AbstractIntegratedModule 0.5.0__tar.gz → 0.5.1__tar.gz

{abstractintegratedmodule-0.5.0 → abstractintegratedmodule-0.5.1/AbstractIntegratedModule.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: AbstractIntegratedModule
-Version: 0.5.0
+Version: 0.5.1
 Summary: Framework for Advanced Integrated Non-LLM AI Module library - Backend Framework For Non-LLM AI Agent
 Author: Micro-Novelty
 Author-email: hernikpuspita5@gmail.com
@@ -42,7 +42,7 @@ https://github.com/Micro-Novelty/IntegratedPipeline-Specialized-Non-LLM-AI-Agent
 #### Note: The README here you are reading is a direct copy from my README Repository, to download the necessary files, you can visit my Repository with the provided link above.
 
 ### Library Short Description:
-- Development Stage: Beta, 0.5.0.
+- Development Stage: Beta, 0.5.1.
 - Maintainer: Micro-Novelty.
 - library Source-Code is Open-sourced on github.
 - Purpose: Specifically Designed for providing Non-LLM AI Agent Framework for edge Devices, Optimized for ARM64 architecture.
@@ -64,6 +64,11 @@ https://github.com/Micro-Novelty/IntegratedPipeline-Specialized-Non-LLM-AI-Agent
    - AWE setup Proven Efficient on Hard-uncontrolled dataset such as Activity Recognition from the given Database.
    - LSTM is Optimized efficiently for scarce data with AWE method.
    - Robust Advanced prediction capabilities proven effective on ARM64 Using MLP + LSTM Architectures.
+- Changelog:
+     - v0.5.1:
+         [=] New features: 
+          - Added new dynamic gate for Transformer Fixed and dynamic switching condition.
+          - Added Capabilities for the Model to save the Transformer weights as binaries in local SQlite database.
 -----
 
 <img width="1280" height="600" alt="WhatsApp Image 2026-05-27 at 07 16 32" src="https://github.com/user-attachments/assets/4b58a556-45a3-419b-96fd-9c1b76cac574" />

{abstractintegratedmodule-0.5.0 → abstractintegratedmodule-0.5.1}/AbstractIntegratedModule.py

@@ -40,7 +40,7 @@ import hmac
 import aiohttp
 import psutil
 from sklearn.preprocessing import StandardScaler
-
+import io
 
 # initial Setup logging for AgentDistributedInference and ModelStorage class logger and security logger
 logger = logging.getLogger(__name__)
@@ -3942,15 +3942,13 @@ class ModelStorage:
                 print('|| Skipping Database Modification...')
                 pass
             c = conn.cursor()
-            c.execute('''CREATE TABLE IF NOT EXISTS agent_attn_storage
-                     (id INTEGER PRIMARY KEY AUTOINCREMENT,
+            c.execute('''CREATE TABLE IF NOT EXISTS weight_storage
+                      (id INTEGER PRIMARY KEY AUTOINCREMENT,
                       memory_name TEXT,
                       model_type TEXT,
-                      model_attn_data TEXT,
-                      model_target_pred TEXT,
-                      agent_id TEXT,
+                      weights TEXT,
                       is_active INTEGER DEFAULT 0,
-                      created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP)''')
+                      created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP)''') 
         
         
             conn.commit()
@@ -4352,16 +4350,142 @@ class ModelStorage:
                 SET is_active = 0 
                 WHERE memory_name = ? AND id != last_insert_rowid()
             """, (memory_name,)) 
+            
+            c.execute("""
+                DELETE FROM weight_storage
+                WHERE memory_name = ?
+                AND model_type = 'Pipeline'
+                AND is_active = 0
+            """, (memory_name,))            
 
             conn.commit()
             conn.close()
 
-            print('[||] Weights dictionary saved!')
+            self.save_transformer_weights(memory_name)
+            print('[||] All Weights dictionary saved!')
 
         except Exception as e:
             print(f'[-] Cant save Weights due to: {e}')
             pass          
-       
+
+
+    def save_transformer_weights(self, memory_name: str):
+        """Save transformer weights as compressed binary blob."""
+        tf = self.pipeline.model2
+
+        try:
+            db_path = self.get_database_path()
+            conn = sqlite3.connect(db_path)
+        except:
+            conn = sqlite3.connect(self.db_path)
+
+        c = conn.cursor()
+
+        buf = io.BytesIO()
+        np.savez_compressed(buf,
+            token_embedding = tf.token_embedding,
+            pos_embedding   = tf.pos_embedding,
+            W_q             = tf.W_q,
+            W_k             = tf.W_k,
+            W_v             = tf.W_v,
+            W_q_fixed       = tf.W_q_fixed,
+            W_k_fixed       = tf.W_k_fixed,
+            W_v_fixed       = tf.W_v_fixed,
+            W_o             = tf.W_o,
+            ffn1            = tf.ffn1,
+            ffn2            = tf.ffn2,
+            ln1_scale       = tf.ln1_scale,
+            ln1_shift       = tf.ln1_shift,
+            ln2_scale       = tf.ln2_scale,
+            ln2_shift       = tf.ln2_shift,
+            output          = tf.output,
+            output_bias     = tf.output_bias
+        )
+        binary_data = buf.getvalue()
+
+        try:
+            c.execute("""
+                INSERT INTO weight_storage
+                (memory_name, model_type, weights, is_active)
+                VALUES (?, ?, ?, ?)
+            """, (memory_name, 'transformer', 
+                sqlite3.Binary(binary_data), 1))
+
+            c.execute("""
+                UPDATE weight_storage SET is_active = 0
+                WHERE memory_name = ?
+                AND model_type = 'transformer'
+                AND id != last_insert_rowid()
+            """, (memory_name,))
+
+            c.execute("""
+                DELETE FROM weight_storage
+                WHERE memory_name = ?
+                AND model_type = 'transformer'
+                AND is_active = 0
+            """, (memory_name,))
+
+            conn.commit()
+            print('[||] Transformer weights saved!')
+
+        except Exception as e:
+            print(f'[!] Transformer weight save failed: {e}')
+            conn.rollback()
+        finally:
+            conn.close()
+
+    def load_transformer_weights(self, memory_name: str) -> bool:
+        try:
+            db_path = self.get_database_path()
+            conn = sqlite3.connect(db_path)
+        except:
+            conn = sqlite3.connect(self.db_path)
+
+        c = conn.cursor()   
+
+        try:
+            c.execute("""
+                SELECT weights FROM weight_storage
+                WHERE memory_name = ? AND model_type = 'transformer' AND is_active = 1
+                ORDER BY id DESC LIMIT 1
+            """, (memory_name,))
+            row = c.fetchone()
+            if not row:
+                print(f'[=] No saved transformer weights for {memory_name}')
+                return False
+
+            buf = io.BytesIO(bytes(row[0]))
+            data = np.load(buf, allow_pickle=False)
+
+            t = self.pipeline.model2
+            t.token_embedding = data['token_embedding']
+            t.pos_embedding   = data['pos_embedding']
+            t.W_q             = data['W_q']
+            t.W_k             = data['W_k']
+            t.W_v             = data['W_v']
+            t.W_q_fixed       = data['W_q_fixed']
+            t.W_k_fixed       = data['W_k_fixed']
+            t.W_v_fixed       = data['W_v_fixed']
+            t.W_o             = data['W_o']
+            t.ffn1            = data['ffn1']
+            t.ffn2            = data['ffn2']
+            t.ln1_scale       = data['ln1_scale']
+            t.ln1_shift       = data['ln1_shift']
+            t.ln2_scale       = data['ln2_scale']
+            t.ln2_shift       = data['ln2_shift']
+            t.output          = data['output']
+            t.output_bias     = data['output_bias']
+
+            print(f'[||] Transformer weights loaded!')
+            return True
+
+        except Exception as e:
+            print(f'[!] Transformer weight load failed: {e}')
+            return False
+        finally:
+            conn.close()
+
+
     def load_weights(self, memory_name):
         """Load weights from database. Returns True if found."""
         result = self.weight_retrieval(memory_name)
@@ -4381,9 +4505,14 @@ class ModelStorage:
             self.pipeline.lstm_engine.n_samples     = weights.get('n_samples', self.pipeline.lstm_engine.n_samples)
             self.pipeline.lstm_engine.quantiles     = {float(k): tuple(v) 
                                 for k, v in weights.get('quantiles', {}).items()}
-        
-            print(f'[=] All Weights loaded for {memory_name} '
-                f'(saved at {weights.get("saved_at", "unknown")})')
+
+            tf_loaded = self.load_transformer_weights(memory_name)
+
+            print(f'[=] Transformer weights loaded: {tf_loaded}')
+            if tf_loaded:
+                print(f'[=] All Weights loaded for {memory_name}  '
+                    f'(saved at {weights.get("saved_at", "unknown")})')
+
         except Exception as e:
             print(f'[!] Cant load any Weights due to: {e}')
             traceback.print_exc()
@@ -7305,6 +7434,7 @@ class IntegratedPipeline:
         self.hidden = 32
         self.output_size = 1
         self.dropout_rate = 0.1
+        self.transformer_training_epochs = 100
 
         # Main component setup
         self.standard_scaler = StandardScaler()
@@ -7365,6 +7495,7 @@ class IntegratedPipeline:
         self.temporary_id = []
 
         self.final_conf_score = 0.0
+        self.timeout = 120
         self.confidence_threshold = 0.45  
         self.peer_assistance_threshold = 0.0              
         self.agent_id = random.randint(0, 10000)
@@ -9501,16 +9632,30 @@ class IntegratedPipeline:
 
         if not unsuitable_training:
             print(f'🚀 Training Transformer with {len(sequence_inputs)} Samples: ')
-            conditional_anisotropy = self.anisotropy_measurement(sequence_inputs)
-            if conditional_anisotropy >= self.confidence_threshold: 
-                print('[+] Dynamic Backward')
+
+            x_conditional_anisotropy = self.anisotropy_measurement(sequence_inputs)
+            s_conditional_anisotropy = self.anisotropy_measurement(X_raw)
+
+            AME_x = self.AME_Encoder(X_raw)
+            AME_s = self.AME_Encoder(sequence_inputs)
+            AMR_x = 1.0 / (1.0 + np.exp(-AME_x))
+            AMR_s = 1.0 / (1.0 + np.exp(-AME_s))
+
+            AMR_ratio = AMR_x / (AMR_s + min_signal)
+            anisotropy_ratio = x_conditional_anisotropy / (s_conditional_anisotropy + min_signal)
+
+            dynamic_complex_environment = (anisotropy_ratio < 0.5 and 
+                                             AMR_ratio < 0.5)                           
+                                               
+            if dynamic_complex_environment: 
+                print('[+] Dynamic Backward for Transformer Initiated')
                 mode = 'dynamic_backward'
             else:
-                print('[-] Fixed Backward')
+                print('[=] Fixed Backward for Transformer initiated')
                 mode = 'fixed_backward'
 
             if self.use_transformer:
-                self.model2.train(sequence_inputs, y_true, epochs=100, mode=mode, lr=lr, embedded=True, batch_size=batch_size)
+                self.model2.train(sequence_inputs, y_true, epochs=self.transformer_training_epochs, mode=mode, lr=lr, embedded=True, batch_size=batch_size)
 
             X_raw_generation, y, n_classes, input_dim = self.mlp_training_features(rules, datasets)
             X_raw_features = self.tfidf.transform(X_raw_generation).toarray()
@@ -9576,7 +9721,7 @@ class IntegratedPipeline:
 
             if self.lstm_engine:
                 self.storage.save_weights(self.memory_name, model_type='Pipeline') 
-
+                
             print('🎉 All Model Trained!')
         else:
             print(f'[=] No suitable condition for training!')
@@ -11015,7 +11160,9 @@ class PipelineAsyncManager:
 
         except asyncio.TimeoutError:
             raise FutureTimeoutError(f"[-] Advanced prediction timed out after {timeout}s")
-            
+        except Exception as e:
+            print(f'[!] Error in asynchronous advanced prediction: {e}')
+                     
     # ============ ADMIN FUNCTIONS (with authentication) ============
     
     def _initialize_bootstrap_security(self):
@@ -11233,7 +11380,7 @@ class PipelineAsyncManager:
             if not self.pipeline.intents:
                 advanced_result = self.predict(
                 text, 
-                timeout=120,
+                timeout=self.pipeline.timeout,
                 retries=None,
                 api_key=api_key,
                 client_ip=client_ip,
@@ -11247,7 +11394,7 @@ class PipelineAsyncManager:
             print('[=] Initiating Batch prediction for multiple texts...')
             results = self.predict_batch(
                 texts=texts,
-                timeout=120,
+                timeout=self.pipeline.timeout,
                 api_key=api_key)
 
             print("========📊 PREDICTION RESULTS============")
@@ -13334,7 +13481,7 @@ class CohesiveAgentDeployment:
         # This runs in a thread pool via asyncio.to_thread
         return self.async_manager.predict(
             texts=texts,
-            timeout=120,
+            timeout=self.pipeline.timeout,
             retries=None,
             api_key=api_key,
             client_ip=client_ip,
@@ -13667,7 +13814,7 @@ class CohesiveAgentDeployment:
 
             result = await asyncio.wait_for(
                 self.predict_with_peers(texts, api_key, method, disable_sync=disable_sync),
-                timeout=120.0
+                timeout=self.pipeline.timeout
             )
             
             # Check if result is valid
@@ -14543,7 +14690,7 @@ def initiate_prediction_usage(pipeline, manager, predict_wrapper, test_titles, l
         texts = {'test_titles': test_titles, 'label_map': label_map, 'rules': rules, "X":X, "y":y,'use_transformer': True}
         regular_predict = wrapper.predict(
         texts=texts, 
-        timeout=120,
+        timeout=pipeline.timeout,
         retries=None,
         api_key=api_key,
         client_ip=None)

{abstractintegratedmodule-0.5.0/AbstractIntegratedModule.egg-info → abstractintegratedmodule-0.5.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: AbstractIntegratedModule
-Version: 0.5.0
+Version: 0.5.1
 Summary: Framework for Advanced Integrated Non-LLM AI Module library - Backend Framework For Non-LLM AI Agent
 Author: Micro-Novelty
 Author-email: hernikpuspita5@gmail.com
@@ -42,7 +42,7 @@ https://github.com/Micro-Novelty/IntegratedPipeline-Specialized-Non-LLM-AI-Agent
 #### Note: The README here you are reading is a direct copy from my README Repository, to download the necessary files, you can visit my Repository with the provided link above.
 
 ### Library Short Description:
-- Development Stage: Beta, 0.5.0.
+- Development Stage: Beta, 0.5.1.
 - Maintainer: Micro-Novelty.
 - library Source-Code is Open-sourced on github.
 - Purpose: Specifically Designed for providing Non-LLM AI Agent Framework for edge Devices, Optimized for ARM64 architecture.
@@ -64,6 +64,11 @@ https://github.com/Micro-Novelty/IntegratedPipeline-Specialized-Non-LLM-AI-Agent
    - AWE setup Proven Efficient on Hard-uncontrolled dataset such as Activity Recognition from the given Database.
    - LSTM is Optimized efficiently for scarce data with AWE method.
    - Robust Advanced prediction capabilities proven effective on ARM64 Using MLP + LSTM Architectures.
+- Changelog:
+     - v0.5.1:
+         [=] New features: 
+          - Added new dynamic gate for Transformer Fixed and dynamic switching condition.
+          - Added Capabilities for the Model to save the Transformer weights as binaries in local SQlite database.
 -----
 
 <img width="1280" height="600" alt="WhatsApp Image 2026-05-27 at 07 16 32" src="https://github.com/user-attachments/assets/4b58a556-45a3-419b-96fd-9c1b76cac574" />

{abstractintegratedmodule-0.5.0 → abstractintegratedmodule-0.5.1}/README.md

@@ -10,7 +10,7 @@ https://github.com/Micro-Novelty/IntegratedPipeline-Specialized-Non-LLM-AI-Agent
 #### Note: The README here you are reading is a direct copy from my README Repository, to download the necessary files, you can visit my Repository with the provided link above.
 
 ### Library Short Description:
-- Development Stage: Beta, 0.5.0.
+- Development Stage: Beta, 0.5.1.
 - Maintainer: Micro-Novelty.
 - library Source-Code is Open-sourced on github.
 - Purpose: Specifically Designed for providing Non-LLM AI Agent Framework for edge Devices, Optimized for ARM64 architecture.
@@ -32,6 +32,11 @@ https://github.com/Micro-Novelty/IntegratedPipeline-Specialized-Non-LLM-AI-Agent
    - AWE setup Proven Efficient on Hard-uncontrolled dataset such as Activity Recognition from the given Database.
    - LSTM is Optimized efficiently for scarce data with AWE method.
    - Robust Advanced prediction capabilities proven effective on ARM64 Using MLP + LSTM Architectures.
+- Changelog:
+     - v0.5.1:
+         [=] New features: 
+          - Added new dynamic gate for Transformer Fixed and dynamic switching condition.
+          - Added Capabilities for the Model to save the Transformer weights as binaries in local SQlite database.
 -----
 
 <img width="1280" height="600" alt="WhatsApp Image 2026-05-27 at 07 16 32" src="https://github.com/user-attachments/assets/4b58a556-45a3-419b-96fd-9c1b76cac574" />

{abstractintegratedmodule-0.5.0 → abstractintegratedmodule-0.5.1}/setup.py

@@ -11,7 +11,7 @@ with open("README.md", encoding="utf-8") as f:
 
 setup(
     name="AbstractIntegratedModule",
-    version="0.5.0",
+    version="0.5.1",
     description="Framework for Advanced Integrated Non-LLM AI Module library - Backend Framework For Non-LLM AI Agent",
     long_description=long_description,
     long_description_content_type="text/markdown",