added support for llama3.2

.gitignore

@@ -165,4 +165,12 @@ cython_debug/
 
 #MacOS
 .DS_Store
-SOURCE_DOCUMENTS/.DS_Store
+SOURCE_DOCUMENTS/.DS_Store
+
+
+
+
+.byaldi/
+sessions/
+static/images/
+uploaded_documents/

models/model_loader.py

@@ -1,8 +1,10 @@
 # models/model_loader.py
 
 import os
-from transformers import Qwen2VLForConditionalGeneration, AutoProcessor
 import torch
+from transformers import Qwen2VLForConditionalGeneration, AutoProcessor
+from transformers import MllamaForConditionalGeneration
+
 from logger import get_logger
 
 logger = get_logger(__name__)
@@ -43,6 +45,7 @@ def load_model(model_choice):
         _model_cache[model_choice] = (model, processor, device)
         logger.info("Qwen model loaded and cached.")
         return _model_cache[model_choice]
+
     elif model_choice == 'gemini':
         # Load Gemini model
         import genai
@@ -52,13 +55,31 @@ def load_model(model_choice):
         _model_cache[model_choice] = (model, processor)
         logger.info("Gemini model loaded and cached.")
         return _model_cache[model_choice]
+
     elif model_choice == 'gpt4':
         # Load OpenAI GPT-4 model
         import openai
         openai.api_key = os.environ.get('OPENAI_API_KEY')
         _model_cache[model_choice] = (None, None)
         logger.info("GPT-4 model ready and cached.")
         return _model_cache[model_choice]
+
+    elif model_choice == 'llama-vision':
+        # Load Llama-Vision model
+        device = detect_device()
+        # model_id = "meta-llama/Llama-3.2-11B-Vision-Instruct"
+        model_id = "alpindale/Llama-3.2-11B-Vision-Instruct"
+        model = MllamaForConditionalGeneration.from_pretrained(
+            model_id,
+            torch_dtype=torch.float16 if device != 'cpu' else torch.float32,
+            device_map="auto"
+        )
+        processor = AutoProcessor.from_pretrained(model_id)
+        model.to(device)
+        _model_cache[model_choice] = (model, processor, device)
+        logger.info("Llama-Vision model loaded and cached.")
+        return _model_cache[model_choice]
+
     else:
         logger.error(f"Invalid model choice: {model_choice}")
         raise ValueError("Invalid model choice.")

models/responder.py

@@ -2,6 +2,7 @@
 
 from models.model_loader import load_model
 from logger import get_logger
+from PIL import Image
 import os
 
 logger = get_logger(__name__)
@@ -65,9 +66,35 @@ def generate_response(images, query, session_id, resized_height=280, resized_wid
             response = generate_gpt4_response(images, query, model)
             logger.info("Response generated using GPT-4 model.")
             return response
+
+        elif model_choice == 'llama-vision':
+            # Load model, processor, and device
+            model, processor, device = load_model('llama-vision')
+
+            # Process images
+            image_paths = [os.path.join('static', image) for image in images]
+            # For simplicity, use the first image
+            image_path = image_paths[0]
+            image = Image.open(image_path).convert('RGB')
+
+            # Prepare messages
+            messages = [
+                {"role": "user", "content": [
+                    {"type": "image"},
+                    {"type": "text", "text": query}
+                ]}
+            ]
+            input_text = processor.apply_chat_template(messages, add_generation_prompt=True)
+            inputs = processor(image, input_text, return_tensors="pt").to(device)
+
+            # Generate response
+            output = model.generate(**inputs, max_new_tokens=512)
+            response = processor.decode(output[0], skip_special_tokens=True)
+            return response
+
         else:
-            logger.error(f"Invalid model choice: {model_choice}")
-            return "Invalid model selected."
+                logger.error(f"Invalid model choice: {model_choice}")
+                return "Invalid model selected."
     except Exception as e:
         logger.error(f"Error generating response: {e}")
         return "An error occurred while generating the response."

templates/settings.html

@@ -25,6 +25,8 @@ <h2>Settings</h2>
                 <option value="qwen" {% if model_choice == 'qwen' %}selected{% endif %}>Qwen2-VL-7B-Instruct</option>
                 <option value="gemini" {% if model_choice == 'gemini' %}selected{% endif %}>Google Gemini</option>
                 <option value="gpt4" {% if model_choice == 'gpt4' %}selected{% endif %}>OpenAI GPT-4</option>
+                <option value="llama-vision" {% if model_choice == 'llama-vision' %}selected{% endif %}>Llama-Vision</option>
+
             </select>
         </div>
         <div class="form-group">