HPE Machine Learning Development Environment: технические характеристики и документация
|
Have feedback on QuickSpecs? We're listening | |
|
HPE Machine Learning Development Environment Software QuickSpecs
A platform that makes it easy for IT and MLOps teams to set up and share AI infrastructure to improve collaboration and productivity for ML teams, and reduce costs. Built upon the open-source Determined model development platform, HPE Machine Learning Development Environment Software allows model developers and researchers to focus on building better models faster, by reducing complexity, and removing the need to write boilerplate code associated with managing ML infrastructure. It easily integrates with popular ML frameworks and tools, and supports cloud or on-premises infrastructure environments, with a consistent user experience (UX). |
|
|
With HPE Machine Learning Development Environment Software, you can easily: - Train models faster - Build better models - Manage and share your AI infrastructure - Track and reproduce your work - Integrate into your enterprise - Bring your own cloud - Evaluate and customize large language models (LLMs) |
|
|
|
What's New - Generative AI studio is a standard feature now available with HPE Machine Learning Development Environment. Generative AI studio is the easiest, and most secure way to build and scale generative AI applications on-premises using foundational models. - Run ML and HPC jobs alongside each other on the same cluster, with support for workload managers like Slurm® or PBS®, and secure container runtimes like Singularity/Apptainer, Podman, or NVIDIA® Enroot. - Use Role-Based Access Controls (RBAC) to authorize development and MLOps teams to securely collaborate and share ML resources and artifacts. - Train models on NVIDIA or AMD® GPUs without any code changes, with foundational support for accelerator heterogeneity. - Use DeepSpeed® for 3D-Parallel (Data-, Model-, and Pipeline-parallel) distributed training, to speed up training of large models like GPT-NeoX. - Use PyTorch® Distributed Data Parallel (DDP) for flexibility and choice of distributed training strategies. - Version, annotate, and organize trained models in the Model Registry so that MLOps teams can effectively collaborate with model developers to manage your models' lifecycle. - Define your own logic to coordinate across multiple trials within an experiment. - Implement your own custom hyperparameter search algorithms, ensembling, active learning, neural architecture search, and reinforcement learning. - New SKUs
|
|
|
|
Train Models Faster Train models faster, and at scale, without changing your model code, or writing cumbersome boilerplate to manage your ML infrastructure.
We take care of ML infrastructure provisioning, networking, data loading, checkpointing, fault-tolerance and high-availability - i.e., we make distributed model training easy, fast, and cost-efficient.
HPE Machine Learning Development Environment Software ships cutting-edge distributed training strategies and techniques: - DeepSpeed: for 3D-Parallel (Data-, Model-, and Pipeline-parallel) distributed training, to speed up training of large models like GPT-NeoX, and beyond! - Horovod: for easy-to-use data-parallel distributed training. - PyTorch Distributed Data Parallel (DDP): for flexibility and choice of distributed training strategies. |
|
|
|
Build Better Models Find better model configurations efficiently with a production-grade implementation from the creators of the cutting-edge Asynchronous Successive Halving (ASHA) Hyperband for hyperparameter tuning.
Define your own logic to coordinate across multiple trials within an experiment.
Implement your own custom hyperparameter search algorithms, ensembling, active learning, neural architecture search, and reinforcement learning methods. |
|
|
|
Efficiently Manage and Share Your AI Infrastructure Efficiently manage and share your on-premises or cloud GPUs and accelerators with Machine Learning workflow-aware smart scheduling and resource management to improve productivity and collaboration for your ML development and operations teams.
Consistent User Experience for deployments ranging from laptop to a supercomputer scale, and everything in-between: - Baremetal - Virtual Machine (incl. cloud-native and on-premises Infrastructure-as-a-Service solutions) - Kubernetes® - Slurm - PBS
Run ML and HPC jobs alongside each other on the same cluster, with support for workload managers like Slurm or PBS, and secure container runtimes like Singularity/Apptainer, Podman, or NVIDIA® Enroot.
Seamlessly use spot or preemptible instances to manage cloud costs.
Train models on NVIDIA or AMD® GPUs without any code changes, with foundational support for accelerator heterogeneity. |
|
|
|
Track and Reproduce Your Work Easily track and reproduce your work with experiment tracking that works out-of-the-box: covering model code, configuration, hyperparameters, metrics, and checkpoints.
Version, annotate, and organize trained models with our built-in Model Registry, enabling MLOps teams to effectively collaborate with model developers to manage your models' lifecycle. |
|
|
|
Integrate Into Your Enterprise Authenticate users using enterprise Single Sign-On (SSO) services provided by Active Directory®, Okta®, PingID®, etc., with support for OpenID Connect (OIDC) and SAML.
Integrate with user provisioning systems to automate the onboarding and offboarding of your teams, with support for SCIM. Use Role-Based Access Controls (RBAC) to authorize development and MLOps teams to securely collaborate and share ML resources and artifacts.
|
|
Bring Your Own Cloud Using HPE Machine Learning Development Environment Software as a managed service, deploy the MLDE core platform within your cloud of choice, whether that is AWS or GCP.
Supported environments - AWS using EC2 - GCP using GKE |
|
|
|
Supported Hardware HPE Machine Learning Development Environment Software can be deployed on hardware equipped with NVIDIA or AMD GPUs, on a variety of on-premises or cloud environments.
Hardware - The master node should be configured with at least four Intel Broadwell or later CPU cores, 8 GB of RAM, and 200 GB of free disk space. The master node does not need GPUs. - Each GPU-equipped compute node should be configured with at least two Intel Broadwell or later CPU cores, 4 GB of RAM, and 50 GB of free disk space.
Notes: Most of the disk space required by the master is for the experiment metadata database. If PostgreSQL is set up on a different machine, the disk space requirements for the master are minimal.
Supported Operating Systems and Platforms - Red Hat® Enterprise Linux (RHEL®); SUSE® Linux Enterprise Server (SLES®); Ubuntu® |
|
|
|
For the most up-to-date information on HPE Services, please refer to the HPE Services - Supplemental QuickSpecs, which provides a comprehensive and regularly updated overview of available services. |
|
HPE Machine Learning Development Environment Software Products HPE Machine Learning Development Environment Software is licensed on a per-GPU basis for the duration of the stated subscription term.
| ||||||||||||||||||
|
|
|
Date |
Version History |
Action |
Description of Change |
|
20-Jul-2026 |
Version 7 |
Changed |
- Updated GreenLake references to align with current branding and terminology standards. - Updated Supplemental Services QuickSpecs content for consistency and accuracy. |
|
16-Feb-2026 |
Version 6 |
Changed |
Visual rebranding only-updated typography, colors, and design elements to align with new HPE brand standards. No technical specifications or content were modified. |
|
03-Sep-2024 |
Version 5 |
Changed |
Standard Features section was updated. |
|
03-Jun-2024 |
Version 4 |
Changed |
Overview and Configuration Information sections were updated. |
|
18-Dec-2023 |
Version 3 |
Changed |
HPE Services Rebranding |
|
05-Sep-2023 |
Version 2 |
Changed |
Overview, Standard Features and Configuration Information sections were updated. |
|
06-Mar-2023 |
Version 1 |
New |
New QuickSpecs |
|
|
|
Have feedback on QuickSpecs? We're listening | |
|
|
|
| |
|
|
|
Chat now | |
|
|
|
© Copyright 2026 Hewlett Packard Enterprise Development LP. The information contained herein is subject to change without notice. The only warranties for Hewlett Packard Enterprise products and services are set forth in the express warranty statements accompanying such products and services. Nothing herein should be construed as constituting an additional warranty. Hewlett Packard Enterprise shall not be liable for technical or editorial errors or omissions contained herein.
For hard drives, 1 GB = 1 billion bytes. Actual formatted capacity is less.
a50006978enw - 17088 - WorldWide - V7 - 20-July-2026 | |
|
|
|
HEWLETT PACKARD ENTERPRISE HPE.com |
|
