daniel-cerebras akhaliq HF staff commited on
Commit
cf545bf
1 Parent(s): 6efec2d

use gradio blocks and chatbot (#2)

Browse files

- use gradio blocks and chatbot (f2b8faf5fd99c7a718e029d1eb2fecc2441ee9e3)


Co-authored-by: AK <akhaliq@users.noreply.huggingface.co>

Files changed (1) hide show
  1. app.py +76 -23
app.py CHANGED
@@ -31,7 +31,7 @@ def generate_response(api_key: str, prompt: str) -> Tuple[List[Tuple[str, str, f
31
  Generate a response to the given prompt using a step-by-step reasoning approach.
32
  """
33
  system_message = """You are an expert AI assistant that explains your reasoning step by step. For each step, provide a title that describes what you're doing in that step, along with the content. Decide if you need another step or if you're ready to give the final answer. Respond in JSON format with 'title', 'content', and 'next_action' (either 'continue' or 'final_answer') keys. USE AS MANY REASONING STEPS AS POSSIBLE. AT LEAST 3. BE AWARE OF YOUR LIMITATIONS AS AN LLM AND WHAT YOU CAN AND CANNOT DO. IN YOUR REASONING, INCLUDE EXPLORATION OF ALTERNATIVE ANSWERS. CONSIDER YOU MAY BE WRONG, AND IF YOU ARE WRONG IN YOUR REASONING, WHERE IT WOULD BE. FULLY TEST ALL OTHER POSSIBILITIES. YOU CAN BE WRONG. WHEN YOU SAY YOU ARE RE-EXAMINING, ACTUALLY RE-EXAMINE, AND USE ANOTHER APPROACH TO DO SO. DO NOT JUST SAY YOU ARE RE-EXAMINING. USE AT LEAST 3 METHODS TO DERIVE THE ANSWER. USE BEST PRACTICES."""
34
-
35
  messages = [
36
  {"role": "system", "content": system_message},
37
  {"role": "user", "content": prompt},
@@ -51,7 +51,7 @@ def generate_response(api_key: str, prompt: str) -> Tuple[List[Tuple[str, str, f
51
  steps.append((f"Step {step_count}: {step_data['title']}", step_data['content'], thinking_time))
52
  messages.append({"role": "assistant", "content": json.dumps(step_data)})
53
 
54
- if step_data['next_action'] == 'final_answer':
55
  break
56
 
57
  step_count += 1
@@ -63,34 +63,87 @@ def generate_response(api_key: str, prompt: str) -> Tuple[List[Tuple[str, str, f
63
  thinking_time = time.time() - start_time
64
  total_thinking_time += thinking_time
65
 
66
- steps.append(("Final Answer", final_data['content'], thinking_time))
67
 
68
  return steps, total_thinking_time
69
 
70
- def generate_ui(api_key: str, prompt: str) -> str:
71
  """
72
  Generate the UI output based on the response to the given prompt.
73
  """
74
  steps, total_time = generate_response(api_key, prompt)
75
- result = "\n\n".join([
76
- f"{'### ' if title.startswith('Final Answer') else '**'}{title}{'**' if not title.startswith('Final Answer') else ''}\n\n{content}"
77
- for title, content, _ in steps
78
- ])
79
- result += f"\n\n**Total thinking time:** {total_time:.2f} seconds"
80
- return result
 
 
 
81
 
82
- # Gradio Interface with an API key input box
83
- iface = gr.Interface(
84
- fn=generate_ui,
85
- inputs=[gr.Textbox(label="API Key", type="password", placeholder="Enter your Cerebras API key"),
86
- gr.Textbox(lines=2, label="Query", placeholder="Enter your query here...")],
87
- outputs="markdown",
88
- title="o1-like chain of thought - llama-3.1 70b on Cerebras",
89
- description="""
90
- Implement Chain of Thought with prompting to improve output accuracy.
91
- It is powered by Cerebras, ensuring fast reasoning steps.
92
- """
93
- )
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
94
 
95
  if __name__ == "__main__":
96
- iface.launch()
 
31
  Generate a response to the given prompt using a step-by-step reasoning approach.
32
  """
33
  system_message = """You are an expert AI assistant that explains your reasoning step by step. For each step, provide a title that describes what you're doing in that step, along with the content. Decide if you need another step or if you're ready to give the final answer. Respond in JSON format with 'title', 'content', and 'next_action' (either 'continue' or 'final_answer') keys. USE AS MANY REASONING STEPS AS POSSIBLE. AT LEAST 3. BE AWARE OF YOUR LIMITATIONS AS AN LLM AND WHAT YOU CAN AND CANNOT DO. IN YOUR REASONING, INCLUDE EXPLORATION OF ALTERNATIVE ANSWERS. CONSIDER YOU MAY BE WRONG, AND IF YOU ARE WRONG IN YOUR REASONING, WHERE IT WOULD BE. FULLY TEST ALL OTHER POSSIBILITIES. YOU CAN BE WRONG. WHEN YOU SAY YOU ARE RE-EXAMINING, ACTUALLY RE-EXAMINE, AND USE ANOTHER APPROACH TO DO SO. DO NOT JUST SAY YOU ARE RE-EXAMINING. USE AT LEAST 3 METHODS TO DERIVE THE ANSWER. USE BEST PRACTICES."""
34
+
35
  messages = [
36
  {"role": "system", "content": system_message},
37
  {"role": "user", "content": prompt},
 
51
  steps.append((f"Step {step_count}: {step_data['title']}", step_data['content'], thinking_time))
52
  messages.append({"role": "assistant", "content": json.dumps(step_data)})
53
 
54
+ if step_data.get('next_action') == 'final_answer':
55
  break
56
 
57
  step_count += 1
 
63
  thinking_time = time.time() - start_time
64
  total_thinking_time += thinking_time
65
 
66
+ steps.append(("Final Answer", final_data.get('content', 'No final answer provided.'), thinking_time))
67
 
68
  return steps, total_thinking_time
69
 
70
+ def generate_ui(api_key: str, prompt: str) -> Tuple[List[Tuple[str, str]], float]:
71
  """
72
  Generate the UI output based on the response to the given prompt.
73
  """
74
  steps, total_time = generate_response(api_key, prompt)
75
+ conversation = []
76
+ for title, content, _ in steps:
77
+ if title.startswith("Step"):
78
+ conversation.append(("Assistant", f"**{title}**\n\n{content}"))
79
+ elif title == "Final Answer":
80
+ conversation.append(("Assistant", f"**{title}**\n\n{content}"))
81
+ else:
82
+ conversation.append(("Assistant", content))
83
+ return conversation, total_time
84
 
85
+ # Gradio Blocks Interface with a Chatbot component and API key input
86
+ def main():
87
+ with gr.Blocks() as demo:
88
+ gr.Markdown("# o1-like Chain of Thought - LLaMA-3.1 70B on Cerebras")
89
+ gr.Markdown("""
90
+ Implement Chain of Thought with prompting to improve output accuracy.
91
+ Powered by Cerebras, ensuring fast reasoning steps.
92
+ """)
93
+
94
+ with gr.Row():
95
+ api_key_input = gr.Textbox(
96
+ label="Cerebras API Key",
97
+ type="password",
98
+ placeholder="Enter your Cerebras API key",
99
+ show_label=True
100
+ )
101
+
102
+ chatbot = gr.Chatbot(label="Conversation")
103
+ with gr.Row():
104
+ user_input = gr.Textbox(
105
+ label="Your Query",
106
+ placeholder="Enter your query here...",
107
+ show_label=True
108
+ )
109
+ submit_btn = gr.Button("Submit")
110
+
111
+ thinking_time_display = gr.Textbox(
112
+ label="Total Thinking Time",
113
+ value="",
114
+ interactive=False
115
+ )
116
+
117
+ def respond(api_key, message, history):
118
+ if not api_key:
119
+ return history, "Please provide a valid Cerebras API key."
120
+
121
+ steps, total_time = generate_response(api_key, message)
122
+ for title, content, _ in steps:
123
+ if title.startswith("Step"):
124
+ history.append(("Assistant", f"**{title}**\n\n{content}"))
125
+ elif title == "Final Answer":
126
+ history.append(("Assistant", f"**{title}**\n\n{content}"))
127
+ else:
128
+ history.append(("Assistant", content))
129
+ return history, f"**Total thinking time:** {total_time:.2f} seconds"
130
+
131
+ submit_btn.click(
132
+ fn=respond,
133
+ inputs=[api_key_input, user_input, chatbot],
134
+ outputs=[chatbot, thinking_time_display],
135
+ queue=True
136
+ )
137
+
138
+ # Optional: Allow pressing Enter to submit
139
+ user_input.submit(
140
+ fn=respond,
141
+ inputs=[api_key_input, user_input, chatbot],
142
+ outputs=[chatbot, thinking_time_display],
143
+ queue=True
144
+ )
145
+
146
+ demo.launch()
147
 
148
  if __name__ == "__main__":
149
+ main()