seqgra.evaluator.gradientbased.smoothgradevaluator module

Smooth Grad Difference Evaluator

class SmoothGradEvaluator(learner: seqgra.learner.learner.Learner, output_dir: str, importance_threshold: Optional[float] = None, stdev_spread: float = 0.15, nsamples: int = 25, magnitude: bool = True, silent: bool = False)[source]

Bases: seqgra.evaluator.gradientbased.abstractdifferencegradientevaluator.AbstractDifferenceGradientEvaluator

Smooth Grad difference gradient evaluator for PyTorch models

modified from https://github.com/PAIR-code/saliency/blob/master/saliency/base.py#L80

evaluate_model(set_name: str = 'test', subset_idx: Optional[List[int]] = None, subset_n: Optional[int] = None, subset_labels: Optional[List[str]] = None, subset_n_per_label: bool = True, subset_shuffle: bool = True, subset_threshold: Optional[float] = None, suppress_plots: bool = False)Any
explain(x, y1, y2)[source]
get_layer(key_list)
select_examples(set_name: str = 'test', labels: Optional[Set[str]] = None, threshold: Optional[float] = None)seqgra.schema.AnnotatedExampleSet

Returns all correctly classified examples that exceed the threshold.

for the specified labels and set that exceed the threshold.

Parameters

TODO

Returns

TODO

select_n_examples(set_name: str = 'test', n: Optional[int] = None, labels: Optional[Set[str]] = None, n_per_label: bool = True, shuffle: bool = True, threshold: Optional[float] = None)seqgra.schema.AnnotatedExampleSet