fix: add cached LMDB environment to support changes in lmdb 2.0.0

Signed-off-by: Luca Foppiano <luca@foppiano.org>

Author: lfoppiano

delft/sequenceLabelling/wrapper.py

@@ -207,7 +207,7 @@ def __init__(
 
     def get_embedding(self, embedding_name, use_cache=True):
         """Return an Embeddings instance for the given name. Override to customize embedding loading."""
-        return Embeddings(embedding_name, resource_registry=self.registry, use_cache=use_cache)
+        return Embeddings.get_or_create(embedding_name, resource_registry=self.registry, use_cache=use_cache)
 
     def train(
         self,
@@ -807,8 +807,7 @@ def load(self, dir_path="data/models/sequenceLabelling/", weight_file=DEFAULT_WE
 
         if self.model_config.embeddings_name is not None:
             # load embeddings
-            # Do not use cache in 'prediction/production' mode
-            self.embeddings = self.get_embedding(self.model_config.embeddings_name, use_cache=False)
+            self.embeddings = self.get_embedding(self.model_config.embeddings_name, use_cache=True)
             self.model_config.word_embedding_size = self.embeddings.embed_size
         else:
             self.embeddings = None

delft/textClassification/wrapper.py

@@ -155,7 +155,7 @@ def __init__(
 
     def get_embedding(self, embedding_name, use_cache=True):
         """Return an Embeddings instance for the given name. Override to customize embedding loading."""
-        return Embeddings(embedding_name, resource_registry=self.registry, use_cache=use_cache)
+        return Embeddings.get_or_create(embedding_name, resource_registry=self.registry, use_cache=use_cache)
 
     def train(self, x_train, y_train, vocab_init=None, incremental=False, callbacks=None):
 
@@ -565,8 +565,7 @@ def load(self, dir_path="data/models/textClassification/"):
 
         if self.model_config.transformer_name is None:
             # load embeddings
-            # Do not use cache in 'production' mode
-            self.embeddings = self.get_embedding(self.model_config.embeddings_name, use_cache=False)
+            self.embeddings = self.get_embedding(self.model_config.embeddings_name, use_cache=True)
             self.model_config.word_embedding_size = self.embeddings.embed_size
         else:
             self.transformer_name = self.model_config.transformer_name

delft/utilities/Embeddings.py

@@ -38,6 +38,22 @@
 
 
 class Embeddings(object):
+    _cache = {}  # class-level cache: (name, lmdb_path) -> Embeddings instance
+
+    @classmethod
+    def get_or_create(cls, name, resource_registry=None, lang="en", extension="vec", use_cache=True, load=True):
+        if use_cache:
+            lmdb_path = resource_registry.get("embedding-lmdb-path") if resource_registry else None
+            cache_key = (name, lmdb_path)
+            if cache_key in cls._cache:
+                return cls._cache[cache_key]
+            instance = cls(name, resource_registry=resource_registry, lang=lang, extension=extension,
+                           use_cache=use_cache, load=load)
+            cls._cache[cache_key] = instance
+            return instance
+        return cls(name, resource_registry=resource_registry, lang=lang, extension=extension,
+                   use_cache=use_cache, load=load)
+
     def __init__(self, name, resource_registry=None, lang="en", extension="vec", use_cache=True, load=True):
 
         self.name = name