Skip to content

Latest commit

 

History

History
38 lines (31 loc) · 1.78 KB

File metadata and controls

38 lines (31 loc) · 1.78 KB

Changelog

[1.1.0] - 2026-02-26

Added

  • pytest test suite covering model output shapes, gradient flow, NaN-aware loss edge cases, dataset utilities, and config loading
  • Sample prediction images (assets/sample_predictions_resnet.png)
  • Training curve visualizations (assets/training_curves.png)
  • Model comparison overlay (assets/model_comparison.png)

Changed

  • Rewrote README with badges (Python, PyTorch, TensorFlow, License, Kaggle RMSE), embedded images, architecture diagrams, and code examples
  • Added parameter counts and strategy details to results table
  • Added references section (He et al. 2016, Kaggle)

Fixed

  • Deprecated df.fillna(method="ffill") replaced with df.ffill() for pandas 2.x compatibility

[1.0.0] - 2026-02-26

Added

  • Modular Python package with keypoints namespace
  • CNN model (Keras/TensorFlow): 3 conv blocks with LeakyReLU, two-phase training (Adam + SGD)
  • ResNet model (PyTorch): 6-stage, 12 residual blocks with batch normalization
  • Custom MSELossIgnoreNan for training with partially-labeled data
  • Shared dataset loading and preprocessing utilities for both frameworks
  • Centralized YAML configuration with typed frozen dataclasses
  • Training entry points with --config and --debug CLI flags
  • Unified prediction script with Kaggle submission generation
  • Visualization utilities: keypoint overlay, training curves, prediction grids, model comparison
  • Architecture documentation with pipeline and block diagrams

Changed

  • Merged code from two separate repositories into a single modular package
  • Extracted all hardcoded hyperparameters to config/default.yaml
  • Replaced print() statements with Python logging module
  • Added type hints and docstrings throughout
  • PyTorch and TensorFlow are optional dependencies with lazy imports