Training

This page outlines the unit test coverage for the train.py module, which handles loading model training configurations and selecting the most recent trained model based on filename timestamps.

Coverage Overview

The tests focus on the following functions:

  • load_train_config: Loads and validates a training configuration file.

  • find_latest_model: Identifies the latest YOLO model file based on naming pattern or falls back to a default model.

Function Tests

load_train_config

  • test_load_valid_config:
    Ensures a valid training config JSON file loads correctly and includes all expected keys and values.

  • test_missing_file_raises_error:
    Confirms that a missing config file raises a FileNotFoundError.

  • test_missing_required_keys:
    Validates that missing required config fields trigger an AssertionError.

find_latest_model

  • test_returns_latest_model:
    Identifies the most recent YOLO model using timestamped filenames matching the pattern *_updated_yolo.pt.

  • test_returns_fallback_if_no_model:
    Returns the fallback path if no matching model is found in the directory.

  • test_model_pattern_only_matches_updated:
    Verifies that only files with _updated_yolo.pt are considered valid candidates for latest model selection.

Summary

This page outlines a focused and reliable test suite that ensures:

  • Training configuration files are present, valid, and complete

  • Only appropriately named model files are selected for retraining or fine-tuning

  • Robust fallback behavior in the absence of model artifacts

These tests help safeguard the training pipeline from misconfiguration or missing artifacts, ensuring repeatable and predictable training runs.