FallnAI/CHATTERBOX

1import gradio as gr2from huggingface_hub import InferenceClient3 4client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")5 6 7def respond(8    message,9    history: list[tuple[str, str]],10    system_message,11    max_tokens,12    temperature,13    top_p,14):15    messages = [{"role": "system", "content": system_message}]16 17    for val in history:18        if val[0]:19            messages.append({"role": "user", "content": val[0]})20        if val[1]:21            messages.append({"role": "assistant", "content": val[1]})22 23    messages.append({"role": "user", "content": message})24 25    response = ""26 27    for message in client.chat_completion(28        messages,29        max_tokens=max_tokens,30        stream=True,31        temperature=temperature,32        top_p=top_p,33    ):34        token = message.choices[0].delta.content35 36        response += token37        yield response38 39 40demo = gr.ChatInterface(41    respond,42    theme='victorrauwcc/RCC',43    additional_inputs=[44        gr.Textbox(value="You are a helpful Chatbot created by FallnAI.", label="System message"),45        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),46        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),47        gr.Slider(48            minimum=0.1,49            maximum=1.0,50            value=0.95,51            step=0.05,52            label="Top-p (nucleus sampling)",53        ),54    ],55)56 57 58if __name__ == "__main__":59    demo.launch(show_api=False)