Hardware requriements

Overview

Model requirements include the GPU, CPU, and RAM needed to run a model. Administrators can create memory and hardware requirement presets for Data Scientists to choose from upon model deployment.

The Model service provides the APIs needed to create and set memory and hardware requirement presets. It provides APIs to list account and model requirement presets, create new presets, and assign and remove them from accounts and models.

The requirements object

{
  "requirementId": 1,
  "gpuUnits": 1,
  "cpuAmount": "...",
  "memoryAmount": "..."
}

Parameter

Type

Description

requirementId

string

The identifier for the memory and hardware requirements set.

gpuUnits

number

An integer number of GPUs required to run the model container. GPU resources are optional.
Example: 1

cpuAmount

string

A number of fractional CPUs required to run the model container. You can limit the percentage of the CPU that the container is allowed to use.
Example: 0.5

memoryAmount

string

The amount of system memory needed to run the model container (e.g. RAM). Measured in bytes, it can be a plain integer or a fixed-point integer using a suffix. Suffixes may be: G, M, K, or their power-of-two equivalents: Gi, Mi, Ki.Mi". Example: the following represent roughly the same value: 128974848, 129e6, 129M, 123Mi.