nasa-nccs-hpda / astrotime / 13397378004

Builds Branch Commit Type Ran Committer Via Coverage
13397378004 main Set up logging in baseline-nc.py Added TensorFlow push ThomasMaxwell github
80.0
13397309927 main Add optional random downsampling to encoder Introduced the `RandomDownsample` filter to the encoder, triggered when sparsity is greater than 0.0. Updated the `refresh` flag to `True`, and added a `sparsity` parameter to control this behavior. Thi... push ThomasMaxwell github
80.0
13397261467 main Add baseline workflow and refactor core components for extensibility Introduced a new baseline workflow script (`baseline-nc.py`) for sinusoid period detection, incorporating loaders, encoders, and model training logic. Refactored core modules (`... push ThomasMaxwell github
80.0
13381028657 main Add ncTrainer class for neural network training Introduced a new `ncTrainer` class in `astrotime/trainers` to handle training and validation of neural network models. This implementation supports metrics logging, gradient computation, and batch p... push ThomasMaxwell github
80.0
13379428293 main Set sequence start index to fixed 0 and adjust slmax value. Replaced random sequence starting index with a fixed value (0) in both wavelet and baseline encoders to ensure deterministic behavior. Additionally, reduced `slmax` in baseline encoder f... push ThomasMaxwell github
80.0
13379187010 main Set default value for mparms in TrainingFilter and subclasses. Updated the `mparms` argument to default to `None` in `TrainingFilter` and `RandomDownsample`. This ensures cleaner initialization and avoids issues when `mparms` is not provided, sim... push ThomasMaxwell github
80.0
13379152444 main Refactor filtering to support axis parameter and numpy. This update enhances the filter functionality by adding axis support and enabling compatibility with numpy arrays. Redundant filter subclasses were commented out push ThomasMaxwell github
80.0
13378407972 main Update model evaluation and remove redundant build calls Added evaluation functionality with `eval_size` and `model.predict` to both baselines and wavelet embedding scripts. Removed unnecessary `model.build()` calls and adjusted checkpoint loadin... push ThomasMaxwell github
80.0
13378336891 main Add model.build() to ensure proper model initialization This update explicitly calls the build() method on models to initialize their structure before training or loading weights. It prevents potential runtime issues when handling model architect... push ThomasMaxwell github
80.0
13378217611 main Add support for applying training filters to encoders Introduced a filtering mechanism for encoders by implementing `add_filters` and `apply_filters` methods. Filters like ` push ThomasMaxwell github
80.0
13377840836 main Remove unused accuracy metrics from model compilation. Removed the 'accuracy' metric from model compilation across multiple scripts to align with updated requirements. Simplified checkpoint initialization by dropping the monitoring of 'val_accura... push ThomasMaxwell github
80.0
13377320918 main Add wavelet-based and baseline training workflows Introduced two new scripts: one for WaveletEncoder-based CNN training and another utilizing a baseline ValueEncoder. Updated training filters to handle NumPy arrays directly and improved gap maski... push ThomasMaxwell github
80.0
13377036645 main Refactor data handling and update checkpoint file naming. Replaced `xa.Dataset` with `Dict[str, np.ndarray]` for dataset handling in filters to improve type consistency and flexibility. Introduced a new `get_dataset` method in `sinusoid` loader f... push ThomasMaxwell github
80.0
13376300298 main Refactor weight loading logic with `load_weights` method. Centralized the weight-loading process into a reusable `load_weights` method within `CheckpointCallback`. Updated baseline and wavelet-embedding scripts to use this method, ensuring cleane... push ThomasMaxwell github
80.0
13376165857 main Add filtering framework and update training configurations Introduced a new filtering framework in `astrotime.transforms.filters`, enabling data preprocessing with push ThomasMaxwell github
80.0