Context:

We are using NVIDIA GPUs and are trying to limit 2 pods max per GPU. Reading through the documentation, we've learned this should be possible with the deviceSplitCount parameter. After following the guide, changing configmaps, restarting containers, the parameter is not honoured (though being correctly modified).

Issue:

We are able to schedule more than 2 pods. On the other hand if we try to schedule more than 10 pods on a GPU, it also sticks to the default value of 10.

https://github.com/Project-HAMi/HAMi/blob/master/charts/hami/templates/scheduler/device-configmap.yaml#L23C7-L23C23