Upgrade business pod room after Hami 2.5

[xinyang09]

hami 2.4.0 to hami 2.5.0

The GPU card numbers seen from hami webui and container environment variables are different. We upgraded hami-2.5 yesterday. And now this service keeps reporting video memory overflow

The service allocated 30GB of video memory before the upgrade, and after the upgrade, the video memory was set to 40GB, which also reported room error

[zhegemingzimeibanquan]

which kube-scheduler version you used