Upload flux_space_model_manager.py with huggingface_hub

flux_space_model_manager.py

@@ -33,6 +33,17 @@ class FluxModelManager:
             }
         }
         
+        # Pre-loaded LoRA models from annoyingpixel
+        self.preloaded_loras = {
+            'T11-Ultra-Portrait-E04': {
+                'repo_id': 'annoyingpixel/T11-Ultra-Portrait.E04.Lora.TA',
+                'model_id': 'annoyingpixel/T11-Ultra-Portrait.E04.Lora.TA',
+                'description': 'Ultra Portrait LoRA for FLUX.1-dev',
+                'trigger_words': 'T11-Ultra-Portrait-E04',
+                'base_model': 'black-forest-labs/FLUX.1-dev'
+            }
+        }
+        
         self.current_model = None
         self.current_pipeline = None
         self.loaded_loras = {}
@@ -186,6 +197,51 @@ class FluxModelManager:
         }
         
         return result.images[0], generation_info
+    
+    def load_preloaded_lora(self, lora_name: str, strength: float = 1.0) -> bool:
+        """
+        Load a pre-loaded LoRA directly from Hugging Face
+        """
+        if lora_name not in self.preloaded_loras:
+            print(f"❌ Pre-loaded LoRA '{lora_name}' not found")
+            return False
+            
+        if self.current_pipeline is None:
+            print("❌ No model loaded. Load a model first.")
+            return False
+            
+        try:
+            print(f"🔄 Loading pre-loaded LoRA: {lora_name}")
+            lora_info = self.preloaded_loras[lora_name]
+            
+            # Load LoRA directly from HF repository
+            self.current_pipeline.load_lora_weights(
+                lora_info['model_id'],
+                weight_name="default",
+                adapter_name=lora_name
+            )
+            
+            # Store LoRA info
+            self.loaded_loras[lora_name] = {
+                'path': lora_info['model_id'],
+                'strength': strength,
+                'trigger_words': lora_info.get('trigger_words', ''),
+                'description': lora_info['description']
+            }
+            
+            print(f"✅ Pre-loaded LoRA '{lora_name}' loaded with strength {strength}")
+            print(f"📝 Trigger words: {lora_info.get('trigger_words', 'None')}")
+            return True
+            
+        except Exception as e:
+            print(f"❌ Error loading pre-loaded LoRA: {e}")
+            return False
+    
+    def get_preloaded_loras(self) -> Dict:
+        """
+        Get information about available pre-loaded LoRAs
+        """
+        return self.preloaded_loras
 
 # Example usage for Gradio integration
 def create_model_manager():