feat: use singleton instead of global (#429)

- where previously global was used now they are singleton class
variables
- this avoids issues with multi-threading env

Author: badGarnet

CHANGELOG.md

@@ -1,3 +1,7 @@
+## 1.0.4
+
+* feat: use singleton instead of `global` to store shared variables
+
 ## 1.0.3
 
 * setting longest_edge=1333 to the table image processor

unstructured_inference/__version__.py

@@ -1 +1 @@
-__version__ = "1.0.3"  # pragma: no cover
+__version__ = "1.0.4"  # pragma: no cover

unstructured_inference/models/base.py

@@ -15,7 +15,28 @@
 
 DEFAULT_MODEL = "yolox"
 
-models: Dict[str, UnstructuredModel] = {}
+
+class Models(object):
+    _instance = None
+
+    def __new__(cls):
+        """return an instance if one already exists otherwise create an instance"""
+        if cls._instance is None:
+            cls._instance = super(Models, cls).__new__(cls)
+            cls.models: Dict[str, UnstructuredModel] = {}
+        return cls._instance
+
+    def __contains__(self, key):
+        return key in self.models
+
+    def __getitem__(self, key: str):
+        return self.models.__getitem__(key)
+
+    def __setitem__(self, key: str, value: UnstructuredModel):
+        self.models[key] = value
+
+
+models: Models = Models()
 
 
 def get_default_model_mappings() -> Tuple[
@@ -46,8 +67,6 @@ def get_model(model_name: Optional[str] = None) -> UnstructuredModel:
     # TODO(alan): These cases are similar enough that we can probably do them all together with
     # importlib
 
-    global models  # noqa
-
     if model_name is None:
         default_name_from_env = os.environ.get("UNSTRUCTURED_DEFAULT_MODEL_NAME")
         model_name = default_name_from_env if default_name_from_env is not None else DEFAULT_MODEL

unstructured_inference/models/tables.py

@@ -27,9 +27,18 @@
 class UnstructuredTableTransformerModel(UnstructuredModel):
     """Unstructured model wrapper for table-transformer."""
 
+    _instance = None
+
     def __init__(self):
         pass
 
+    @classmethod
+    def instance(cls):
+        """return an instance if one already exists otherwise create an instance"""
+        if cls._instance is None:
+            cls._instance = cls.__new__(cls)
+        return cls._instance
+
     def predict(
         self,
         x: PILImage.Image,
@@ -72,7 +81,8 @@ def initialize(
             cached_current_verbosity = logging.get_verbosity()
             logging.set_verbosity_error()
             self.model = TableTransformerForObjectDetection.from_pretrained(
-                model, device_map=self.device
+                model,
+                device_map=self.device,
             )
             logging.set_verbosity(cached_current_verbosity)
             self.model.eval()
@@ -135,12 +145,11 @@ def run_prediction(
         return prediction
 
 
-tables_agent: UnstructuredTableTransformerModel = UnstructuredTableTransformerModel()
+tables_agent: UnstructuredTableTransformerModel = UnstructuredTableTransformerModel.instance()
 
 
 def load_agent():
-    """Loads the Table agent as a global variable to ensure that we only load it once."""
-    global tables_agent  # noqa
+    """Loads the Table agent."""
 
     if not hasattr(tables_agent, "model"):
         logger.info("Loading the Table agent ...")