.

app.py

@@ -33,14 +33,10 @@ tokenizer.padding_side = "right"
 pipe = pipeline("text-generation", model=model, tokenizer=tokenizer)
 
 
-def respond(
-    message,
-    history,
-    system_message,
-    max_tokens=256,
-    temperature=0.3,
-    top_p=0.95,
-):
+def respond(message, history):
+    system_message = """You are General Knowledge Assistant. 
+    Answer the questions based on the provided information. 
+    Be succinct and use first-principles thinking to answer the questions."""
     # Construct the chat list
     chat_list = [{"role": "system", "content": system_message}]
     for user, assistant in history:
@@ -52,37 +48,23 @@ def respond(
         )
     chat_list.append({"role": "user", "content": message})
 
-    # Manually construct the prompt
-    prompt = ""
-    for chat in chat_list:
-        prompt += f"{chat['role']}: {chat['content']}\n"
-    prompt += "assistant:"
-
-    # Get the input length
-    input_ids = pipe.tokenizer.encode(prompt, return_tensors="pt")
-    input_length = input_ids.shape[1]
-
-    # Ensure max_tokens is a positive integer
-    max_tokens = max(1, int(max_tokens))
-
-    # Adjust max_length to be greater than input_length
-    max_length = input_length + max_tokens
+    prompt = pipe.tokenizer.apply_chat_template(
+        chat_list, tokenize=False, add_generation_prompt=True
+    )
 
     outputs = pipe(
         prompt,
-        max_new_tokens=max_tokens,
-        max_length=max_length,
+        max_new_tokens=256,
         num_beams=1,
         do_sample=True,
-        temperature=temperature,
-        top_p=top_p,
-        top_k=50
+        temperature=0.3,
+        top_p=0.95,
+        top_k=50,
     )
-    new_text = outputs[0]['generated_text'][len(prompt):]
+    new_text = outputs[0]["generated_text"][len(prompt) :]
     return new_text.strip()
 
 
-
 examples = [
     ["Suggest some breeds that get along with each other"],
     ["Explain LLM in AI"],
@@ -95,21 +77,6 @@ demo = gr.ChatInterface(
         placeholder="Enter your message here...", container=False, scale=7
     ),
     examples=examples,
-    additional_inputs=[
-        gr.Textbox(
-            value="You are General Knowledge Assistant. Answer the questions based on the provided information. Be succinct and use first-principles thinking to answer the questions.",
-            label="System message",
-        ),
-        gr.Slider(minimum=1, maximum=2048, value=256, step=1, label="Max new tokens"),
-        gr.Slider(minimum=0.1, maximum=2.0, value=0.3, step=0.1, label="Temperature"),
-        gr.Slider(
-            minimum=0.1,
-            maximum=1.0,
-            value=0.95,
-            step=0.05,
-            label="Top-p (nucleus sampling)",
-        ),
-    ],
     title="General Knowledge Assistant",
     description="Ask me anything about general knowledge. I'll try to answer succinctly using first principles.",
 )