snorkel.classification.Task

class snorkel.classification.Task(name, module_pool, op_sequence, scorer=<snorkel.analysis.scorer.Scorer object>, loss_func=None, output_func=None)[source]

Bases: object

A single task (a collection of modules and specified path through them).

Parameters
  • name (str) – The name of the task

  • module_pool (ModuleDict) – A ModuleDict mapping module names to the modules themselves

  • op_sequence (Sequence[Operation]) – A list of Operations to execute in order, defining the flow of information through the network for this task

  • scorer (Scorer) – A Scorer with the desired metrics to calculate for this task

  • loss_func (Optional[Callable[…, Tensor]]) – A function that converts final logits into loss values. Defaults to F.cross_entropy() if none is provided. To use probalistic labels for training, use the Snorkel-defined method cross_entropy_with_probs() instead.

  • output_func (Optional[Callable[…, Tensor]]) – A function that converts final logits into ‘outputs’ (e.g. probabilities) Defaults to F.softmax(…, dim=1).

name[source]

See above

module_pool[source]

See above

op_sequence[source]

See above

scorer[source]

See above

loss_func[source]

See above

output_func[source]

See above

__init__(name, module_pool, op_sequence, scorer=<snorkel.analysis.scorer.Scorer object>, loss_func=None, output_func=None)[source]

Initialize self. See help(type(self)) for accurate signature.

Return type

None

Methods

__init__(name, module_pool, op_sequence[, …])

Initialize self.