Update app.py

app.py

@@ -6,29 +6,28 @@ import torch
 import spaces
 
 # Initialize the client with your model
-client = InferenceClient("karpathy/gpt2_1558M_final2_hf") # Replace with your model's name or endpoint
+client = InferenceClient("karpathy/gpt2_1558M_final2_hf")
 
 default_system = 'You are a helpful assistant'
 
 @spaces.GPU
 def respond(message, history, system_message, max_tokens, temperature, top_p):
-    messages = [{"role": "system", "content": system_message}]
+    # Combine system message, history, and new message
+    full_prompt = f"{system_message}\n\n"
     for user, assistant in history:
-        messages.append({"role": "user", "content": user})
-        messages.append({"role": "assistant", "content": assistant})
-    messages.append({"role": "user", "content": message})
+        full_prompt += f"Human: {user}\nAssistant: {assistant}\n"
+    full_prompt += f"Human: {message}\nAssistant:"
 
     response = ""
-    for chunk in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
+    for chunk in client.text_generation(
+        full_prompt,
+        max_new_tokens=max_tokens,
         stream=True,
         temperature=temperature,
         top_p=top_p,
     ):
-        if chunk.choices[0].delta.content is not None:
-            token = chunk.choices[0].delta.content
-            response += token
+        if chunk.token.text:
+            response += chunk.token.text
             yield history + [(message, response)]
 
     # If the response is empty, yield a default message
@@ -43,21 +42,15 @@ def modify_system_session(system):
         system = default_system
     return system, system, []
 
-def use_example(example):
-    return example
-
-def set_unicorn_example():
-    return unicorn_example
-
-def set_time_travel_example():
-    return time_travel_example
+def use_example_and_respond(example, history, system_message, max_tokens, temperature, top_p):
+    return example, respond(example, history, system_message, max_tokens, temperature, top_p)
 
 # Define example prompts
 unicorn_example = "In a shocking finding, scientist discovered a herd of unicorns living in a remote, previously unexplored valley, in the Andes Mountains. Even more surprising to the researchers was the fact that the unicorns spoke perfect English."
 time_travel_example = "Explain the grandfather paradox in time travel and propose a potential resolution."
 
 with gr.Blocks() as demo:
-    gr.Markdown("<h1 style='text-align: center;'>LLM.C 1.5B Chat Demo</h1>")
+    gr.Markdown("<h1 style='text-align: center;'>GPT-2 1.5B Chat Demo</h1>")
     
     with gr.Row():
         with gr.Column(scale=3):
@@ -66,7 +59,7 @@ with gr.Blocks() as demo:
             modify_system = gr.Button("🛠️ Set system prompt and clear history")
     
     system_state = gr.Textbox(value=default_system, visible=False)
-    chatbot = gr.Chatbot(label='LLM.C Chat')
+    chatbot = gr.Chatbot(label='GPT-2 1.5B Chat')
     message = gr.Textbox(lines=1, label='Your message')
     
     with gr.Row():
@@ -89,14 +82,23 @@ with gr.Blocks() as demo:
     submit.click(respond, inputs=[message, chatbot, system_state, max_tokens, temperature, top_p], outputs=[chatbot])
     clear_history.click(fn=clear_session, inputs=[], outputs=[message, chatbot])
     modify_system.click(fn=modify_system_session, inputs=[system_input], outputs=[system_state, system_input, chatbot])
-
-    example1.click(fn=set_unicorn_example, inputs=[], outputs=[message])
-    example2.click(fn=set_time_travel_example, inputs=[], outputs=[message])
+    
+    # New event handlers for example prompts
+    example1.click(
+        fn=use_example_and_respond, 
+        inputs=[lambda: unicorn_example, chatbot, system_state, max_tokens, temperature, top_p], 
+        outputs=[message, chatbot]
+    )
+    example2.click(
+        fn=use_example_and_respond, 
+        inputs=[lambda: time_travel_example, chatbot, system_state, max_tokens, temperature, top_p], 
+        outputs=[message, chatbot]
+    )
 
     gr.Markdown(
         """
-        ## About LLM.C
-        some stuff about llmc
+        ## About GPT-2 1.5B
+        This is a large language model trained by OpenAI and fine-tuned by Andrej Karpathy. It's capable of generating human-like text based on the input it receives.
         """
     )