View
Window
Models
One card per served model. Power is node GPU power while the model is serving; a tensor-parallel model sums every node it spans.
Loading…
One card per served model. Power is node GPU power while the model is serving; a tensor-parallel model sums every node it spans.
Loading…