pr fixes

rectools/dataset/dataset.py

@@ -31,20 +31,25 @@
 from .identifiers import ExternalId, IdMap
 from .interactions import Interactions
 
-DenseOrSparseFeatureName = tp.Union[str, SparseFeatureName]
+AnyFeatureName = tp.Union[str, SparseFeatureName]
 
 
-def _serialize_feature_name(spec: DenseOrSparseFeatureName) -> Hashable:
+def _serialize_feature_name(spec: tp.Union[AnyFeatureName, tp.Any]) -> Hashable:
     if isinstance(spec, tuple):
         return tuple(_serialize_feature_name(item) for item in spec)
-    if isinstance(spec, (int, float, str)):
+    if isinstance(spec, (int, float, str, bool)):
         return spec
-    if np.issubdtype(spec, np.number):  # type:ignore[unreachable]
+    if not isinstance(spec, np.ndarray) and np.issubdtype(spec, np.number) or np.issubdtype(spec, np.bool_):
         return spec.item()
-    return "unsupported feature name"
+    raise ValueError(
+        f"""
+        Serialization for feature name {spec} is not supported.
+        Please convert your feature names and values to strings, numbers, booleans or their tuples or lists.
+    """
+    )
 
 
-FeatureName = tpe.Annotated[DenseOrSparseFeatureName, PlainSerializer(_serialize_feature_name, when_used="json")]
+FeatureName = tpe.Annotated[AnyFeatureName, PlainSerializer(_serialize_feature_name, when_used="json")]
 DatasetSchemaDict = tp.Dict[str, tp.Any]
 
 

rectools/models/nn/bert4rec.py

@@ -169,8 +169,8 @@ class BERT4RecModel(TransformerModelBase[BERT4RecModelConfig]):
     BERT4Rec model: transformer-based sequential model with bidirectional attention mechanism and
     "MLM" (masked item in user sequence) training objective.
     Our implementation covers multiple loss functions and a variable number of negatives for them.
-    
-    Notes
+
+    References
     ----------
     Transformers tutorial: https://rectools.readthedocs.io/en/stable/examples/tutorials/transformers_tutorial.html
     Advanced training guide:
@@ -217,8 +217,6 @@ class BERT4RecModel(TransformerModelBase[BERT4RecModelConfig]):
         Verbosity level.
         Enables progress bar, model summary and logging in default lightning trainer when set to a
         positive integer.
-        Enables automatic lightning checkpointing when set to 100 or higher. This will save the most
-        the most recent model to a single checkpoint after each epoch.
         Will be omitted if `get_trainer_func` is specified.
     dataloader_num_workers : int, default 0
         Number of loader worker processes.

rectools/models/nn/sasrec.py

@@ -201,7 +201,7 @@ class SASRecModel(TransformerModelBase[SASRecModelConfig]):
     "Shifted Sequence" training objective.
     Our implementation covers multiple loss functions and a variable number of negatives for them.
 
-    Notes
+    References
     ----------
     Transformers tutorial: https://rectools.readthedocs.io/en/stable/examples/tutorials/transformers_tutorial.html
     Advanced training guide:
@@ -246,8 +246,6 @@ class SASRecModel(TransformerModelBase[SASRecModelConfig]):
         Verbosity level.
         Enables progress bar, model summary and logging in default lightning trainer when set to a
         positive integer.
-        Enables automatic lightning checkpointing when set to 100 or higher. This will save the most
-        the most recent model to a single checkpoint after each epoch.
         Will be omitted if `get_trainer_func` is specified.
     dataloader_num_workers : int, default 0
         Number of loader worker processes.

rectools/models/nn/transformer_base.py

@@ -730,7 +730,7 @@ def _init_trainer(self) -> None:
                 enable_progress_bar=self.verbose > 0,
                 enable_model_summary=self.verbose > 0,
                 logger=self.verbose > 0,
-                enable_checkpointing=self.verbose > 99,
+                enable_checkpointing=False,
                 devices=1,
             )
         else:
@@ -931,12 +931,12 @@ def _model_from_checkpoint(cls, checkpoint: tp.Dict[str, tp.Any]) -> tpe.Self:
 
     def __getstate__(self) -> object:
         if self.is_fitted:
+            if self.fit_trainer is None:
+                raise RuntimeError("Model that was loaded from checkpoint cannot be saved without being fitted again")
             with NamedTemporaryFile() as f:
-                if self.fit_trainer is None:
-                    raise TypeError("Model that was loaded from checkpoint cannot be saved without being fitted again")
                 self.fit_trainer.save_checkpoint(f.name)
                 checkpoint = Path(f.name).read_bytes()
-                state: tp.Dict[str, tp.Any] = {"fitted_checkpoint": checkpoint}
+            state: tp.Dict[str, tp.Any] = {"fitted_checkpoint": checkpoint}
             return state
         state = {"model_config": self.get_config()}
         return state