replay-rec 0.20.0__tar.gz → 0.20.0rc0__tar.gz

{replay_rec-0.20.0 → replay_rec-0.20.0rc0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: replay-rec
-Version: 0.20.0
+Version: 0.20.0rc0
 Summary: RecSys Library
 License-Expression: Apache-2.0
 License-File: LICENSE
@@ -14,29 +14,23 @@ Classifier: Intended Audience :: Developers
 Classifier: Intended Audience :: Science/Research
 Classifier: Natural Language :: English
 Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
-Provides-Extra: spark
-Provides-Extra: torch
-Provides-Extra: torch-cpu
-Requires-Dist: lightning (<2.6.0) ; extra == "torch" or extra == "torch-cpu"
-Requires-Dist: lightning ; extra == "torch"
-Requires-Dist: lightning ; extra == "torch-cpu"
+Requires-Dist: d3rlpy (>=2.8.1,<2.9)
+Requires-Dist: implicit (>=0.7.2,<0.8)
+Requires-Dist: lightautoml (>=0.4.1,<0.5)
+Requires-Dist: lightning (>=2.0.2,<=2.4.0)
+Requires-Dist: numba (>=0.50,<1)
 Requires-Dist: numpy (>=1.20.0,<2)
 Requires-Dist: pandas (>=1.3.5,<2.4.0)
 Requires-Dist: polars (<2.0)
-Requires-Dist: psutil (<=7.0.0) ; extra == "spark"
-Requires-Dist: psutil ; extra == "spark"
+Requires-Dist: psutil (<=7.0.0)
 Requires-Dist: pyarrow (<22.0)
-Requires-Dist: pyspark (>=3.0,<3.5) ; extra == "spark"
-Requires-Dist: pyspark ; extra == "spark"
-Requires-Dist: pytorch-optimizer (>=3.8.0,<3.9.0) ; extra == "torch" or extra == "torch-cpu"
-Requires-Dist: pytorch-optimizer ; extra == "torch"
-Requires-Dist: pytorch-optimizer ; extra == "torch-cpu"
+Requires-Dist: pyspark (>=3.0,<3.5)
+Requires-Dist: pytorch-optimizer (>=3.8.0,<4)
+Requires-Dist: sb-obp (>=0.5.10,<0.6)
 Requires-Dist: scikit-learn (>=1.6.1,<1.7.0)
 Requires-Dist: scipy (>=1.13.1,<1.14)
 Requires-Dist: setuptools
-Requires-Dist: torch (>=1.8,<3.0.0) ; extra == "torch" or extra == "torch-cpu"
-Requires-Dist: torch ; extra == "torch"
-Requires-Dist: torch ; extra == "torch-cpu"
+Requires-Dist: torch (>=1.8,<3.0.0)
 Requires-Dist: tqdm (>=4.67,<5)
 Project-URL: Homepage, https://sb-ai-lab.github.io/RePlay/
 Project-URL: Repository, https://github.com/sb-ai-lab/RePlay
@@ -245,6 +239,12 @@ pip install openvino onnx
 pip install hnswlib fixed-install-nmslib
 ```
 
+4) (Experimental) LightFM model support:
+```bash
+pip install ligfhtfm
+```
+> **_NOTE_** : LightFM is not officially supported for Python 3.12 due to discontinued maintenance of the library. If you wish to install it locally, you'll have to use a patched fork of LightFM, such as the [one used internally](https://github.com/daviddavo/lightfm).
+
 
 <a name="examples"></a>
 ## 📑  Resources

{replay_rec-0.20.0 → replay_rec-0.20.0rc0}/README.md

@@ -201,6 +201,12 @@ pip install openvino onnx
 pip install hnswlib fixed-install-nmslib
 ```
 
+4) (Experimental) LightFM model support:
+```bash
+pip install ligfhtfm
+```
+> **_NOTE_** : LightFM is not officially supported for Python 3.12 due to discontinued maintenance of the library. If you wish to install it locally, you'll have to use a patched fork of LightFM, such as the [one used internally](https://github.com/daviddavo/lightfm).
+
 
 <a name="examples"></a>
 ## 📑  Resources

{replay_rec-0.20.0 → replay_rec-0.20.0rc0}/pyproject.toml

@@ -40,19 +40,19 @@ dependencies = [
     "scikit-learn (>=1.6.1,<1.7.0)",
     "pyarrow (<22.0)",
     "tqdm (>=4.67,<5)",
-    "pyspark (>=3.0,<3.5); extra == 'spark'",
-    "psutil (<=7.0.0); extra == 'spark'",
-    "torch (>=1.8, <3.0.0); extra == 'torch' or extra == 'torch-cpu'",
-    "pytorch-optimizer (>=3.8.0,<3.9.0); extra == 'torch' or extra == 'torch-cpu'",
-    "lightning (<2.6.0); extra == 'torch' or extra == 'torch-cpu'",
+    "torch (>=1.8,<3.0.0)",
+    "lightning (>=2.0.2,<=2.4.0)",
+    "pytorch-optimizer (>=3.8.0,<4)",
+    "lightautoml (>=0.4.1,<0.5)",
+    "numba (>=0.50,<1)",
+    "sb-obp (>=0.5.10,<0.6)",
+    "d3rlpy (>=2.8.1,<2.9)",
+    "implicit (>=0.7.2,<0.8)",
+    "pyspark (>=3.0,<3.5)",
+    "psutil (<=7.0.0)",
 ]
 dynamic = ["dependencies"]
-version = "0.20.0"
-
-[project.optional-dependencies]
-spark = ["pyspark", "psutil"]
-torch = ["torch", "pytorch-optimizer", "lightning"]
-torch-cpu = ["torch", "pytorch-optimizer", "lightning"]
+version = "0.20.0.preview"
 
 [project.urls]
 homepage = "https://sb-ai-lab.github.io/RePlay/"
@@ -66,13 +66,6 @@ target-version = ["py39", "py310", "py311", "py312"]
 packages = [{include = "replay"}]
 exclude = [
     "replay/conftest.py",
-    "replay/experimental",
-]
-
-[tool.poetry.dependencies]
-torch = [
-    {markers = "extra == 'torch-cpu' and extra !='torch'", source = "torch-cpu-mirror"},
-    {markers = "extra == 'torch' and extra !='torch-cpu'", source = "PyPI"},
 ]
 
 [tool.poetry.group.dev.dependencies]
@@ -95,11 +88,6 @@ docutils = "0.16"
 data-science-types = "0.2.23"
 filelock = "~3.14.0"
 
-[[tool.poetry.source]]
-name = "torch-cpu-mirror"
-url = "https://download.pytorch.org/whl/cpu"
-priority = "explicit"
-
 [tool.poetry-dynamic-versioning]
 enable = false
 format-jinja = """0.20.0{{ env['PACKAGE_SUFFIX'] }}"""

{replay_rec-0.20.0 → replay_rec-0.20.0rc0}/replay/__init__.py

@@ -4,4 +4,4 @@
 # functionality removed in Python 3.12 is used in downstream packages (like lightfm)
 import setuptools as _
 
-__version__ = "0.20.0"
+__version__ = "0.20.0.preview"

{replay_rec-0.20.0 → replay_rec-0.20.0rc0}/replay/data/dataset.py

@@ -5,8 +5,9 @@
 from __future__ import annotations
 
 import json
+from collections.abc import Iterable, Sequence
 from pathlib import Path
-from typing import Callable, Dict, Iterable, List, Optional, Sequence, Union
+from typing import Callable, Optional, Union
 
 import numpy as np
 from pandas import read_parquet as pd_read_parquet
@@ -315,7 +316,7 @@ class Dataset:
         :returns: Loaded Dataset.
         """
         base_path = Path(path).with_suffix(".replay").resolve()
-        with open(base_path / "init_args.json", "r") as file:
+        with open(base_path / "init_args.json") as file:
             dataset_dict = json.loads(file.read())
 
         if dataframe_type not in ["pandas", "spark", "polars", None]:
@@ -436,14 +437,14 @@ class Dataset:
         )
 
     def _get_feature_source_map(self):
-        self._feature_source_map: Dict[FeatureSource, DataFrameLike] = {
+        self._feature_source_map: dict[FeatureSource, DataFrameLike] = {
             FeatureSource.INTERACTIONS: self.interactions,
             FeatureSource.QUERY_FEATURES: self.query_features,
             FeatureSource.ITEM_FEATURES: self.item_features,
         }
 
     def _get_ids_source_map(self):
-        self._ids_feature_map: Dict[FeatureHint, DataFrameLike] = {
+        self._ids_feature_map: dict[FeatureHint, DataFrameLike] = {
             FeatureHint.QUERY_ID: self.query_features if self.query_features is not None else self.interactions,
             FeatureHint.ITEM_ID: self.item_features if self.item_features is not None else self.interactions,
         }
@@ -499,10 +500,10 @@ class Dataset:
             )
         return FeatureSchema(features_list=features_list + filled_features)
 
-    def _fill_unlabeled_features_sources(self, feature_schema: FeatureSchema) -> List[FeatureInfo]:
+    def _fill_unlabeled_features_sources(self, feature_schema: FeatureSchema) -> list[FeatureInfo]:
         features_list = list(feature_schema.all_features)
 
-        source_mapping: Dict[str, FeatureSource] = {}
+        source_mapping: dict[str, FeatureSource] = {}
         for source in FeatureSource:
             dataframe = self._feature_source_map[source]
             if dataframe is not None:
@@ -524,7 +525,7 @@ class Dataset:
         self._set_cardinality(features_list=features_list)
         return features_list
 
-    def _get_unlabeled_columns(self, source: FeatureSource, feature_schema: FeatureSchema) -> List[FeatureInfo]:
+    def _get_unlabeled_columns(self, source: FeatureSource, feature_schema: FeatureSchema) -> list[FeatureInfo]:
         set_source_dataframe_columns = set(self._feature_source_map[source].columns)
         set_labeled_dataframe_columns = set(feature_schema.columns)
         unlabeled_columns = set_source_dataframe_columns - set_labeled_dataframe_columns
@@ -534,13 +535,13 @@ class Dataset:
         ]
         return unlabeled_features_list
 
-    def _fill_unlabeled_features(self, source: FeatureSource, feature_schema: FeatureSchema) -> List[FeatureInfo]:
+    def _fill_unlabeled_features(self, source: FeatureSource, feature_schema: FeatureSchema) -> list[FeatureInfo]:
         unlabeled_columns = self._get_unlabeled_columns(source=source, feature_schema=feature_schema)
         self._set_features_source(feature_list=unlabeled_columns, source=source)
         self._set_cardinality(features_list=unlabeled_columns)
         return unlabeled_columns
 
-    def _set_features_source(self, feature_list: List[FeatureInfo], source: FeatureSource) -> None:
+    def _set_features_source(self, feature_list: list[FeatureInfo], source: FeatureSource) -> None:
         for feature in feature_list:
             feature._set_feature_source(source)
 

{replay_rec-0.20.0 → replay_rec-0.20.0rc0}/replay/data/dataset_utils/dataset_label_encoder.py

@@ -6,7 +6,8 @@ Contains classes for encoding categorical data
 """
 
 import warnings
-from typing import Dict, Iterable, Iterator, Optional, Sequence, Set, Union
+from collections.abc import Iterable, Iterator, Sequence
+from typing import Optional, Union
 
 from replay.data import Dataset, FeatureHint, FeatureSchema, FeatureSource, FeatureType
 from replay.preprocessing import LabelEncoder, LabelEncodingRule, SequenceEncodingRule
@@ -45,9 +46,9 @@ class DatasetLabelEncoder:
         """
         self._handle_unknown_rule = handle_unknown_rule
         self._default_value_rule = default_value_rule
-        self._encoding_rules: Dict[str, LabelEncodingRule] = {}
+        self._encoding_rules: dict[str, LabelEncodingRule] = {}
 
-        self._features_columns: Dict[Union[FeatureHint, FeatureSource], Sequence[str]] = {}
+        self._features_columns: dict[Union[FeatureHint, FeatureSource], Sequence[str]] = {}
 
     def fit(self, dataset: Dataset) -> "DatasetLabelEncoder":
         """
@@ -161,7 +162,7 @@ class DatasetLabelEncoder:
         """
         self._check_if_initialized()
 
-        columns_set: Set[str]
+        columns_set: set[str]
         columns_set = {columns} if isinstance(columns, str) else {*columns}
 
         def get_encoding_rules() -> Iterator[LabelEncodingRule]:

{replay_rec-0.20.0 → replay_rec-0.20.0rc0}/replay/data/nn/schema.py

@@ -1,17 +1,8 @@
+from collections import OrderedDict
+from collections.abc import ItemsView, Iterable, Iterator, KeysView, Mapping, Sequence, ValuesView
 from typing import (
-    Dict,
-    ItemsView,
-    Iterable,
-    Iterator,
-    KeysView,
-    List,
-    Mapping,
     Optional,
-    OrderedDict,
-    Sequence,
-    Set,
     Union,
-    ValuesView,
 )
 
 import torch
@@ -20,7 +11,7 @@ from replay.data import FeatureHint, FeatureSource, FeatureType
 
 # Alias
 TensorMap = Mapping[str, torch.Tensor]
-MutableTensorMap = Dict[str, torch.Tensor]
+MutableTensorMap = dict[str, torch.Tensor]
 
 
 class TensorFeatureSource:
@@ -79,7 +70,7 @@ class TensorFeatureInfo:
         feature_type: FeatureType,
         is_seq: bool = False,
         feature_hint: Optional[FeatureHint] = None,
-        feature_sources: Optional[List[TensorFeatureSource]] = None,
+        feature_sources: Optional[list[TensorFeatureSource]] = None,
         cardinality: Optional[int] = None,
         padding_value: int = 0,
         embedding_dim: Optional[int] = None,
@@ -154,13 +145,13 @@ class TensorFeatureInfo:
         self._feature_hint = hint
 
     @property
-    def feature_sources(self) -> Optional[List[TensorFeatureSource]]:
+    def feature_sources(self) -> Optional[list[TensorFeatureSource]]:
         """
         :returns: List of sources feature came from.
         """
         return self._feature_sources
 
-    def _set_feature_sources(self, sources: List[TensorFeatureSource]) -> None:
+    def _set_feature_sources(self, sources: list[TensorFeatureSource]) -> None:
         self._feature_sources = sources
 
     @property
@@ -276,7 +267,7 @@ class TensorSchema(Mapping[str, TensorFeatureInfo]):
 
         :returns: New tensor schema of given features.
         """
-        features: Set[TensorFeatureInfo] = set()
+        features: set[TensorFeatureInfo] = set()
         for feature_name in features_to_keep:
             features.add(self._tensor_schema[feature_name])
         return TensorSchema(list(features))
@@ -432,7 +423,7 @@ class TensorSchema(Mapping[str, TensorFeatureInfo]):
             return None
         return rating_features.item().name
 
-    def _get_object_args(self) -> Dict:
+    def _get_object_args(self) -> dict:
         """
         Returns list of features represented as dictionaries.
         """
@@ -456,7 +447,7 @@ class TensorSchema(Mapping[str, TensorFeatureInfo]):
         return features
 
     @classmethod
-    def _create_object_by_args(cls, args: Dict) -> "TensorSchema":
+    def _create_object_by_args(cls, args: dict) -> "TensorSchema":
         features_list = []
         for feature_data in args:
             feature_data["feature_sources"] = (

{replay_rec-0.20.0 → replay_rec-0.20.0rc0}/replay/data/nn/sequence_tokenizer.py

@@ -2,8 +2,9 @@ import abc
 import json
 import pickle
 import warnings
+from collections.abc import Sequence
 from pathlib import Path
-from typing import TYPE_CHECKING, Dict, Generic, List, Optional, Sequence, Set, Tuple, TypeVar, Union
+from typing import TYPE_CHECKING, Generic, Optional, TypeVar, Union
 
 import numpy as np
 import polars as pl
@@ -187,7 +188,7 @@ class SequenceTokenizer:
     def _group_dataset(
         self,
         dataset: Dataset,
-    ) -> Tuple[SequenceDataFrameLike, Optional[SequenceDataFrameLike], Optional[SequenceDataFrameLike]]:
+    ) -> tuple[SequenceDataFrameLike, Optional[SequenceDataFrameLike], Optional[SequenceDataFrameLike]]:
         from replay.data.nn.utils import ensure_pandas, groupby_sequences
 
         grouped_interactions = groupby_sequences(
@@ -268,13 +269,13 @@ class SequenceTokenizer:
         tensor_schema: "TensorSchema",
         query_id_column: str,
         item_id_column: str,
-    ) -> Set[str]:
+    ) -> set[str]:
         # We need only features, which related to tensor schema, otherwise feature should
         # be ignored for efficiency reasons. The code below does feature filtering, and
         # keeps features used as a source in tensor schema.
 
         # Query and item IDs are always needed
-        features_subset: List[str] = [
+        features_subset: list[str] = [
             query_id_column,
             item_id_column,
         ]
@@ -303,7 +304,7 @@ class SequenceTokenizer:
                 msg = "All tensor features must have sources defined"
                 raise ValueError(msg)
 
-            source_tables: List[FeatureSource] = [s.source for s in feature_sources]
+            source_tables: list[FeatureSource] = [s.source for s in feature_sources]
 
             unexpected_tables = list(filter(lambda x: not isinstance(x, FeatureSource), source_tables))
             if len(unexpected_tables) > 0:
@@ -327,7 +328,7 @@ class SequenceTokenizer:
         tensor_features_to_keep: Optional[Sequence[str]] = None,
     ) -> None:
         # Check if all source columns specified in tensor schema exist in provided data frames
-        sources_for_tensors: List["TensorFeatureSource"] = []
+        sources_for_tensors: list["TensorFeatureSource"] = []
         for tensor_feature_name, tensor_feature in tensor_schema.items():
             if tensor_features_to_keep is not None and tensor_feature_name not in tensor_features_to_keep:
                 continue
@@ -421,7 +422,7 @@ class SequenceTokenizer:
 
         if not use_pickle:
             base_path = Path(path).with_suffix(".replay").resolve()
-            with open(base_path / "init_args.json", "r") as file:
+            with open(base_path / "init_args.json") as file:
                 tokenizer_dict = json.loads(file.read())
 
             # load tensor_schema, tensor_features
@@ -625,7 +626,7 @@ class _PandasSequenceProcessor(_BaseSequenceProcessor[PandasDataFrame]):
         """
         :returns: processed Pandas DataFrame with all features from tensor schema.
         """
-        all_features: Dict[str, Union[np.ndarray, List[np.ndarray]]] = {}
+        all_features: dict[str, Union[np.ndarray, list[np.ndarray]]] = {}
         all_features[self._query_id_column] = self._grouped_interactions[self._query_id_column].values
 
         for tensor_feature_name in self._tensor_schema:
@@ -635,7 +636,7 @@ class _PandasSequenceProcessor(_BaseSequenceProcessor[PandasDataFrame]):
 
     def _process_num_interaction_feature(
         self, tensor_feature: "TensorFeatureInfo"
-    ) -> Union[List[np.ndarray], List[List]]:
+    ) -> Union[list[np.ndarray], list[list]]:
         """
         Process numerical interaction feature.
 
@@ -656,7 +657,7 @@ class _PandasSequenceProcessor(_BaseSequenceProcessor[PandasDataFrame]):
                 values.append(np.array(sequence))
         return values
 
-    def _process_num_item_feature(self, tensor_feature: "TensorFeatureInfo") -> Union[List[np.ndarray], List[List]]:
+    def _process_num_item_feature(self, tensor_feature: "TensorFeatureInfo") -> Union[list[np.ndarray], list[list]]:
         """
         Process numerical feature from item features dataset.
 
@@ -682,7 +683,7 @@ class _PandasSequenceProcessor(_BaseSequenceProcessor[PandasDataFrame]):
 
         return values
 
-    def _process_num_query_feature(self, tensor_feature: "TensorFeatureInfo") -> List[np.ndarray]:
+    def _process_num_query_feature(self, tensor_feature: "TensorFeatureInfo") -> list[np.ndarray]:
         """
         Process numerical feature from query features dataset.
 
@@ -694,7 +695,7 @@ class _PandasSequenceProcessor(_BaseSequenceProcessor[PandasDataFrame]):
 
     def _process_cat_interaction_feature(
         self, tensor_feature: "TensorFeatureInfo"
-    ) -> Union[List[np.ndarray], List[List]]:
+    ) -> Union[list[np.ndarray], list[list]]:
         """
         Process categorical interaction feature.
 
@@ -715,7 +716,7 @@ class _PandasSequenceProcessor(_BaseSequenceProcessor[PandasDataFrame]):
                 values.append(np.array(sequence))
         return values
 
-    def _process_cat_query_feature(self, tensor_feature: "TensorFeatureInfo") -> List[np.ndarray]:
+    def _process_cat_query_feature(self, tensor_feature: "TensorFeatureInfo") -> list[np.ndarray]:
         """
         Process categorical feature from query features dataset.
 
@@ -744,7 +745,7 @@ class _PandasSequenceProcessor(_BaseSequenceProcessor[PandasDataFrame]):
                 ]
         return [np.array([query_feature[i]]).reshape(-1) for i in range(len(self._grouped_interactions))]
 
-    def _process_cat_item_feature(self, tensor_feature: "TensorFeatureInfo") -> Union[List[np.ndarray], List[List]]:
+    def _process_cat_item_feature(self, tensor_feature: "TensorFeatureInfo") -> Union[list[np.ndarray], list[list]]:
         """
         Process categorical feature from item features dataset.
 
@@ -760,7 +761,7 @@ class _PandasSequenceProcessor(_BaseSequenceProcessor[PandasDataFrame]):
         assert source is not None
 
         item_feature = self._item_features[source.column]
-        values: List[np.ndarray] = []
+        values: list[np.ndarray] = []
 
         for item_id_sequence in self._grouped_interactions[self._item_id_column]:
             feature_sequence = item_feature.loc[item_id_sequence].values

{replay_rec-0.20.0 → replay_rec-0.20.0rc0}/replay/data/nn/sequential_dataset.py

@@ -1,7 +1,7 @@
 import abc
 import json
 from pathlib import Path
-from typing import TYPE_CHECKING, Tuple, Union
+from typing import TYPE_CHECKING, Union
 
 import numpy as np
 import pandas as pd
@@ -90,7 +90,7 @@ class SequentialDataset(abc.ABC):
     @staticmethod
     def keep_common_query_ids(
         lhs: "SequentialDataset", rhs: "SequentialDataset"
-    ) -> Tuple["SequentialDataset", "SequentialDataset"]:
+    ) -> tuple["SequentialDataset", "SequentialDataset"]:
         """
         Returns `SequentialDatasets` that contain query ids from both datasets.
 
@@ -203,7 +203,7 @@ class PandasSequentialDataset(SequentialDataset):
         from replay.data.nn import TensorSchema
 
         base_path = Path(path).with_suffix(".replay").resolve()
-        with open(base_path / "init_args.json", "r") as file:
+        with open(base_path / "init_args.json") as file:
             sequential_dict = json.loads(file.read())
 
         sequences = pd.read_json(base_path / sequential_dict["init_args"]["sequences_path"])
@@ -287,7 +287,7 @@ class PolarsSequentialDataset(PandasSequentialDataset):
         from replay.data.nn import TensorSchema
 
         base_path = Path(path).with_suffix(".replay").resolve()
-        with open(base_path / "init_args.json", "r") as file:
+        with open(base_path / "init_args.json") as file:
             sequential_dict = json.loads(file.read())
 
         sequences = pl.DataFrame(pd.read_json(base_path / sequential_dict["init_args"]["sequences_path"]))

{replay_rec-0.20.0 → replay_rec-0.20.0rc0}/replay/data/nn/torch_sequential_dataset.py

@@ -1,4 +1,5 @@
-from typing import TYPE_CHECKING, Generator, NamedTuple, Optional, Sequence, Tuple, Union, cast
+from collections.abc import Generator, Sequence
+from typing import TYPE_CHECKING, NamedTuple, Optional, Union, cast
 
 import numpy as np
 import torch
@@ -110,7 +111,7 @@ class TorchSequentialDataset(TorchDataset):
             return sequence
 
         # form shape for padded_sequence. Now supported one and two-dimentions features
-        padded_sequence_shape: Union[Tuple[int, int], Tuple[int]]
+        padded_sequence_shape: Union[tuple[int, int], tuple[int]]
         if len(sequence.shape) == 1:
             padded_sequence_shape = (self._max_sequence_length,)
         elif len(sequence.shape) == 2:
@@ -134,10 +135,10 @@ class TorchSequentialDataset(TorchDataset):
             return torch.float32
         assert False, "Unknown tensor feature type"
 
-    def _build_index2sequence_map(self) -> Sequence[Tuple[int, int]]:
+    def _build_index2sequence_map(self) -> Sequence[tuple[int, int]]:
         return list(self._iter_with_window())
 
-    def _iter_with_window(self) -> Generator[Tuple[int, int], None, None]:
+    def _iter_with_window(self) -> Generator[tuple[int, int], None, None]:
         for i in range(len(self._sequential)):
             actual_seq_len = self._sequential.get_sequence_length(i)
             left_seq_len = actual_seq_len - self._max_sequence_length

{replay_rec-0.20.0 → replay_rec-0.20.0rc0}/replay/data/nn/utils.py

@@ -1,4 +1,5 @@
-from typing import Iterable, Optional
+from collections.abc import Iterable
+from typing import Optional
 
 import polars as pl
 

{replay_rec-0.20.0 → replay_rec-0.20.0rc0}/replay/data/schema.py

@@ -1,18 +1,9 @@
+from collections.abc import ItemsView, Iterable, Iterator, KeysView, Mapping, Sequence, ValuesView
 from enum import Enum
 from typing import (
     Callable,
-    Dict,
-    ItemsView,
-    Iterable,
-    Iterator,
-    KeysView,
-    List,
-    Mapping,
     Optional,
-    Sequence,
-    Set,
     Union,
-    ValuesView,
 )
 
 
@@ -162,7 +153,7 @@ class FeatureSchema(Mapping[str, FeatureInfo]):
             in original schema to keep in subset.
         :returns: new feature schema of given features.
         """
-        features: Set[FeatureInfo] = set()
+        features: set[FeatureInfo] = set()
         for feature_column in features_to_keep:
             if feature_column in self._features_schema:
                 features.add(self._features_schema[feature_column])
@@ -438,7 +429,7 @@ class FeatureSchema(Mapping[str, FeatureInfo]):
         """
         unique_columns = set()
         duplicates = set()
-        item_query_names: Dict[FeatureHint, List[str]] = {
+        item_query_names: dict[FeatureHint, list[str]] = {
             FeatureHint.ITEM_ID: [],
             FeatureHint.QUERY_ID: [],
         }

replay_rec-0.20.0rc0/replay/experimental/metrics/__init__.py

@@ -0,0 +1,62 @@
+"""
+Most metrics require dataframe with recommendations
+and dataframe with ground truth values —
+which objects each user interacted with.
+
+- recommendations (Union[pandas.DataFrame, spark.DataFrame]):
+    predictions of a recommender system,
+    DataFrame with columns ``[user_id, item_id, relevance]``
+- ground_truth (Union[pandas.DataFrame, spark.DataFrame]):
+    test data, DataFrame with columns
+    ``[user_id, item_id, timestamp, relevance]``
+
+Metric is calculated for all users, presented in ``ground_truth``
+for accurate metric calculation in case when the recommender system generated
+recommendation not for all users.  It is assumed, that all users,
+we want to calculate metric for, have positive interactions.
+
+But if we have users, who observed the recommendations, but have not responded,
+those users will be ignored and metric will be overestimated.
+For such case we propose additional optional parameter ``ground_truth_users``,
+the dataframe with all users, which should be considered during the metric calculation.
+
+- ground_truth_users (Optional[Union[pandas.DataFrame, spark.DataFrame]]):
+    full list of users to calculate metric for, DataFrame with ``user_id`` column
+
+Every metric is calculated using top ``K`` items for each user.
+It is also possible to calculate metrics
+using multiple values for ``K`` simultaneously.
+In this case the result will be a dictionary and not a number.
+
+Make sure your recommendations do not contain user-item duplicates
+as duplicates could lead to the wrong calculation results.
+
+- k (Union[Iterable[int], int]):
+    a single number or a list, specifying the
+    truncation length for recommendation list for each user
+
+By default, metrics are averaged by users,
+but you can alternatively use method ``metric.median``.
+Also, you can get the lower bound
+of ``conf_interval`` for a given ``alpha``.
+
+Diversity metrics require extra parameters on initialization stage,
+but do not use ``ground_truth`` parameter.
+
+For each metric, a formula for its calculation is given, because this is
+important for the correct comparison of algorithms, as mentioned in our
+`article <https://arxiv.org/abs/2206.12858>`_.
+"""
+
+from replay.experimental.metrics.base_metric import Metric, NCISMetric
+from replay.experimental.metrics.coverage import Coverage
+from replay.experimental.metrics.hitrate import HitRate
+from replay.experimental.metrics.map import MAP
+from replay.experimental.metrics.mrr import MRR
+from replay.experimental.metrics.ncis_precision import NCISPrecision
+from replay.experimental.metrics.ndcg import NDCG
+from replay.experimental.metrics.precision import Precision
+from replay.experimental.metrics.recall import Recall
+from replay.experimental.metrics.rocauc import RocAuc
+from replay.experimental.metrics.surprisal import Surprisal
+from replay.experimental.metrics.unexpectedness import Unexpectedness