PyTorch Lightning

MetricValCallback

class oml.lightning.callbacks.metric.MetricValCallback(metric: IBasicMetric, log_images: bool = False, loader_idx: int = 0, samples_in_getitem: int = 1)[source]

Bases: Callback

This is a wrapper which allows to use IBasicMetric with PyTorch Lightning.

__init__(metric: IBasicMetric, log_images: bool = False, loader_idx: int = 0, samples_in_getitem: int = 1)[source]
Parameters

  • metric – Metric

  • log_images – Set True if you want to have visual logging

  • loader_idx – Idx of the loader to calculate metric for

  • samples_in_getitem – Some of the datasets return several samples when calling __getitem__, so we need to handle it for the proper calculation. For most of the cases this value equals to 1, but for the dataset which explicitly return triplets, this value must be equal to 3, for a dataset of pairs it must be equal to 2.

ExtractorModule

class oml.lightning.modules.extractor.ExtractorModule(extractor: IExtractor, criterion: Optional[Module] = None, optimizer: Optional[Optimizer] = None, scheduler: Optional[_LRScheduler] = None, scheduler_interval: str = 'step', scheduler_frequency: int = 1, input_tensors_key: str = 'input_tensors', labels_key: str = 'labels', embeddings_key: str = 'embeddings', scheduler_monitor_metric: Optional[str] = None, freeze_n_epochs: int = 0)[source]

Bases: LightningModule

This is a base module to train your model with Lightning.

__init__(extractor: IExtractor, criterion: Optional[Module] = None, optimizer: Optional[Optimizer] = None, scheduler: Optional[_LRScheduler] = None, scheduler_interval: str = 'step', scheduler_frequency: int = 1, input_tensors_key: str = 'input_tensors', labels_key: str = 'labels', embeddings_key: str = 'embeddings', scheduler_monitor_metric: Optional[str] = None, freeze_n_epochs: int = 0)[source]
Parameters

  • extractor – Extractor to train

  • criterion – Criterion to optimize

  • optimizer – Optimizer

  • scheduler – Learning rate scheduler

  • scheduler_interval – Interval of calling scheduler (must be step or epoch)

  • scheduler_frequency – Frequency of calling scheduler

  • input_tensors_key – Key to get tensors from the batches

  • labels_key – Key to get labels from the batches

  • embeddings_key – Key to get embeddings from the batches

  • scheduler_monitor_metric – Metric to monitor for the schedulers that depend on the metric value

  • freeze_n_epochs – number of epochs to freeze model (for n > 0 model has to be a successor of IFreezable interface). When current_epoch >= freeze_n_epochs model is unfreezed. Note that epochs are starting with 0.

extractor_training_pipeline

oml.lightning.pipelines.train.extractor_training_pipeline(cfg: Union[Dict[str, Any], DictConfig]) → None[source]

This pipeline allows you to train and validate a feature extractor which represents images as feature vectors.

The config can be specified as a dictionary or with hydra: https://hydra.cc/. For more details look at pipelines/features_extraction/README.md

extractor_validation_pipeline

oml.lightning.pipelines.validate.extractor_validation_pipeline(cfg: Union[Dict[str, Any], DictConfig]) → Tuple[Trainer, Dict[str, Any]][source]

This pipeline allows you to validate a feature extractor which represents images as feature vectors.

The config can be specified as a dictionary or with hydra: https://hydra.cc/. For more details look at pipelines/features_extraction/README.md

extractor_prediction_pipeline

oml.lightning.pipelines.predict.extractor_prediction_pipeline(cfg: Union[Dict[str, Any], DictConfig]) → None[source]

This pipeline allows you to save features extracted by a feature extractor.

The config can be specified as a dictionary or with hydra: https://hydra.cc/. For more details look at pipelines/features_extraction/README.md

postprocessor_training_pipeline

oml.lightning.pipelines.train_postprocessor.postprocessor_training_pipeline(cfg: DictConfig) → None[source]

This pipeline allows you to train and validate a pairwise postprocessor which fixes mistakes of a feature extractor in retrieval setup.

The config can be specified as a dictionary or with hydra: https://hydra.cc/. For more details look at pipelines/postprocessing/pairwise_postprocessing/README.md