Version: Increased supported version of numpy and pytables

Author: milankures

cesnet_tszoo/configs/base_config.py

@@ -451,7 +451,7 @@ def _set_default_values(self, default_values: dict[str, Number]) -> None:
 
         elif self.default_values is None or math.isnan(self.default_values) or np.isnan(self.default_values):
             # If default_values is None or NaN, assign NaN to each feature
-            self.default_values = {feature: np.NaN for feature in self.features_to_take_without_ids}
+            self.default_values = {feature: np.nan for feature in self.features_to_take_without_ids}
             self.logger.debug("Assigned NaN as the default value for all features because 'default_values' is None or NaN.")
 
         # Convert the default values into a NumPy array for consistent data handling

cesnet_tszoo/pytables_data/base_datasets/base_dataset.py

@@ -71,7 +71,7 @@ def cleanup(self) -> None:
     def load_data_from_table(self, ts_row_ranges_to_take: np.ndarray, time_indices_to_take: np.ndarray, fillers_to_use: np.ndarray[Filler] | None) -> np.ndarray:
         """Return data from table. Missing values are filled with `fillers_to_use` and `default_values`. """
 
-        result = np.full((len(ts_row_ranges_to_take), len(time_indices_to_take), len(self.features_to_take)), fill_value=np.NaN, dtype=np.float64)
+        result = np.full((len(ts_row_ranges_to_take), len(time_indices_to_take), len(self.features_to_take)), fill_value=np.nan, dtype=np.float64)
         result[:, :, self.offset_exclude_feature_ids:] = self.default_values
 
         full_missing_indices = np.arange(0, len(time_indices_to_take))
@@ -119,7 +119,7 @@ def load_data_from_table(self, ts_row_ranges_to_take: np.ndarray, time_indices_t
 
             # Get valid times
             filtered_rows = rows[mask].view()
-            filtered_rows[ID_TIME_COLUMN_NAME] -= first_time_index
+            filtered_rows[ID_TIME_COLUMN_NAME] = filtered_rows[ID_TIME_COLUMN_NAME] - first_time_index
             existing_indices = filtered_rows[ID_TIME_COLUMN_NAME].view()
 
             missing_values_mask = np.ones(len(time_indices_to_take), dtype=bool)

cesnet_tszoo/pytables_data/base_datasets/initializer_dataset.py

@@ -54,7 +54,7 @@ def cleanup(self) -> None:
     def load_data_from_table(self, identifier_row_range_to_take: np.ndarray, idx: int) -> np.ndarray:
         """Returns data from table. Missing values are filled fillers and `default_values`. Prepares fillers."""
 
-        result = np.full((len(self.time_period), len(self.features_to_take)), fill_value=np.NaN, dtype=np.float64)
+        result = np.full((len(self.time_period), len(self.features_to_take)), fill_value=np.nan, dtype=np.float64)
         result[:, self.offset_exclude_feature_ids:] = self.default_values
 
         expected_offset = np.uint32(len(self.time_period))
@@ -95,7 +95,7 @@ def load_data_from_table(self, identifier_row_range_to_take: np.ndarray, idx: in
 
         # Get valid times
         filtered_rows = rows[mask].view()
-        filtered_rows[ID_TIME_COLUMN_NAME] -= first_time_index
+        filtered_rows[ID_TIME_COLUMN_NAME] = filtered_rows[ID_TIME_COLUMN_NAME] - first_time_index
         existing_indices = filtered_rows[ID_TIME_COLUMN_NAME].view()
 
         missing_values_mask = np.ones(len(self.time_period), dtype=bool)

cesnet_tszoo/utils/scaler.py

@@ -28,7 +28,7 @@ def partial_fit(self, data: np.ndarray) -> None:
             def transform(self, data: np.ndarray):
                 log_data = np.ma.log(data)
 
-                return log_data.filled(np.NaN)
+                return log_data.filled(np.nan)
     """
 
     @abstractmethod
@@ -133,7 +133,7 @@ def transform(self, data: np.ndarray):
 
 class LogScaler(Scaler):
     """
-    Tranforms data with natural logarithm. Zero or invalid values are set to `np.NaN`.
+    Tranforms data with natural logarithm. Zero or invalid values are set to `np.nan`.
 
     Corresponds to enum [`ScalerType.LOG_SCALER`][cesnet_tszoo.utils.enums.ScalerType] or literal `log_scaler`.
     """
@@ -147,7 +147,7 @@ def partial_fit(self, data: np.ndarray) -> None:
     def transform(self, data: np.ndarray):
         log_data = np.ma.log(data)
 
-        return log_data.filled(np.NaN)
+        return log_data.filled(np.nan)
 
 
 class L2Normalizer(Scaler):

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "cesnet_tszoo"
-version = "0.1.1"
+version = "0.1.2"
 authors = [
     { name = "Milan Kureš", email = "[email protected]" },
     { name = "Josef Koumar", email = "[email protected]" },
@@ -33,13 +33,13 @@ classifiers = [
 ]
 requires-python = ">=3.10"
 dependencies = [
-    "numpy<2.0",
+    "numpy>= 2.0, ==2.*",
     "pandas",
     "scikit-learn",
-    "tables>=3.8.0,<=3.9.2",
-    "torch>=1.10",
+    "tables >= 3.10.0",
+    "torch >= 1.10",
     "tqdm",
-    "nbformat>=4.2.0",
+    "nbformat >= 4.2.0",
     "plotly",
     "PyYAML",
     "requests",