Proposal to add nameplate/design power to documentation

[dslik]

This PR adds text developed by the MLPerf Power WG describing how nameplate and design power is calculated, reported and displayed in the MLPerf Inference results table.

Outstanding items:

• Review by MLPerf Inference WG
• Addition of image showing mockup of results table (if required)

[github-actions]

MLCommons CLA bot All contributors have signed the MLCommons CLA ✍️ ✅

[guthrieg]

This PR will make it incredibly easy for organizations to include power measurements in their inference submissions to align with regulations of the upper bound power measurements. I am in support of this.

[arav-agarwal2]

For validation purposes, is the intent to validate what's expected in the RedFish PowerSupply schema for each item, or does "based on the PowerSupply schema" here mean that there's some list of required subfields?

I'm asking this as the PowerSupply schema only requires the following fields, according to my understanding of their JSON schema:

• @odata.id
• @odata.type
• Id
• Name

The @OData fields are used to identify the schema and the version of the schema used, so while I can omit them we'd need to pick a version of the PowerSupply schema for our purposes, or decide to have a YAML-wide @odata.type field to record the PowerSupply schema version.

I'm more worried about the fields, like PowerCapacityWatts, that aren't required by the resource schema but are probably required to make the data useful for everyone. If we want validation on this YAML, can we include which fields are required for submission?