Simplified Node Configuration

[kozlov721]

• metrics, losses, and visualizers can be defined under their respective nodes instead of in separate fields under model
  • No need to specify the attached_to field for such modules
  • The old way is still supported, but raises a deprecation warning
• node.inputs not required for simple models following head -> (neck?) -> head1, head2, ... topology
  • The nodes must be defined in the config in this exact order
  • For multihead networks, the heads must have "Head" in their name for this to work

Example

Old Config

model:
  name: resnet50_classification
  nodes:
    - name: ResNet
      params:
        variant: "50"
        download_weights: True

    - name: ClassificationHead
      inputs:
        - ResNet

  losses:
    - name: CrossEntropyLoss
      attached_to: ClassificationHead

  metrics:
    - name: Accuracy
      is_main_metric: true
      attached_to: ClassificationHead
    - name: Recall
      attached_to: ClassificationHead


  visualizers:
    - name: ClassificationVisualizer
      attached_to: ClassificationHead

New Config

model:
  name: resnet50_classification
  nodes:
    - name: ResNet
      params:
        variant: "50"
        download_weights: True

    - name: ClassificationHead

      losses:
        name: CrossEntropyLoss

      metrics:
        - name: Accuracy
          is_main_metric: true
        - name: Recall

      visualizers:
        name: ClassificationVisualizer

[klemen1999]

LGTM

[github-actions]

Test Results

  1 files    1 suites   23m 38s ⏱️
129 tests 129 ✅ 0 💤 0 ❌
180 runs  180 ✅ 0 💤 0 ❌

Results for commit 9c84111.

[github-actions]

☂️ Python Coverage

current status: ✅

Overall Coverage

Lines	Covered	Coverage	Threshold	Status
5342	5147	96%	90%	🟢

New Files

No new covered files...

Modified Files

File	Coverage	Status
luxonis_train/config/config.py	93%	🟢
TOTAL	93%	🟢

updated for commit: 9c84111 by action🐍