I’ve attempted to run the VIA engine inside k8s cluster. It’s ok to run the application in the Pod via the following yaml file.
apiVersion: v1
kind: Pod
metadata:
name: via-engine
spec:
containers:
- name: via-engine
image: nvcr.io/metropolis/via-dp/via-engine:2.0-dp
resources:
limits:
nvidia.com/gpu: 1
requests:
nvidia.com/gpu: 1
ports:
- containerPort: 8000
name: backend-port
- containerPort: 9000
name: frontend-port
env:
- name: BACKEND_PORT
value: "8000"
- name: FRONTEND_PORT
value: "9000"
- name: NVIDIA_API_KEY
value: "<NVIDIA_API_KEY>"
- name: NGC_API_KEY
value: "<NGC_API_KEY>"
- name: VLM_MODEL_TO_USE
value: "vita-2.0"
- name: MODEL_PATH
value: "ngc:nvidia/tao/vita:2.0.1"
restartPolicy: Never
It’s ok to load the GUI.
But when upload a small video file from GUI, it will just stuck there.
Then I’ve directly uploaded the video file via VIA’s backend API. It’s fine.
Not sure where is the problem in VIA engine, the network? or any permission? Seems like it’s a frontend issue.