From 69f963244e53ab82573f2895ee21c520078fe4c4 Mon Sep 17 00:00:00 2001
From: JarbasAI <33701864+JarbasAl@users.noreply.github.com>
Date: Wed, 11 Dec 2024 14:06:37 +0000
Subject: [PATCH] performance: paralelize inference (#32)

* performance: paralelize inference

* Update intent_manager.py
---
 ovos_padatious/intent_manager.py | 52 +++++++++++++++++++++++++++-----
 ovos_padatious/simple_intent.py  |  2 +-
 2 files changed, 45 insertions(+), 9 deletions(-)

diff --git a/ovos_padatious/intent_manager.py b/ovos_padatious/intent_manager.py
index fa15588..ddfcb3e 100644
--- a/ovos_padatious/intent_manager.py
+++ b/ovos_padatious/intent_manager.py
@@ -11,22 +11,58 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
-
+import time
+from concurrent.futures import ThreadPoolExecutor
+from typing import List
 from ovos_padatious.intent import Intent
 from ovos_padatious.match_data import MatchData
 from ovos_padatious.training_manager import TrainingManager
 from ovos_padatious.util import tokenize
+from ovos_utils.log import LOG
 
 
 class IntentManager(TrainingManager):
-    def __init__(self, cache):
-        super(IntentManager, self).__init__(Intent, cache)
+    """
+    Manages intents and performs matching using the OVOS Padatious framework.
+
+    Args:
+        cache (str): Path to the cache directory for storing trained models.
+    """
+    def __init__(self, cache: str, debug: bool = False):
+        super().__init__(Intent, cache)
+        self.debug = debug
+
+    def calc_intents(self, query: str, entity_manager) -> List[MatchData]:
+        """
+        Calculate matches for the given query against all registered intents.
+
+        Args:
+            query (str): The input query to match.
+            entity_manager: The entity manager for resolving entities in the query.
 
-    def calc_intents(self, query, entity_manager):
+        Returns:
+            List[MatchData]: A list of matches sorted by confidence.
+        """
         sent = tokenize(query)
         matches = []
-        for i in self.objects:
-            match = i.match(sent, entity_manager)
-            match.detokenize()
-            matches.append(match)
+
+        def match_intent(intent):
+            start_time = time.monotonic()
+            try:
+                match = intent.match(sent, entity_manager)
+                match.detokenize()
+                if self.debug:
+                    LOG.debug(f"Inference for intent '{intent.name}' took {time.monotonic() - start_time} seconds")
+                return match
+            except Exception as e:
+                LOG.error(f"Error processing intent '{intent.name}': {e}")
+                return None
+
+        # Parallelize matching
+        with ThreadPoolExecutor() as executor:
+            matches = list(executor.map(match_intent, self.objects))
+
+        # Filter out None results from failed matches
+        matches = [match for match in matches if match]
+
         return matches
diff --git a/ovos_padatious/simple_intent.py b/ovos_padatious/simple_intent.py
index 1cd97bc..d03117d 100644
--- a/ovos_padatious/simple_intent.py
+++ b/ovos_padatious/simple_intent.py
@@ -118,7 +118,7 @@ def calc_weight(w): return pow(len(w), 3.0)
 
         train_data = fann.training_data()
         train_data.set_train_data(inputs, outputs)
-        LOG.debug(f"Training {self.name} with samples: {n_pos} positive + {n_neg} negative")
+        LOG.debug(f"Training {self.name} with {len(self.ids)} inputs and samples: {n_pos} positive + {n_neg} negative")
         for _ in range(10):
             self.configure_net()
             self.net.train_on_data(train_data, 1000, 0, 0)