This section explains what environments are and how to create and use them.
Environments are one type of workload assets. An environment consists of a configuration that simplifies how workloads are submitted and can be used by AI practitioners when they submit their workloads.
An environment asset is a preconfigured building block that encapsulates aspects for the workload such as:
Container image and container configuration
The type of workload it serves
The Environments table can be found under Workload manager in the NVIDIA Run:ai platform.
The Environment table provides a list of all the environment defined in the platform and allows you to manage them.
The Environments table consists of the following columns:
The name of the environment
A description of the environment
The scope of this environment within the organizational tree. Click the name of the scope to view the organizational tree diagram
The application or service to be run by the workload
This can be either standard for running workloads on a single node or distributed for running distributed workloads on multiple nodes
The tools and connection types the environment exposes
The list of existing workloads that use the environment
The workload types that can use the environment (Workspace/ Training / Inference)
The list of workload templates that use this environment
The user who created the environment. By default NVIDIA Run:ai UI comes with preinstalled environments created by NVIDIA Run:ai
The timestamp of when the environment was created
The timestamp of when the environment was last updated
The cluster with which the environment is associated
Tools Associated with the EnvironmentClick one of the values in the tools column to view the list of tools and their connection type.
The name of the tool or application AI practitioner can set up within the environment. For more information, see Integrations.
The method by which you can access and interact with the running workload. It's essentially the "doorway" through which you can reach and use the tools the workload provide. (E.g node port, external URL, etc)
Workloads Associated with the EnvironmentClick one of the values in the Workload(s) column to view the list of workloads and their parameters.
The workload that uses the environment
The workload type (Workspace/Training/Inference)
Customizing the Table ViewFilter - Click ADD FILTER, select the column to filter by, and enter the filter values
Search - Click SEARCH and type the value to search by
Sort - Click each column header to sort by
Column selection - Click COLUMNS and select the columns to display in the table
Download table - Click MORE and then Click Download as CSV. Export to CSV is limited to 20,000 rows.
When installing NVIDIA Run:ai, you automatically get the environments created by NVIDIA Run:ai to ease up the onboarding process and support different use cases out of the box. These environments are created at the scope of the account.
Note
The environments listed below are available based on your cluster settings. Some environments, such as vscode and rstudio, are only available in clusters with host-based routing.
nvcr.io/nvidia/clara/bionemo-framework:2.5
A framework developed by NVIDIA for large-scale biomolecular models, optimized to support drug discovery, genomics, and protein structure prediction
runai.jfrog.io/core-llm/llm-app
A user interface for interacting with chat-based AI models, often used for testing and deploying chatbot applications
runai.jfrog.io/core-llm/quickstart-inference:gpt2-cpu
A package containing an inference server, GPT2 model and chat UI often used for quick demos
jupyter-lab / jupyter-scipy
An interactive development environment for Jupyter notebooks, code, and data visualization
gcr.io/run-ai-demo/jupyter-tensorboard
An integrated combination of the interactive Jupyter development environment and TensorFlow's visualization toolkit for monitoring and analyzing ML models
runai.jfrog.io/core-llm/runai-vllm:v0.6.4-0.10.0
A vLLM-based server that hosts and serves large language models for inference, enabling API-based access to AI models
nvcr.io/nvidia/nemo:25.02
A framework for training and deploying LLMs and generative AI developed by NVIDIA with automated data processing, model training techniques, and flexible deployment options
nvcr.io/nvidia/pytorch:25.02-py3
An integrated deep learning framework accelerated by NVIDIA, built for dynamic training and seamless compatibility with Python tools like NumPy and SciPy
An integrated development environment (IDE) for R, commonly used for statistical computing and data analysis
tensorboard / tensorboad-tensorflow
tensorflow/tensorflow:latest
A visualization toolkit for TensorFlow that helps users monitor and analyze ML models, displaying various metrics and model architecture
ghcr.io/coder/code-server
A fast, lightweight code editor with powerful features like intelligent code completion, debugging, Git integration, and extensions, ideal for web development, data science, and more
Environment creation is limited to specific roles
To add a new environment:
Go to the Environments table
Select under which cluster to create the environment
Enter a name for the environment. The name must be unique.
Optional: Provide a description of the essence of the environment
Enter the Image URL If a token or secret is required to pull the image, it is possible to create it via credentials of type docker registry. These credentials are automatically used once the image is pulled (which happens when the workload is submitted)
Set the image pull policy - the condition for when to pull the image from the registry
Set the workload architecture:
Standard Only standard workloads can use the environment. A standard workload consists of a single process.
Distributed Only distributed workloads can use the environment. A distributed workload consists of multiple processes working together. These processes can run on different nodes.
Select a framework from the list.
Set the workload type:
Inference
When inference is selected, define the endpoint of the model by providing both the protocol and the container’s serving port
Optional: Set the connection for your tool(s). The tools must be configured in the image. When submitting a workload using the environment, it is possible to connect to these tools
Select the tool from the list (the available tools varies from IDE, experiment tracking, and more, including a custom tool for your choice)
Select the connection type
External URL
Auto generate A unique URL is automatically created for each workload using the environment
Custom URL The URL is set manually
Node port
Auto generate A unique port is automatically exposed for each workload using the environment
Custom URL Set the port manually
Optional: Set a command and arguments for the container running the pod
When no command is added, the default command of the image is used (the image entrypoint)
The command can be modified while submitting a workload using the environment
The argument(s) can be modified while submitting a workload using the environment
Optional: Set the environment variable(s)
Click +ENVIRONMENT VARIABLE
Select the source for the environment variable
Custom
Add instructions for the expected value if any
Credentials - Select an existing credential as the environment variable
Select a credential name To add new credentials to the credentials list, and for additional information, see Credentials.
ConfigMap - Select a predefined ConfigMap
The environment variables can be modified and new variables can be added while submitting a workload using the environment
Optional: Set the container’s working directory to define where the container’s process starts running. When left empty, the default directory is used.
Optional: Set where the UID, GID and supplementary groups are taken from, this can be:
From the IdP token (only available in an SSO installations)
Custom (manually set) - decide whether the submitter can modify these value upon submission.
Set the User ID (UID), Group ID (GID) and the supplementary groups that can run commands in the container
Add Supplementary groups (multiple groups can be added, separated by commas)
Disable Allow the values above to be modified within the workload if you want the above values to be used as the default
Optional: Select Linux capabilities - Grant certain privileges to a container without granting all the privileges of the root user.
To edit an existing environment:
Select the environment you want to edit
Update the environment and click SAVE ENVIRONMENT
Note
The already bound workload that is using this asset will not be affected.
llm-server and chatbot-ui environments cannot be edited.
To copy an existing environment:
Select the environment you want to copy
Enter a name for the environment. The name must be unique.
Update the environment and click CREATE ENVIRONMENT
To delete an environment:
Select the environment you want to delete
On the dialog, click DELETE to confirm
Note
The already bound workload that is using this asset will not be affected.
Go to the Environment API reference to view the available actions
RetroSearch is an open source project built by @garambo | Open a GitHub Issue
Search and Browse the WWW like it's 1997 | Search results from DuckDuckGo
HTML:
3.2
| Encoding:
UTF-8
| Version:
0.7.4