def __init__(self, num_classes: int, encoder: torch.nn.Module,
freeze_encoder: bool, class_weights: Optional[torch.Tensor]):
super().__init__()
self.num_classes = num_classes
self.encoder = encoder
self.freeze_encoder = freeze_encoder
self.class_weights = class_weights
self.encoder.eval()
self.classifier_head = SSLEvaluator(n_input=get_encoder_output_dim(
self.encoder),
n_hidden=None,
n_classes=num_classes,
p=0.20)
if self.num_classes == 2:
self.train_metrics = ModuleList([
AreaUnderRocCurve(),
AreaUnderPrecisionRecallCurve(),
Accuracy05()
])
self.val_metrics = ModuleList([
AreaUnderRocCurve(),
AreaUnderPrecisionRecallCurve(),
Accuracy05()
])
else:
# Note that for multi-class, Accuracy05 is the standard multi-class accuracy.
self.train_metrics = ModuleList([Accuracy05()])
self.val_metrics = ModuleList([Accuracy05()])
def __init__(self,
learning_rate: float,
class_weights: Optional[torch.Tensor] = None,
**kwargs: Any) -> None:
"""
Creates a hook to evaluate a linear model on top of an SSL embedding.
:param class_weights: The class weights to use when computing the cross entropy loss. If set to None,
no weighting will be done.
:param length_linear_head_loader: The maximum number of batches in the dataloader for the linear head.
"""
super().__init__(**kwargs)
self.weight_decay = 1e-4
self.learning_rate = learning_rate
self.train_metrics: List[Metric] = [AreaUnderRocCurve(), AreaUnderPrecisionRecallCurve(),
Accuracy05()] \
if self.num_classes == 2 else [Accuracy05()]
self.val_metrics: List[Metric] = [AreaUnderRocCurve(), AreaUnderPrecisionRecallCurve(),
Accuracy05()] \
if self.num_classes == 2 else [Accuracy05()]
self.class_weights = class_weights
self.evaluator_state: Optional[OrderedDict] = None
self.optimizer_state: Optional[OrderedDict] = None
def _get_metrics_computers(self) -> ModuleList:
"""
Gets the objects that compute metrics for the present kind of models, for a single prediction target.
"""
if self.is_classification_model:
return ModuleList([Accuracy05(),
AccuracyAtOptimalThreshold(),
OptimalThreshold(),
FalsePositiveRateOptimalThreshold(),
FalseNegativeRateOptimalThreshold(),
AreaUnderRocCurve(),
AreaUnderPrecisionRecallCurve(),
BinaryCrossEntropy()])
else:
return ModuleList([MeanAbsoluteError(), MeanSquaredError(), ExplainedVariance()])
def create_metric_computers(self) -> ModuleDict:
return ModuleDict({MetricsDict.DEFAULT_HUE_KEY: ModuleList([Accuracy05()])})