fix

app.py

@@ -6,10 +6,8 @@ def main():
     model_options = list(LlamaCppGemmaModel.AVAILABLE_MODELS.keys())
     task_options = ["Question Answering", "Text Generation", "Code Completion"]
     
-    model = LlamaCppGemmaModel(name="gemma-2b-it")
-    model.load_model()
     prompt_manager = PromptManager(task="Question Answering")
-    chat = GradioChat(model=model, prompt_manager=prompt_manager, model_options=model_options, task_options=task_options)
+    chat = GradioChat(prompt_manager=prompt_manager, model_options=model_options, task_options=task_options)
     chat.run()
 
 if __name__ == "__main__":

gemmademo/_chat.py

@@ -9,22 +9,28 @@ class GradioChat:
 
     Features:
     - A Gradio-based chatbot UI.
-    - Maintains chat history automatically.
-    - Uses Gemma (Hugging Face) model for generating responses.
+    - Dynamically loads models based on user selection.
+    - Uses Gemma (llama.cpp) for generating responses.
     - Formats user inputs before sending them to the model.
     """
 
-    def __init__(self, model: LlamaCppGemmaModel, prompt_manager: PromptManager, model_options: list[str], task_options: list[str]):
-        self.model = model
+    def __init__(self, prompt_manager: PromptManager, model_options: list[str], task_options: list[str]):
         self.prompt_manager = prompt_manager
         self.model_options = model_options
         self.task_options = task_options
+        self.current_model_name = "gemma-2b-it"  # Default model
+        self.model = self._load_model(self.current_model_name)
 
-    def run(self):
-        self._chat()
+    def _load_model(self, model_name: str):
+        """Loads the model dynamically when switching models."""
+        return LlamaCppGemmaModel(name=model_name).load_model()
 
     def _chat(self):
-        def chat_fn(history, message):
+        def chat_fn(history, message, selected_model):
+            if selected_model != self.current_model_name:
+                self.current_model_name = selected_model
+                self.model = self._load_model(selected_model)  # Reload model when changed
+
             prompt = self.prompt_manager.get_prompt(user_input=message)
             response = self.model.generate_response(prompt)
             return response
@@ -33,8 +39,11 @@ class GradioChat:
             chat_fn,
             textbox=gr.Textbox(placeholder="What is up?", container=False),
             additional_inputs=[
-                gr.Dropdown(choices=self.model_options, value="gemma-2b-it", label="Select Gemma Model"),
+                gr.Dropdown(choices=self.model_options, value=self.current_model_name, label="Select Gemma Model"),
                 gr.Dropdown(choices=self.task_options, value="Question Answering", label="Select Task"),
             ],
         )
         chat_interface.launch()
+
+    def run(self):
+        self._chat()