Skip to main content
Version: 0.93

snorkelflow.client.models.add_predictions

snorkelflow.client.models.add_predictions(node, *, model_uid, x_uids, predicted_labels=None, predicted_probs=None, user_format=False)

Register predictions generated by models (and optionally, ground truth labels). To learn more about the Label format for your use case you can view Label Spaces GT formats.

Parameters:

  • node (int) – UID of the node for which we’re adding predictions

  • model_uid (int) – ID corresponding to the model that produced predictions.

  • x_uids (Union[List[str], ndarray]) – UIDs of data points for which predictions were made.

  • predicted_labels (Union[List[int], ndarray, None], default: None) – Predicted labels corresponding to x_uids.

  • predicted_probs (Union[List[List[float]], ndarray, None], default: None) – Predicted label probabilities corresponding to x_uids. These probabilities should be a list of lists or 2-D numpy array of dimension (len(x_uids)) x (num_classes). All probabilities must be greater than zero and those for each data point should sum to 1.

  • user_format (bool, default: False) – True if predictions and ground truth are provided in user format, False otherwise.

Return type:

None