Docs

Volcano Engine Function as a Service

Serverless GPU

Managing GPU functions

This topic describes how to perform basic management operations on GPU functions, including creating GPU functions, updating function settings, updating images, and deleting functions.

Create GPU functions

veFaaS supports creating GPU functions using container images. This section explains how to create a GPU function.

tip

To use a custom container image for function deployment, refer to Native runtimes to complete code development, image building, and image pushing first.

In the top navigation bar, select the target region.

In the left navigation pane, click Function > Function management.

In the function list, click Create function and select the target creation method. Web applications, microservices, and tasks all support container image-based deployment. Select one that suits your actual business scenario.

Basic information

Parameter	Description
Function name	Specify a custom name for the function. Function names cannot be repeated within the same account. After a function name is created, it cannot be modified.
Description	Enter a description for the function. You can leave this field empty.

Function code

Parameter	Description
Runtime	In this scenario, only Native runtime is supported.
Deployment method	In this scenario, select Image. CR: Select the specific image to be deployed from Volcano Engine Container Registry (CR). You can filter out a specific image by instance, namespace, and repository type (public or private). Public image: Select a pre-built, ready-to-use image provided by veFaaS. It can be used directly, eliminating the need for you to develop code and build an image. tip Public images are contributed by community developers. veFaaS does not maintain them. If you encounter issues during software usage, provide feedback to the community.

Advanced settings

The table describes only GPU-related parameters. For other parameters, see Function configuration.

Parameter	Description
Compute mode	In this scenario, select GPU acceleration.
GPU card type	The following GPU types are supported: NVIDIA A10. Each GPU instance supports only one GPU with 24 GB of video memory and 112 GB of memory. It supports NVIDIA RTX™ features and provides high-performance 3D graphics virtualization capabilities. This GPU type is suitable for deep learning inference and small-scale training scenarios. NVIDIA GPU: Each GPU instance supports only one GPU with 24 GB of video memory and 60 GB of memory. It supports NVIDIA RTX™ features and provides high-performance 3D graphics virtualization capabilities. This GPU type is suitable for scenarios such as large-scale deployments of generative AI inference applications.

Click Confirm to create the GPU function. After the function is created, it appears in the function list.

Update settings

You can modify function settings, such as execution timeout, single-instance concurrency, environment variables, VPC networking, and log delivery. After the update, you need to release the function again for the update to take effect.

In the function list, find the target function and click its name to enter the basic information tab.

Update function settings as needed, such as the VPC network and log delivery settings.

Update images

You can update the container image used by a function. After the update, you need to release the function again for the update to take effect.

In the function list, find the target function and click its name to enter the basic information tab.

Click the "Code" tab and click Modify image.

In the pop-up confirmation box, click Confirm.

On the image modification page, select the desired different image version from CR.

Click Confirm to start updating the image.

The system starts synchronizing the new image. Please wait.

Function deletion

danger

Deleting a function will cause its instances to be immediately released or destroyed. This operation cannot be undone. Proceed with caution.

In the function list, select the target function, and in the Actions column, click ... > Delete.

In the pop-up confirmation dialog box, verify the function information. To proceed with the deletion, click Delete.

Last updated: 2026.08.19 14:06:06