Set dtype=object to create numpy array

In order to create an array from list of inhomogeneus objects explicitly dtype to object Since numpy 1.24 numpy raises ValueError when creating Ragged array numpy/numpy#22004
RasaHQ · Feb 28, 2023 · 0c45413 · 0c45413
1 parent 28eb117
commit 0c45413
Show file tree

Hide file tree

Showing 4 changed files with 18 additions and 12 deletions.
diff --git a/pyproject.toml b/pyproject.toml
@@ -139,7 +139,7 @@ version = ">=1.19.2,<1.22.0"
 python = "~=3.7.0"
 
 [[tool.poetry.dependencies.numpy]]
-version = ">=1.19.2,<1.24.0"
+version = ">=1.19.2,<1.25.0"
 python = ">=3.8,<3.11"
 
 [[tool.poetry.dependencies.scipy]]

diff --git a/rasa/utils/tensorflow/model_data.py b/rasa/utils/tensorflow/model_data.py
@@ -771,7 +771,7 @@ def balanced_data(self, data: Data, batch_size: int, shuffle: bool) -> Data:
                 for f in features:
                     final_data[key][sub_key].append(
                         FeatureArray(
-                            np.concatenate(np.array(f)),
+                            np.concatenate(np.array(f, dtype=object)),
                             number_of_dimensions=f[0].number_of_dimensions,
                         )
                     )

diff --git a/rasa/utils/tensorflow/model_data_utils.py b/rasa/utils/tensorflow/model_data_utils.py
@@ -385,19 +385,19 @@ def _feature_arrays_for_attribute(
     for key, values in _sparse_features.items():
         if consider_dialogue_dimension:
             sparse_features[key] = FeatureArray(
-                np.array(values), number_of_dimensions=4
+                np.array(values, dtype=object), number_of_dimensions=4
             )
         else:
             sparse_features[key] = FeatureArray(
-                np.array([v[0] for v in values]), number_of_dimensions=3
+                np.array([v[0] for v in values], dtype=object), number_of_dimensions=3
             )
 
     for key, values in _dense_features.items():
         if consider_dialogue_dimension:
-            dense_features[key] = FeatureArray(np.array(values), number_of_dimensions=4)
+            dense_features[key] = FeatureArray(np.array(values, dtype=object), number_of_dimensions=4)
         else:
             dense_features[key] = FeatureArray(
-                np.array([v[0] for v in values]), number_of_dimensions=3
+                np.array([v[0] for v in values], dtype=object), number_of_dimensions=3
             )
 
     attribute_to_feature_arrays = {

diff --git a/tests/utils/tensorflow/conftest.py b/tests/utils/tensorflow/conftest.py
@@ -21,7 +21,8 @@ async def model_data() -> RasaModelData:
                                 np.random.rand(3, 14),
                                 np.random.rand(1, 14),
                                 np.random.rand(3, 14),
-                            ]
+                            ],
+                            dtype=object,
                         ),
                         number_of_dimensions=3,
                     ),
@@ -43,7 +44,8 @@ async def model_data() -> RasaModelData:
                                 scipy.sparse.csr_matrix(
                                     np.random.randint(5, size=(3, 10))
                                 ),
-                            ]
+                            ],
+                            dtype=object,
                         ),
                         number_of_dimensions=3,
                     ),
@@ -106,7 +108,8 @@ async def model_data() -> RasaModelData:
                                         np.random.randint(5, size=(7, 10))
                                     ),
                                 ],
-                            ]
+                            ],
+                            dtype=object,
                         ),
                         number_of_dimensions=4,
                     ),
@@ -132,7 +135,8 @@ async def model_data() -> RasaModelData:
                                     np.random.rand(1, 14),
                                     np.random.rand(7, 14),
                                 ],
-                            ]
+                            ],
+                            dtype=object,
                         ),
                         number_of_dimensions=4,
                     ),
@@ -148,7 +152,8 @@ async def model_data() -> RasaModelData:
                                 np.random.randint(2, size=(3, 10)),
                                 np.random.randint(2, size=(1, 10)),
                                 np.random.randint(2, size=(3, 10)),
-                            ]
+                            ],
+                            dtype=object,
                         ),
                         number_of_dimensions=3,
                     )
@@ -167,7 +172,8 @@ async def model_data() -> RasaModelData:
                                 np.array([[0], [1], [1]]),
                                 np.array([[0], [1]]),
                                 np.array([[0], [0], [0]]),
-                            ]
+                            ],
+                            dtype=object,
                         ),
                         number_of_dimensions=3,
                     )