Versioned asset that pairs a Docker image with a pip or conda specification, so every job or deployment using it gets identical dependencies. You point to it by name plus a version number, or by name with @latest.
Also called environment.
Read more: Microsoft Learn
In the Ultra Transcenders books
Each book explains Azure Machine Learning environment in context, with comparison tables and the common traps.
Terms in this definition
- Conda specification
To build a versioned Azure Machine Learning environment, this YAML file names the Python packages to be layered onto a base Docker image.
- Job
A sequence of steps executed together on one agent or runner, or on the server for agentless work. While running, each occupies one of your parallel jobs.
- Model deployment
Inside a resource, each deployment is a named copy of a Foundry model with its own TPM quota and deployment type. Requests identify the model by this deployment name.
Related terms
- Accessibility level
Whether a Container Apps environment is reachable publicly or only privately, fixed when the environment is made. With external, its virtual IP sits on a public address; with internal, an internal load balancer in your own virtual network holds it.
- App settings
Environment variables injected into an App Service app as name/value pairs, taking precedence over values in appsettings.json or Web.config. Each can be marked as specific to a slot.
- Approvals and checks
Resource owners configure these in the web portal rather than YAML, on environments, agent pools, repositories, secure files, variable groups and service connections; a stage can't use the resource until all pass. GitHub's equivalent: environment protection rules.
- Azure Artifacts Credential Provider
Signs dotnet and NuGet in to Artifacts feeds; unattended jobs supply credentials through environment variables, people sign in interactively.
- Azure Databricks workspace
A deployed Azure Databricks environment where a group of people build and run notebooks, jobs and compute. Accounts often hold several, each attached to a Unity Catalog metastore within its own Azure region.
- Azure landing zone
The architecture Microsoft recommends for running a multi-subscription Azure environment that is governed, secure and able to scale. It consists of a single platform landing zone plus workload landing zones operating inside the guardrails it provides.
- Azure Machine Learning component
Reusable building block for a pipeline, versioned and self-contained, much like a function: it declares a name, inputs and outputs, and bundles the command, code and environment it runs.
- Azure NAT Gateway
Gives every resource in a subnet a managed way out to the internet through one or more fixed public IP addresses, while accepting no unsolicited inbound connections. Azure Container Apps can use it only when the environment is a workload profiles one.