Train label model. Train label model to estimate mu, the parameters used to combine LFs. Parameters ---------- L_tra

Evaluate a standard metric on a set of predictions/probabilities. Parameters ---------- golds An array of gold (int) labels p

Train a MultitaskClassifier. Parameters ---------- model The model to train dataloaders A lis

Update the count and total number.

Label list of data points or a NumPy array with LFs. Parameters ---------- data_points List of data points or Num

Calculate scores for one or more user-specified metrics. Parameters ---------- golds An array of gold (int) label

Return predicted labels, with ties broken according to policy. Policies to break ties include: - "abstain": return an abstain vote (

Calculate cross-entropy loss when targets are probabilities (floats), not ints. PyTorch's F.cross_entropy() method requires integer labels; it do

Check if current counts trigger checkpointing.

Check if current counts trigger evaluation.

Set a prior for the class balance. In order of preference: 1) Use user-provided class_balance 2) Estimate balance from Y_dev

Label Pandas DataFrame of data points with LFs. Parameters ---------- df Pandas DataFrame containing data points

Convert an array to a (possibly flattened) array of ints. Cast all values to ints and possibly flatten [n, 1] arrays to [n]. This method is t

Label Dask DataFrame of data points with LFs. Parameters ---------- df Dask DataFrame containing data points to b

Check if iteration and current metrics necessitate a checkpoint. Parameters ---------- iteration Current training

Load existing label model. Parameters ---------- source Filename to load model from Example ----

Convert an array of probabilistic labels into an array of predictions. Policies to break ties include: "abstain": return an abstain vote (-1)

Calculate scores for the provided DictDataLoaders. Parameters ---------- dataloaders A list of DictDataLoaders to

Get a hashable version of a potentially unhashable object. This helper is used for caching mapper outputs of data points. For common data poi

Convert an array of predictions into an array of probabilistic labels. Parameters ---------- pred A [num_datapoints] or [num_data

Label PySpark RDD of data points with LFs. Parameters ---------- data_points PySpark RDD containing data points t

Calculate the loss for each task and the number of data points contributing. Parameters ---------- X_dict A dict

Convert a list of torch.Tensor into a single torch.Tensor.

Print all metrics in metrics_dict to screen. Parameters ---------- metrics_dict Dictionary of metric names (keys)

Convert the batch into a padded tensor and mask tensor. Parameters ---------- batch The data for padding max_len Max

Reset the memoization cache.

r"""Heuristically choose amongst (possibly) several valid mu values. If there are several values of mu that equivalently satisfy the optimiza

Augment a list of data points using TFs and policy. Parameters ---------- data_points List containing data points

Augment a Pandas DataFrame of data points using TFs and policy. Parameters ---------- df Pandas DataFrame contain

r"""Return the estimated conditional probabilities table. Return the estimated conditional probabilites table cprobs, where cprobs is an

Return data point indices bucketed by label combinations. Parameters ---------- *y A list of np.ndarray of (int) labels Retu

Return instances in x with the specified combination of labels. Parameters ---------- bucket A tuple of label values correspondin

Calculate probabilities, (optionally) predictions, and pull out gold labels. Parameters ---------- dataloader A D

Calculate one or more scores from user-specified and/or user-defined metrics. Parameters ---------- L An [n,m] ma

Generate overlaps and conflicts matrix from label matrix. Parameters ---------- L An [n,m] label matrix with valu

Create augmented version of label matrix. In augmented version, each column is an indicator for whether a certain source or clique of

r"""Overall mu loss. Parameters ---------- l2 A float or np.array representing the per-source regularization

Augment a list of data points using TFs and policy in batches. This method acts as a generator, yielding augmented data points for a

Augment a Pandas DataFrame of data points using TFs and policy in batches. This method acts as a generator, yielding augmented data points fo

Check that operator names are unique.

Return output_dict outputs specified by suffix, ordered by sorted flow_name.

Do a forward pass through the network for all specified tasks. Parameters ---------- X_dict A dict of data fields

Compute frac. of examples each LF labels. Returns ------- numpy.ndarray Fraction of labeled examples for each LF

Compute frac. of examples each LF labels that are labeled by another LF. An overlapping example is one that at least one other LF returns a

Create a pandas DataFrame with the various per-LF statistics. Parameters ---------- Y [n] or [n, 1] np.ndarray of

Convert ``Mapper`` to be compatible with PySpark. Parameters ---------- mapper Mapper to make compatible with PySpark

Calculate user-specified and/or user-defined metrics overall + slices. Parameters ---------- S A recarray with en

Map each label to its corresponding task outputs based on whether the task is available. If remap_labels specified, overrides specific label

r"""Initialize the learned params. - \mu is the primary learned parameter, where each row corresponds to the probability of a clique

r"""L2 loss centered around mu_init, scaled optionally per-source. In other words, diagonal Tikhonov regularization, ||D(\mu-\mu_

Move the model to the device specified in the model config.

Reset the loss counters.

Filter out examples from arrays based on specified labels to filter. The most common use of this method is to remove examples whose gold label is

Generate a synthetic label matrix with true parameters and labels. This function generates a set of labeling function conditional probability tab

Compute frac. of examples each LF labels and labeled differently by another LF. A conflicting example is one that at least one other LF retur

Load trainer config and optimizer state from the specified json file path to the trainer object. The optimizer state is stored, too. However, it only

Create DictDataLoader with slice labels, initialized from specified dataset. Parameters ---------- dataset A Dict

Merge a (potentially nested) dict of kwargs into a config (NamedTuple). Parameters ---------- config An instantiated Config to up

Build mask applied to O^{-1}, O for the matrix approx constraint.

Get indicator vector z where z_i = 1 if x_i is labeled differently by two LFs.

Get indicator vector z where z_i = 1 if x_i is labeled by more than one LF.

Test generated label matrix L for consistency with P, Y. This tests for consistency between the true conditional LF probabilities, P,

Log a scalar variable. Parameters ---------- name Name of the scalar collection value Value o

Modify a dataloader in-place, adding labels for slice tasks. Parameters ---------- dataloader A DictDataLoader whose dataset.Y_di

Label Pandas DataFrame of data points with LFs in parallel using Dask. Parameters ---------- df Pandas DataFrame

Label a single data point with a set of LFs. Parameters ---------- x Data point to label index Index of the data poin

Clear existing checkpoint files, besides the best-yet model.

Add slice labels to dataloader and creates new slice tasks (including base slice). Each slice will get two slice-specific heads: - an indicat

Create uniform X data from [-1, 1] on both axes. Create labels with linear decision boundaries related to the two coordinates of X.