Does vllm support deploy models on vuda?

[Gaohang0804]

I'm trying to deploy internVL2 model with vllm on k8s，I have succeed on 1 A800 80G GPU，but because the low memory required of my model，many gpu memory is wasted.
I just wondering could deploy my model on virtualized cuda，and required lower memory on one k8s pod.

        limits:
          cpu: '15'
          memory: 16Gi
          tencent.com/vcuda-core: '100'
          tencent.com/vcuda-memory: '320'
        requests:
          cpu: '15'
          memory: 16Gi
          tencent.com/vcuda-core: '100'
          tencent.com/vcuda-memory: '320'

[Qhr77]

遇到同样问题了，在k8s上 tencent.com/vcuda-core 设成100 可以部署 改成25 这样报错 。请问你解决了吗