Azure Container for PyTorch (ACPT)
Azure Container for PyTorch is a lightweight, standalone environment that includes needed components to effectively run optimized training for large models on Azure Machine Learning. The Azure Machine Learning curated environments are available in the user’s workspace by default and are backed by cached Docker images that use the latest version of the Azure Machine Learning SDK. It helps with reducing preparation costs and faster deployment time. ACPT can be used to quickly get started with various deep learning tasks with PyTorch on Azure.
Note
Use the Python SDK, CLI, or Azure Machine Learning studio to get the full list of environments and their dependencies. For more information, see the environments article.
Why should I use ACPT?
- Flexibility: Use as-is with preinstalled packages or build on top of the curated environment.
- Ease of use: All components are installed and validated against dozens of Microsoft workloads to reduce setup costs and accelerate time to value.
- Efficiency: Avoid unnecessary image builds and only have required dependencies that are accessible right in the image/container.
- Optimized training framework: Set up, develop, and accelerate PyTorch models on large workloads, and improve training and deployment success rate.
- Up-to-date stack: Access the latest compatible versions of Ubuntu, Python, PyTorch, CUDA/RocM, etc.
- Latest training optimization technologies: Make use of ONNX Runtime , DeepSpeed, MSCCL, and more.
- Integration with Azure Machine Learning: Track your PyTorch experiments on Azure Machine Learning studio or using the SDK. Azure customer support also reduces training and deployment latency.
- Availability as DSVM: The image is also available as a Data Science Virtual Machine (DSVM). To learn more about Data Science Virtual Machines, see the DSVM overview documentation.
Important
To view more information about curated environment packages and versions, visit the Environments tab in the Azure Machine Learning studio.
Supported configurations for Azure Container for PyTorch (ACPT)
Description: The Azure Curated Environment for PyTorch is our latest PyTorch curated environment. It's optimized for large, distributed deep learning workloads and comes prepackaged with the best of Microsoft technologies for accelerated training (e.g., Onnx Runtime Training (ORT), DeepSpeed, MSCCL, etc.).
The following configurations are supported:
| Environment Name | OS | GPU Version | Python Version | PyTorch Version | ORT-training Version | DeepSpeed Version | torch-ort Version | Nebula Version |
|---|---|---|---|---|---|---|---|---|
| acpt-pytorch-2.2-cuda12.1 | Ubuntu 20.04 | cu121 | 3.10 | 2.2.2 | 1.17.3 | 0.13.1 | 1.17.3 | 0.16.11 |
| acpt-pytorch-2.1-cuda12.1 | Ubuntu 20.04 | cu121 | 3.10 | 2.1.2 | 1.17.3 | 0.13.1 | 1.17.3 | 0.16.11 |
| acpt-pytorch-2.0-cuda11.7 | Ubuntu 20.04 | cu117 | 3.10 | 2.0.1 | 1.17.3 | 0.13.1 | 1.17.3 | 0.16.11 |
| acpt-pytorch-1.13-cuda11.7 | Ubuntu 20.04 | cu117 | 3.10 | 1.13.1 | 1.17.3 | 0.13.1 | 1.17.3 | 0.16.11 |
Other packages like fairscale, horovod, msccl, protobuf, pyspark, pytest, pytorch-lightning, tensorboard, NebulaML, torchvision, and torchmetrics are provided to support all training needs.
To learn more, see Create custom ACPT curated environments.
Support
Version updates for supported environments, including the base images they reference, are released every two weeks to address vulnerabilities no older than 30 days. Based on usage, some environments may be deprecated (hidden from the product but usable) to support more common machine learning scenarios.