Skip to content

From v7.2 to v8.1

On this page, find out how to upgrade from the older Face SDK version 7.2 to version 8.1.

Web Service

Before upgrading the Web Service from Face SDK v7.2 to v8.1, update the system dependencies on all hosts where the service runs.

GPU

1. Update GPU and image-processing dependencies.

Make sure the GPU is in the list of supported NVIDIA GPU architectures (Turing, Ampere, Ada Lovelace, Hopper, Blackwell) and the following components are upgraded to the versions required by Face SDK v8.1:

Container installations:

  • NVIDIA driver >= 570

Package installations:

  • NVIDIA driver >= 570
  • CUDA 12.8
  • cuDNN 9.17.1

2. After updating the dependencies:

  • Install the latest (8.1) version of Face SDK
  • Deploy the updated Web Service to all environments (test, staging, production) and verify that it starts successfully and passes basic health checks.

Helm chart

Helm chart versions earlier than 3.0.0 are not compatible with Face SDK Web Service v8. Thus, before upgrading to the latest 8.1 version ensure you have the latest faceapi 3.0.0 chart version:

helm repo update
helm search repo faceapi  # CHART VERSION >= 3.0.0
helm upgrade --install my-release regulaforensics/faceapi ...

If you use the embedded Milvus subchart (milvus.enabled: true), additional migration steps may be required when upgrading Milvus (including support for 2.6.4). Follow the steps from the Milvus upgrade guide: Milvus cluster upgrade (Helm).

The subcharts are used for demonstration and Dev/Test purposes related to the liveness and search capabilities. We strongly recommend deploying separate installations of the VectorDatabase (search) and DB (liveness/search) in Production.

Here are the additional resources:

Python WebClient

In v8.1, the generated Python WebClient module structure has changed: imports now use gen.models (plural) instead of gen.model (singular).

Before (v7.2)

from regula.facesdk.webclient.gen.model.process_param import ProcessParam

Now (v8.1)

from regula.facesdk.webclient.gen.models.process_param import ProcessParam