTorch Custom Curated environment
+titleSuffix: Azure Machine Learning
+description: Create custom curated Azure Container for PyTorch environments in Azure Machine Learning studio to run your machine learning models and reuse it in different scenarios.
+services: machine-learning
+author: sheetalarkadam
+ms.author: parinitarahi
+ms.reviewer: ssalgado
+ms.service: machine-learning
+ms.subservice: core
+ms.topic: how-to
+ms.date: 03/20/2023
+---
+
+# Create custom curated Azure Container for PyTorch (ACPT) environments in Azure Machine Learning studio
+
+If you're looking to extend curated environment and add Hugging Face (HF) transformers or datasets or any other external packages to be installed, Azure Machine Learning offers to create a new env with docker context containing ACPT curated environment as base image and additional packages on top of it as below.
+
+## Prerequisites
+
+Before following the steps in this article, make sure you have the following prerequisites:
+
+- An Azure subscription. If you don't have an Azure subscription, create a free account before you begin. Try the [free or paid version of Azure Machine Learning](https://azure.microsoft.com/free/).
+
+- An Azure Machine Learning workspace. If you don't have one, use the steps in the [Quickstart: Create workspace resources](quickstart-create-resources.md) article to create one.
+
+## Navigate to environments
+
+In the [Azure Machine Learning studio](https://ml.azure.com/registries/environments), navigate to the "Environments" section by selecting the "Environments" option.
+
+:::image type="content" source="./media/how-to-azure-container-for-pytorch-environment/navigate-to-environments.png" alt-text="Screenshot of navigating to environments from Azure Machine Learning studio." lightbox= "./media/how-to-azure-container-for-pytorch-environment/navigate-to-environments.png":::
+
+## Navigate to curated environments
+
+Navigate to curated environments and search "acpt" to list all the available ACPT curated environments. Selecting the environment shows details of the environment.
+
+:::image type="content" source="./media/how-to-azure-container-for-pytorch-environment/navigate-to-curated-environments.png" alt-text="Screenshot of navigating to curated environments." lightbox= "./media/how-to-azure-container-for-pytorch-environment/navigate-to-curated-environments.png":::
+
+
+## Get details of the curated environments
+
+To create custom environment, you need the base docker image repository, which can be found in the "Description" section as "Azure Container Registry". Copy the "Azure Container Registry" name, which is used later when you create a new custom environment.
+
+:::image type="content" source="./media/how-to-azure-container-for-pytorch-environment/get-details-curated-environments.png" alt-text="Screenshot of getting container registry name." lightbox= "./media/how-to-azure-container-for-pytorch-environment/get-details-curated-environments.png":::
+
+## Navigate to custom environments
+
+Go back and select the " Custom Environments" tab.
+
+:::image type="content" source="./media/how-to-azure-container-for-pytorch-environment/navigate-to-custom-environment.png" alt-text="Screenshot of navigating to custom environments." lightbox= "./media/how-to-azure-container-for-pytorch-environment/navigate-to-custom-environment.png":::
+
+## Create custom environments
+
+Select **+ Create**. In the "Create Environment" window, name the environment, description and select "Create a new docker context" in Select environments type section.
+
+:::image type="content" source="./media/how-to-azure-container-for-pytorch-environment/create-environment-window.png" alt-text="Screenshot of creating custom environment." lightbox= "./media/how-to-azure-container-for-pytorch-environment/create-environment-window.png":::
+
+Paste the docker image name that you copied in previously. Configure your environment by declaring the base image and add any env variables you want to use and the packages that you want to include.
+
+:::image type="content" source="./media/how-to-azure-container-for-pytorch-environment/configure-environment.png" alt-text="Screenshot of configuring the environment with name, packages with docker context." lightbox= "./media/how-to-azure-container-for-pytorch-environment/configure-environment.png":::
+
+Review your environment settings, add any tags if needed and select on the **Create** button to create your custom environment.
+
+That's it! You've now created a custom environment in Azure Machine Learning studio and can use it to run your machine learning models.
+
+## Next steps
+
+- Learn more about environment objects:
+    - [What are Azure Machine Learning environments? ](concept-environments.md).
+    -  Learn more about [curated environments](concept-environments.md).
+- Learn more about [training models in Azure Machine Learning](concept-train-machine-learning-model.md).
+- [Azure Container for PyTorch (ACPT) reference](resource-azure-container-for-pytorch.md)

articles/machine-learning/how-to-train-distributed-gpu.md

@@ -121,14 +121,14 @@ Azure Machine Learning will set the `MASTER_ADDR`, `MASTER_PORT`, `WORLD_SIZE`,
 
 ## DeepSpeed
 
-[DeepSpeed](https://www.deepspeed.ai/tutorials/azure/) is supported as a first-class citizen within Azure Machine Learning to run distributed jobs with near linear scalabibility in terms of 
+[DeepSpeed](https://www.deepspeed.ai/tutorials/azure/) is supported as a first-class citizen within Azure Machine Learning to run distributed jobs with near linear scalability in terms of 
 
 * Increase in model size
 * Increase in number of GPUs
 
 `DeepSpeed` can be enabled using either Pytorch distribution or MPI for running distributed training. Azure Machine Learning supports the `DeepSpeed` launcher to launch distributed training as well as autotuning to get optimal `ds` configuration.
 
-You can use a [curated environment](resource-curated-environments.md#azure-container-for-pytorch-acpt-preview) for an out of the box environment with the latest state of art technologies including `DeepSpeed`, `ORT`, `MSSCCL`, and `Pytorch` for your DeepSpeed training jobs.
+You can use a [curated environment](resource-curated-environments.md#azure-container-for-pytorch-acpt) for an out of the box environment with the latest state of art technologies including `DeepSpeed`, `ORT`, `MSSCCL`, and `Pytorch` for your DeepSpeed training jobs.
 
 ### DeepSpeed example
 

articles/machine-learning/media/how-to-azure-container-for-pytorch-environment/configure-environment.png

@@ -78,7 +78,7 @@ Nebula can
 * An Azure subscription and an Azure Machine Learning workspace. See [Create workspace resources](./quickstart-create-resources.md) for more information about workspace resource creation
 * An Azure Machine Learning compute target. See [Manage training & deploy computes](./how-to-create-attach-compute-studio.md) to learn more about compute target creation
 * A training script that uses **PyTorch**.
-* ACPT-curated (Azure Container for Pytorch) environment. See [Curated environments](resource-curated-environments.md#azure-container-for-pytorch-acpt-preview) to obtain the ACPT image. Learn how to use the curated environment [here](./how-to-use-environments.md)
+* ACPT-curated (Azure Container for Pytorch) environment. See [Curated environments](resource-curated-environments.md#azure-container-for-pytorch-acpt) to obtain the ACPT image. Learn how to [use the curated environment](./how-to-use-environments.md)
 * An Azure Machine Learning script run configuration file. If you don’t have one, you can follow [this resource](./how-to-set-up-training-targets.md)
 
 ## How to Use Nebula
@@ -90,7 +90,7 @@ Nebula use involves:
 - [API calls to save and load checkpoints](#call-apis-to-save-and-load-checkpoints)
 
 ### Using ACPT environment
-[Azure Container for PyTorch (ACPT)](how-to-manage-environments-v2.md?tabs=cli#curated-environments), a curated environment for PyTorch model training, includes Nebula as a preinstalled, dependent Python package. See [Azure Container for PyTorch (ACPT)](resource-curated-environments.md#azure-container-for-pytorch-acpt-preview) to view the curated environment, and [Enabling Deep Learning with Azure Container for PyTorch in Azure Machine Learning](https://techcommunity.microsoft.com/t5/ai-machine-learning-blog/enabling-deep-learning-with-azure-container-for-pytorch-in-azure/ba-p/3650489) to learn more about the ACPT image.
+[Azure Container for PyTorch (ACPT)](how-to-manage-environments-v2.md?tabs=cli#curated-environments), a curated environment for PyTorch model training, includes Nebula as a preinstalled, dependent Python package. See [Azure Container for PyTorch (ACPT)](resource-curated-environments.md#azure-container-for-pytorch-acpt) to view the curated environment, and [Enabling Deep Learning with Azure Container for PyTorch in Azure Machine Learning](https://techcommunity.microsoft.com/t5/ai-machine-learning-blog/enabling-deep-learning-with-azure-container-for-pytorch-in-azure/ba-p/3650489) to learn more about the ACPT image.
 
 ### Initializing Nebula