Status: Needs Review

This page has not been reviewed for accuracy and completeness. Content may be outdated or contain errors.

Configuration Groups¶

Hydra config groups keep current experiments modular and composable.

Current Group Layout¶

configs/
├── data/
├── pipeline/
│   ├── anomaly/
│   └── sam3/
├── plugins/
├── training/
└── trainrun/

Standard Trainrun Composition¶

# @package _global_
defaults:
  - /pipeline/anomaly/rx@pipeline: rx_statistical
  - /data@data: lentils
  - /training@training: default
  - _self_

name: rx_demo
output_dir: ./outputs/${name}

Pipeline Group¶

Current pipeline families:

configs/pipeline/anomaly/rx/
configs/pipeline/anomaly/deep_svdd/
configs/pipeline/anomaly/adaclip/
configs/pipeline/sam3/

Override example:

pipeline:
  nodes:
    - name: RXGlobal
      hparams:
        eps: 1.0e-08

Data Group¶

Use configs/data/ for the data module, split selectors, and loader settings.

Example:

data:
  data_module: cu3s
  splits:
    train:
      - {kind: file_indices, source: data/Lentils/Lentils_000.cu3s, ids: [0, 2, 3]}
    val:
      - {kind: file_indices, source: data/Lentils/Lentils_000.cu3s, ids: [1]}
    test:
      - {kind: file_indices, source: data/Lentils/Lentils_000.cu3s, ids: [5]}
  batch_size: 2
  params:
    cu3s_file_path: data/Lentils/Lentils_000.cu3s
    annotation_json_path: data/Lentils/Lentils_000.json

Training Group¶

Use configs/training/ for trainer and optimizer settings.

Example:

training:
  trainer:
    max_epochs: 50
    accelerator: auto
    devices: 1
  optimizer:
    name: adamw
    lr: 0.001

Plugin Group¶

Use the narrow manifest required by the workflow:

configs/plugins/adaclip.yaml
configs/plugins/bytetrack.yaml
configs/plugins/deepeiou.yaml
configs/plugins/detr.yaml
configs/plugins/sam3.yaml
configs/plugins/trackeval.yaml
configs/plugins/ultralytics.yaml

Each plugin ships its own plugins.yaml manifest; reference the ones you need from your config. The full list of plugin-supplied nodes lives in the Nodes catalog.