Skip to content

Refactor metadata structure to nest profiling and runtime metrics #261

@keighrim

Description

@keighrim

Currently, when the hwFetch and runningTime parameters are requested, the SDK performs basic profiling and records the results in a flat structure within the app metadata. This mixes dynamic runtime metrics (like execution time and hardware usage) with static configuration data at the root level.

  "metadata": {
    "appRunningTime": "0:07:29.515738",
    "appRunningHardware": {
      "architecture": "x86_64", 
      "cuda": [
        "NVIDIA GeForce RTX 3060, With 12.00 GiB, Used 3.37 GiB"
      ]
    },
    ...
    "appConfiguration": {
      "useClassifier": true,
      "tpModelName": "convnextv2_tiny",
      "tpUsePosModel": true,
      "tpSampleRate": 100,

To improve organization and separate execution statistics from configuration, we should nest profiling results under a dedicated field, such as appProfiling or runtimeMetrics. (Or any other suggestions?)

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    Status

    Done

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions