from finetune.classifier import Classifier, ClassificationPipeline
from finetune.regressor import Regressor, RegressionPipeline
from finetune.base import BaseModel
class MultiFieldClassificationPipeline(ClassificationPipeline):
def _format_for_encoding(self, X):
return [X]
class MultiFieldRegressionPipeline(RegressionPipeline):
def _format_for_encoding(self, X):
return [X]
[docs]class MultiFieldClassifier(Classifier):
"""
Classifies a set of documents into 1 of N classes.
:param config: A :py:class:`finetune.config.Settings` object or None (for default config).
:param \**kwargs: key-value pairs of config items to override.
"""
def _get_input_pipeline(self):
return MultiFieldClassificationPipeline(self.config)
[docs] def finetune(self, Xs, Y=None, batch_size=None):
"""
:param \*Xs: lists of text inputs, shape [batch, n_fields]
:param Y: integer or string-valued class labels. It is necessary for the items of Y to be sortable.
:param batch_size: integer number of examples per batch. When N_GPUS > 1, this number
corresponds to the number of training examples provided to each GPU.
"""
return BaseModel.finetune(self, Xs, Y=Y, batch_size=batch_size)
[docs] def predict(self, Xs):
"""
Produces list of most likely class labels as determined by the fine-tuned model.
:param \*Xs: lists of text inputs, shape [batch, n_fields]
:returns: list of class labels.
"""
return BaseModel.predict(self, Xs)
[docs] def predict_proba(self, Xs):
"""
Produces probability distribution over classes for each example in X.
:param \*Xs: lists of text inputs, shape [batch, n_fields]
:returns: list of dictionaries. Each dictionary maps from X2 class label to its assigned class probability.
"""
return BaseModel.predict_proba(self, Xs)
[docs] def featurize(self, Xs):
"""
Embeds inputs in learned feature space. Can be called before or after calling :meth:`finetune`.
:param \*Xs: lists of text inputs, shape [batch, n_fields]
:returns: np.array of features of shape (n_examples, embedding_size).
"""
return BaseModel.featurize(self, Xs)
[docs]class MultiFieldRegressor(Regressor):
"""
Regresses one or more floating point values given a set of documents per example.
:param config: A :py:class:`finetune.config.Settings` object or None (for default config).
:param \**kwargs: key-value pairs of config items to override.
"""
def _get_input_pipeline(self):
return MultiFieldRegressionPipeline(self.config)
[docs] def finetune(self, Xs, Y=None, batch_size=None):
"""
:param \*Xs: lists of text inputs, shape [batch, n_fields]
:param Y: floating point targets
:param batch_size: integer number of examples per batch. When N_GPUS > 1, this number
corresponds to the number of training examples provided to each GPU.
"""
return BaseModel.finetune(self, Xs, Y=Y, batch_size=batch_size)
[docs] def predict(self, Xs):
"""
Produces list of most likely class labels as determined by the fine-tuned model.
:param \*Xs: lists of text inputs, shape [batch, n_fields]
:returns: list of class labels.
"""
return BaseModel.predict(self, Xs)
[docs] def predict_proba(self, Xs):
"""
Produces probability distribution over classes for each example in X.
:param \*Xs: lists of text inputs, shape [batch, n_fields]
:returns: list of dictionaries. Each dictionary maps from X2 class label to its assigned class probability.
"""
return BaseModel.predict_proba(self, Xs)
[docs] def featurize(self, Xs):
"""
Embeds inputs in learned feature space. Can be called before or after calling :meth:`finetune`.
:param \*Xs: lists of text inputs, shape [batch, n_fields]
:returns: np.array of features of shape (n_examples, embedding_size).
"""
return BaseModel.featurize(self, Xs)