import streamlit as st from huggingface_hub import InferenceClient client = InferenceClient("mistralai/Mistral-7B-Instruct-v0.1") def format_prompt(message, history): prompt = "" for user_prompt, bot_response in history: prompt += f"[INST] {user_prompt} [/INST]" prompt += f" {bot_response} " prompt += f"[INST] {message} [/INST]" return prompt def generate(prompt, history, temperature=0.9, max_new_tokens=256, top_p=0.95, repetition_penalty=1.0): temperature = float(temperature) if temperature < 1e-2: temperature = 1e-2 top_p = float(top_p) generate_kwargs = dict( temperature=temperature, max_new_tokens=max_new_tokens, top_p=top_p, repetition_penalty=repetition_penalty, do_sample=True, seed=42, ) formatted_prompt = format_prompt(prompt, history) response = client.text_generation(formatted_prompt, **generate_kwargs) if isinstance(response, list) and len(response) > 0 and 'choices' in response[0]: return response[0]['choices'][0]['text'] elif 'choices' in response: return response['choices'][0]['text'] else: return response['text'] def main(): st.title("Mistral 7B Chat Interface") # Sidebar for adjusting parameters st.sidebar.header("Adjust Parameters") temperature = st.sidebar.slider("Temperature", 0.0, 1.0, 0.9, step=0.05) max_new_tokens = st.sidebar.slider("Max new tokens", 0, 1048, 256, step=64) top_p = st.sidebar.slider("Top-p (nucleus sampling)", 0.0, 1.0, 0.90, step=0.05) repetition_penalty = st.sidebar.slider("Repetition penalty", 1.0, 2.0, 1.2, step=0.05) # Chat interface user_input = st.text_area("User Input:", "") history = [] # You need to manage the conversation history here if st.button("Send"): history.append(("User", user_input)) bot_response = generate(user_input, history, temperature, max_new_tokens, top_p, repetition_penalty) history.append(("Bot", bot_response)) st.text("Chat History:") for role, message in history: st.write(f"{role}: {message}") if __name__ == "__main__": main()