using keras model

app.py

@@ -1,11 +1,20 @@
 import gradio as gr
-from huggingface_hub import InferenceClient
+import keras_nlp
 
 """
 For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
 """
-client = InferenceClient("bhashwarsengupta/gemma-finance")
 
+model = keras_nlp.models.GemmaCausalLM.from_preset("kaggle://bhashwar22/gemma-for-finance/keras/gemma-for-finance")
+print("model successfully loaded!")
+
+context = """You are an intelligent personal finance assistant
+designed to help users understand various financial concepts.
+You are supposed to provide concise and easy-to-understand explanations for the requested questions,
+ensuring the users feel informed and confident about managing their money.
+Keep your answers limited to 50-100 words.
+If you receive any non-finance related query, please return the following response:
+\"Unrelated Topic\""""
 
 def respond(
     message,
@@ -15,29 +24,27 @@ def respond(
     temperature,
     top_p,
 ):
-    messages = [{"role": "system", "content": system_message}]
+    messages = f"Context: {context}\n"
 
     for val in history:
         if val[0]:
-            messages.append({"role": "user", "content": val[0]})
+            messages += f"Question: {val[0]}\n"
         if val[1]:
-            messages.append({"role": "assistant", "content": val[1]})
-
-    messages.append({"role": "user", "content": message})
+            messages += f"Answer: {val[1]}\n"
 
-    response = ""
+    messages += f"Question: {message}\nAnswer: "
 
-    for message in client.chat_completion(
+    output = model.generate(
         messages,
         max_tokens=max_tokens,
-        stream=True,
         temperature=temperature,
-        top_p=top_p,
-    ):
-        token = message.choices[0].delta.content
+        top_p=top_o
+    )
 
-        response += token
-        yield response
+    # Split by "Answer:" from the right and get the last part
+    response = output.rsplit("Answer: ", 1)[-1]
+    
+    return response    
 
 
 """