rasa-pro 3.11.0__py3-none-any.whl → 3.11.0a1__py3-none-any.whl

rasa/nlu/classifiers/logistic_regression_classifier.py

@@ -1,21 +1,22 @@
 from typing import Any, Text, Dict, List, Type, Tuple
 
+import joblib
 import structlog
 from scipy.sparse import hstack, vstack, csr_matrix
 from sklearn.exceptions import NotFittedError
 from sklearn.linear_model import LogisticRegression
 from sklearn.utils.validation import check_is_fitted
 
-from rasa.engine.graph import ExecutionContext, GraphComponent
-from rasa.engine.recipes.default_recipe import DefaultV1Recipe
 from rasa.engine.storage.resource import Resource
 from rasa.engine.storage.storage import ModelStorage
+from rasa.engine.recipes.default_recipe import DefaultV1Recipe
+from rasa.engine.graph import ExecutionContext, GraphComponent
 from rasa.nlu.classifiers import LABEL_RANKING_LENGTH
-from rasa.nlu.classifiers.classifier import IntentClassifier
 from rasa.nlu.featurizers.featurizer import Featurizer
-from rasa.shared.nlu.constants import TEXT, INTENT
-from rasa.shared.nlu.training_data.message import Message
+from rasa.nlu.classifiers.classifier import IntentClassifier
 from rasa.shared.nlu.training_data.training_data import TrainingData
+from rasa.shared.nlu.training_data.message import Message
+from rasa.shared.nlu.constants import TEXT, INTENT
 from rasa.utils.tensorflow.constants import RANKING_LENGTH
 
 structlogger = structlog.get_logger()
@@ -183,11 +184,9 @@ class LogisticRegressionClassifier(IntentClassifier, GraphComponent):
 
     def persist(self) -> None:
         """Persist this model into the passed directory."""
-        import skops.io as sio
-
         with self._model_storage.write_to(self._resource) as model_dir:
-            path = model_dir / f"{self._resource.name}.skops"
-            sio.dump(self.clf, path)
+            path = model_dir / f"{self._resource.name}.joblib"
+            joblib.dump(self.clf, path)
             structlogger.debug(
                 "logistic_regression_classifier.persist",
                 event_info=f"Saved intent classifier to '{path}'.",
@@ -203,21 +202,9 @@ class LogisticRegressionClassifier(IntentClassifier, GraphComponent):
         **kwargs: Any,
     ) -> "LogisticRegressionClassifier":
         """Loads trained component (see parent class for full docstring)."""
-        import skops.io as sio
-
         try:
             with model_storage.read_from(resource) as model_dir:
-                classifier_file = model_dir / f"{resource.name}.skops"
-                unknown_types = sio.get_untrusted_types(file=classifier_file)
-
-                if unknown_types:
-                    structlogger.error(
-                        f"Untrusted types found when loading {classifier_file}!",
-                        unknown_types=unknown_types,
-                    )
-                    raise ValueError()
-
-                classifier = sio.load(classifier_file, trusted=unknown_types)
+                classifier = joblib.load(model_dir / f"{resource.name}.joblib")
                 component = cls(
                     config, execution_context.node_name, model_storage, resource
                 )

rasa/nlu/classifiers/sklearn_intent_classifier.py

@@ -1,6 +1,6 @@
 from __future__ import annotations
-
 import logging
+from rasa.nlu.featurizers.dense_featurizer.dense_featurizer import DenseFeaturizer
 import typing
 import warnings
 from typing import Any, Dict, List, Optional, Text, Tuple, Type
@@ -8,18 +8,18 @@ from typing import Any, Dict, List, Optional, Text, Tuple, Type
 import numpy as np
 
 import rasa.shared.utils.io
+import rasa.utils.io as io_utils
 from rasa.engine.graph import GraphComponent, ExecutionContext
 from rasa.engine.recipes.default_recipe import DefaultV1Recipe
 from rasa.engine.storage.resource import Resource
 from rasa.engine.storage.storage import ModelStorage
-from rasa.nlu.classifiers import LABEL_RANKING_LENGTH
-from rasa.nlu.classifiers.classifier import IntentClassifier
-from rasa.nlu.featurizers.dense_featurizer.dense_featurizer import DenseFeaturizer
 from rasa.shared.constants import DOCS_URL_TRAINING_DATA_NLU
+from rasa.nlu.classifiers import LABEL_RANKING_LENGTH
 from rasa.shared.exceptions import RasaException
 from rasa.shared.nlu.constants import TEXT
-from rasa.shared.nlu.training_data.message import Message
+from rasa.nlu.classifiers.classifier import IntentClassifier
 from rasa.shared.nlu.training_data.training_data import TrainingData
+from rasa.shared.nlu.training_data.message import Message
 from rasa.utils.tensorflow.constants import FEATURIZERS
 
 logger = logging.getLogger(__name__)
@@ -266,20 +266,14 @@ class SklearnIntentClassifier(GraphComponent, IntentClassifier):
 
     def persist(self) -> None:
         """Persist this model into the passed directory."""
-        import skops.io as sio
-
         with self._model_storage.write_to(self._resource) as model_dir:
             file_name = self.__class__.__name__
-            classifier_file_name = model_dir / f"{file_name}_classifier.skops"
-            encoder_file_name = model_dir / f"{file_name}_encoder.json"
+            classifier_file_name = model_dir / f"{file_name}_classifier.pkl"
+            encoder_file_name = model_dir / f"{file_name}_encoder.pkl"
 
             if self.clf and self.le:
-                # convert self.le.classes_ (numpy array of strings) to a list in order
-                # to use json dump
-                rasa.shared.utils.io.dump_obj_as_json_to_file(
-                    encoder_file_name, list(self.le.classes_)
-                )
-                sio.dump(self.clf.best_estimator_, classifier_file_name)
+                io_utils.json_pickle(encoder_file_name, self.le.classes_)
+                io_utils.json_pickle(classifier_file_name, self.clf.best_estimator_)
 
     @classmethod
     def load(
@@ -292,36 +286,21 @@ class SklearnIntentClassifier(GraphComponent, IntentClassifier):
     ) -> SklearnIntentClassifier:
         """Loads trained component (see parent class for full docstring)."""
         from sklearn.preprocessing import LabelEncoder
-        import skops.io as sio
 
         try:
             with model_storage.read_from(resource) as model_dir:
                 file_name = cls.__name__
-                classifier_file = model_dir / f"{file_name}_classifier.skops"
+                classifier_file = model_dir / f"{file_name}_classifier.pkl"
 
                 if classifier_file.exists():
-                    unknown_types = sio.get_untrusted_types(file=classifier_file)
-
-                    if unknown_types:
-                        logger.error(
-                            f"Untrusted types ({unknown_types}) found when "
-                            f"loading {classifier_file}!"
-                        )
-                        raise ValueError()
-                    else:
-                        classifier = sio.load(classifier_file, trusted=unknown_types)
-
-                    encoder_file = model_dir / f"{file_name}_encoder.json"
-                    classes = rasa.shared.utils.io.read_json_file(encoder_file)
+                    classifier = io_utils.json_unpickle(classifier_file)
 
+                    encoder_file = model_dir / f"{file_name}_encoder.pkl"
+                    classes = io_utils.json_unpickle(encoder_file)
                     encoder = LabelEncoder()
-                    intent_classifier = cls(
-                        config, model_storage, resource, classifier, encoder
-                    )
-                    # convert list of strings (class labels) back to numpy array of
-                    # strings
-                    intent_classifier.transform_labels_str2num(classes)
-                    return intent_classifier
+                    encoder.classes_ = classes
+
+                    return cls(config, model_storage, resource, classifier, encoder)
         except ValueError:
             logger.debug(
                 f"Failed to load '{cls.__name__}' from model storage. Resource "

rasa/nlu/extractors/crf_entity_extractor.py

@@ -4,9 +4,9 @@ from collections import OrderedDict
 from enum import Enum
 import logging
 import typing
-from typing import Any, Dict, List, Optional, Text, Tuple, Callable, Type
 
 import numpy as np
+from typing import Any, Dict, List, Optional, Text, Tuple, Callable, Type
 
 import rasa.nlu.utils.bilou_utils as bilou_utils
 import rasa.shared.utils.io
@@ -41,9 +41,6 @@ if typing.TYPE_CHECKING:
     from sklearn_crfsuite import CRF
 
 
-CONFIG_FEATURES = "features"
-
-
 class CRFToken:
     def __init__(
         self,
@@ -63,29 +60,6 @@ class CRFToken:
         self.entity_role_tag = entity_role_tag
         self.entity_group_tag = entity_group_tag
 
-    def to_dict(self) -> Dict[str, Any]:
-        return {
-            "text": self.text,
-            "pos_tag": self.pos_tag,
-            "pattern": self.pattern,
-            "dense_features": [str(x) for x in list(self.dense_features)],
-            "entity_tag": self.entity_tag,
-            "entity_role_tag": self.entity_role_tag,
-            "entity_group_tag": self.entity_group_tag,
-        }
-
-    @classmethod
-    def create_from_dict(cls, data: Dict[str, Any]) -> "CRFToken":
-        return cls(
-            data["text"],
-            data["pos_tag"],
-            data["pattern"],
-            np.array([float(x) for x in data["dense_features"]]),
-            data["entity_tag"],
-            data["entity_role_tag"],
-            data["entity_group_tag"],
-        )
-
 
 class CRFEntityExtractorOptions(str, Enum):
     """Features that can be used for the 'CRFEntityExtractor'."""
@@ -114,6 +88,8 @@ class CRFEntityExtractorOptions(str, Enum):
 class CRFEntityExtractor(GraphComponent, EntityExtractorMixin):
     """Implements conditional random fields (CRF) to do named entity recognition."""
 
+    CONFIG_FEATURES = "features"
+
     function_dict: Dict[Text, Callable[[CRFToken], Any]] = {  # noqa: RUF012
         CRFEntityExtractorOptions.LOW: lambda crf_token: crf_token.text.lower(),
         CRFEntityExtractorOptions.TITLE: lambda crf_token: crf_token.text.istitle(),
@@ -161,7 +137,7 @@ class CRFEntityExtractor(GraphComponent, EntityExtractorMixin):
             # "is the preceding token in title case?"
             # POS features require SpacyTokenizer
             # pattern feature require RegexFeaturizer
-            CONFIG_FEATURES: [
+            CRFEntityExtractor.CONFIG_FEATURES: [
                 [
                     CRFEntityExtractorOptions.LOW,
                     CRFEntityExtractorOptions.TITLE,
@@ -224,7 +200,7 @@ class CRFEntityExtractor(GraphComponent, EntityExtractorMixin):
         )
 
     def _validate_configuration(self) -> None:
-        if len(self.component_config.get(CONFIG_FEATURES, [])) % 2 != 1:
+        if len(self.component_config.get(self.CONFIG_FEATURES, [])) % 2 != 1:
             raise ValueError(
                 "Need an odd number of crf feature lists to have a center word."
             )
@@ -275,11 +251,9 @@ class CRFEntityExtractor(GraphComponent, EntityExtractorMixin):
         ]
         dataset = [self._convert_to_crf_tokens(example) for example in entity_examples]
 
-        self.entity_taggers = self.train_model(
-            dataset, self.component_config, self.crf_order
-        )
+        self._train_model(dataset)
 
-        self.persist(dataset)
+        self.persist()
 
         return self._resource
 
@@ -325,9 +299,7 @@ class CRFEntityExtractor(GraphComponent, EntityExtractorMixin):
             if include_tag_features:
                 self._add_tag_to_crf_token(crf_tokens, predictions)
 
-            features = self._crf_tokens_to_features(
-                crf_tokens, self.component_config, include_tag_features
-            )
+            features = self._crf_tokens_to_features(crf_tokens, include_tag_features)
             predictions[tag_name] = entity_tagger.predict_marginals_single(features)
 
         # convert predictions into a list of tags and a list of confidences
@@ -417,25 +389,27 @@ class CRFEntityExtractor(GraphComponent, EntityExtractorMixin):
         **kwargs: Any,
     ) -> CRFEntityExtractor:
         """Loads trained component (see parent class for full docstring)."""
+        import joblib
+
         try:
+            entity_taggers = OrderedDict()
             with model_storage.read_from(resource) as model_dir:
-                dataset = rasa.shared.utils.io.read_json_file(
-                    model_dir / "crf_dataset.json"
-                )
-                crf_order = rasa.shared.utils.io.read_json_file(
-                    model_dir / "crf_order.json"
-                )
-
-                dataset = [
-                    [CRFToken.create_from_dict(token_data) for token_data in sub_list]
-                    for sub_list in dataset
-                ]
+                # We have to load in the same order as we persisted things as otherwise
+                # the predictions might be off
+                file_names = sorted(model_dir.glob("**/*.pkl"))
+                if not file_names:
+                    logger.debug(
+                        "Failed to load model for 'CRFEntityExtractor'. "
+                        "Maybe you did not provide enough training data and "
+                        "no model was trained."
+                    )
+                    return cls(config, model_storage, resource)
 
-                entity_taggers = cls.train_model(dataset, config, crf_order)
+                for file_name in file_names:
+                    name = file_name.stem[1:]
+                    entity_taggers[name] = joblib.load(file_name)
 
-                entity_extractor = cls(config, model_storage, resource, entity_taggers)
-                entity_extractor.crf_order = crf_order
-                return entity_extractor
+                return cls(config, model_storage, resource, entity_taggers)
         except ValueError:
             logger.warning(
                 f"Failed to load {cls.__name__} from model storage. Resource "
@@ -443,29 +417,23 @@ class CRFEntityExtractor(GraphComponent, EntityExtractorMixin):
             )
             return cls(config, model_storage, resource)
 
-    def persist(self, dataset: List[List[CRFToken]]) -> None:
+    def persist(self) -> None:
         """Persist this model into the passed directory."""
-        with self._model_storage.write_to(self._resource) as model_dir:
-            data_to_store = [
-                [token.to_dict() for token in sub_list] for sub_list in dataset
-            ]
+        import joblib
 
-            rasa.shared.utils.io.dump_obj_as_json_to_file(
-                model_dir / "crf_dataset.json", data_to_store
-            )
-            rasa.shared.utils.io.dump_obj_as_json_to_file(
-                model_dir / "crf_order.json", self.crf_order
-            )
+        with self._model_storage.write_to(self._resource) as model_dir:
+            if self.entity_taggers:
+                for idx, (name, entity_tagger) in enumerate(
+                    self.entity_taggers.items()
+                ):
+                    model_file_name = model_dir / f"{idx}{name}.pkl"
+                    joblib.dump(entity_tagger, model_file_name)
 
-    @classmethod
     def _crf_tokens_to_features(
-        cls,
-        crf_tokens: List[CRFToken],
-        config: Dict[str, Any],
-        include_tag_features: bool = False,
+        self, crf_tokens: List[CRFToken], include_tag_features: bool = False
     ) -> List[Dict[Text, Any]]:
         """Convert the list of tokens into discrete features."""
-        configured_features = config[CONFIG_FEATURES]
+        configured_features = self.component_config[self.CONFIG_FEATURES]
         sentence_features = []
 
         for token_idx in range(len(crf_tokens)):
@@ -476,31 +444,28 @@ class CRFEntityExtractor(GraphComponent, EntityExtractorMixin):
             half_window_size = window_size // 2
             window_range = range(-half_window_size, half_window_size + 1)
 
-            token_features = cls._create_features_for_token(
+            token_features = self._create_features_for_token(
                 crf_tokens,
                 token_idx,
                 half_window_size,
                 window_range,
                 include_tag_features,
-                config,
             )
 
             sentence_features.append(token_features)
 
         return sentence_features
 
-    @classmethod
     def _create_features_for_token(
-        cls,
+        self,
         crf_tokens: List[CRFToken],
         token_idx: int,
         half_window_size: int,
         window_range: range,
         include_tag_features: bool,
-        config: Dict[str, Any],
     ) -> Dict[Text, Any]:
         """Convert a token into discrete features including words before and after."""
-        configured_features = config[CONFIG_FEATURES]
+        configured_features = self.component_config[self.CONFIG_FEATURES]
         prefixes = [str(i) for i in window_range]
 
         token_features = {}
@@ -540,13 +505,13 @@ class CRFEntityExtractor(GraphComponent, EntityExtractorMixin):
                         # set in the training data, 'matched' is either 'True' or
                         # 'False' depending on whether the token actually matches the
                         # pattern or not
-                        regex_patterns = cls.function_dict[feature](token)
+                        regex_patterns = self.function_dict[feature](token)
                         for pattern_name, matched in regex_patterns.items():
                             token_features[f"{prefix}:{feature}:{pattern_name}"] = (
                                 matched
                             )
                     else:
-                        value = cls.function_dict[feature](token)
+                        value = self.function_dict[feature](token)
                         token_features[f"{prefix}:{feature}"] = value
 
         return token_features
@@ -670,46 +635,38 @@ class CRFEntityExtractor(GraphComponent, EntityExtractorMixin):
 
         return tags
 
-    @classmethod
-    def train_model(
-        cls,
-        df_train: List[List[CRFToken]],
-        config: Dict[str, Any],
-        crf_order: List[str],
-    ) -> OrderedDict[str, CRF]:
+    def _train_model(self, df_train: List[List[CRFToken]]) -> None:
         """Train the crf tagger based on the training data."""
         import sklearn_crfsuite
 
-        entity_taggers = OrderedDict()
+        self.entity_taggers = OrderedDict()
 
-        for tag_name in crf_order:
+        for tag_name in self.crf_order:
             logger.debug(f"Training CRF for '{tag_name}'.")
 
             # add entity tag features for second level CRFs
             include_tag_features = tag_name != ENTITY_ATTRIBUTE_TYPE
             X_train = (
-                cls._crf_tokens_to_features(sentence, config, include_tag_features)
+                self._crf_tokens_to_features(sentence, include_tag_features)
                 for sentence in df_train
             )
             y_train = (
-                cls._crf_tokens_to_tags(sentence, tag_name) for sentence in df_train
+                self._crf_tokens_to_tags(sentence, tag_name) for sentence in df_train
             )
 
             entity_tagger = sklearn_crfsuite.CRF(
                 algorithm="lbfgs",
                 # coefficient for L1 penalty
-                c1=config["L1_c"],
+                c1=self.component_config["L1_c"],
                 # coefficient for L2 penalty
-                c2=config["L2_c"],
+                c2=self.component_config["L2_c"],
                 # stop earlier
-                max_iterations=config["max_iterations"],
+                max_iterations=self.component_config["max_iterations"],
                 # include transitions that are possible, but not observed
                 all_possible_transitions=True,
             )
             entity_tagger.fit(X_train, y_train)
 
-            entity_taggers[tag_name] = entity_tagger
+            self.entity_taggers[tag_name] = entity_tagger
 
             logger.debug("Training finished.")
-
-        return entity_taggers

rasa/nlu/featurizers/sparse_featurizer/count_vectors_featurizer.py

@@ -1,32 +1,30 @@
 from __future__ import annotations
-
 import logging
 import re
-from typing import Any, Dict, List, Optional, Text, Tuple, Set, Type, Union
-
-import numpy as np
 import scipy.sparse
-from sklearn.exceptions import NotFittedError
-from sklearn.feature_extraction.text import CountVectorizer
+from typing import Any, Dict, List, Optional, Text, Tuple, Set, Type
+from rasa.nlu.tokenizers.tokenizer import Tokenizer
 
 import rasa.shared.utils.io
 from rasa.engine.graph import GraphComponent, ExecutionContext
 from rasa.engine.recipes.default_recipe import DefaultV1Recipe
 from rasa.engine.storage.resource import Resource
 from rasa.engine.storage.storage import ModelStorage
+from rasa.nlu.featurizers.sparse_featurizer.sparse_featurizer import SparseFeaturizer
+from rasa.nlu.utils.spacy_utils import SpacyModel
+from rasa.shared.constants import DOCS_URL_COMPONENTS
+import rasa.utils.io as io_utils
+from sklearn.exceptions import NotFittedError
+from sklearn.feature_extraction.text import CountVectorizer
+from rasa.shared.nlu.training_data.training_data import TrainingData
+from rasa.shared.nlu.training_data.message import Message
+from rasa.shared.exceptions import RasaException, FileIOException
 from rasa.nlu.constants import (
     TOKENS_NAMES,
     MESSAGE_ATTRIBUTES,
     DENSE_FEATURIZABLE_ATTRIBUTES,
 )
-from rasa.nlu.featurizers.sparse_featurizer.sparse_featurizer import SparseFeaturizer
-from rasa.nlu.tokenizers.tokenizer import Tokenizer
-from rasa.nlu.utils.spacy_utils import SpacyModel
-from rasa.shared.constants import DOCS_URL_COMPONENTS
-from rasa.shared.exceptions import RasaException, FileIOException
 from rasa.shared.nlu.constants import TEXT, INTENT, INTENT_RESPONSE_KEY, ACTION_NAME
-from rasa.shared.nlu.training_data.message import Message
-from rasa.shared.nlu.training_data.training_data import TrainingData
 
 BUFFER_SLOTS_PREFIX = "buf_"
 
@@ -690,31 +688,6 @@ class CountVectorsFeaturizer(SparseFeaturizer, GraphComponent):
         """Check if any model got trained."""
         return any(value is not None for value in attribute_vocabularies.values())
 
-    @staticmethod
-    def convert_vocab(
-        vocab: Dict[str, Union[int, Optional[Dict[str, int]]]], to_int: bool
-    ) -> Dict[str, Union[None, int, np.int64, Dict[str, Union[int, np.int64]]]]:
-        """Converts numpy integers in the vocabulary to Python integers."""
-
-        def convert_value(value: int) -> Union[int, np.int64]:
-            """Helper function to convert a single value based on to_int flag."""
-            return int(value) if to_int else np.int64(value)
-
-        result_dict: Dict[
-            str, Union[None, int, np.int64, Dict[str, Union[int, np.int64]]]
-        ] = {}
-        for key, sub_dict in vocab.items():
-            if isinstance(sub_dict, int):
-                result_dict[key] = convert_value(sub_dict)
-            elif not sub_dict:
-                result_dict[key] = None
-            else:
-                result_dict[key] = {
-                    sub_key: convert_value(value) for sub_key, value in sub_dict.items()
-                }
-
-        return result_dict
-
     def persist(self) -> None:
         """Persist this model into the passed directory.
 
@@ -728,18 +701,17 @@ class CountVectorsFeaturizer(SparseFeaturizer, GraphComponent):
             attribute_vocabularies = self._collect_vectorizer_vocabularies()
             if self._is_any_model_trained(attribute_vocabularies):
                 # Definitely need to persist some vocabularies
-                featurizer_file = model_dir / "vocabularies.json"
+                featurizer_file = model_dir / "vocabularies.pkl"
 
                 # Only persist vocabulary from one attribute if `use_shared_vocab`.
                 # Can be loaded and distributed to all attributes.
-                loaded_vocab = (
+                vocab = (
                     attribute_vocabularies[TEXT]
                     if self.use_shared_vocab
                     else attribute_vocabularies
                 )
-                vocab = self.convert_vocab(loaded_vocab, to_int=True)
 
-                rasa.shared.utils.io.dump_obj_as_json_to_file(featurizer_file, vocab)
+                io_utils.json_pickle(featurizer_file, vocab)
 
                 # Dump OOV words separately as they might have been modified during
                 # training
@@ -814,9 +786,8 @@ class CountVectorsFeaturizer(SparseFeaturizer, GraphComponent):
         """Loads trained component (see parent class for full docstring)."""
         try:
             with model_storage.read_from(resource) as model_dir:
-                featurizer_file = model_dir / "vocabularies.json"
-                vocabulary = rasa.shared.utils.io.read_json_file(featurizer_file)
-                vocabulary = cls.convert_vocab(vocabulary, to_int=False)
+                featurizer_file = model_dir / "vocabularies.pkl"
+                vocabulary = io_utils.json_unpickle(featurizer_file)
 
                 share_vocabulary = config["use_shared_vocab"]