kkasemos/experiments

1import gradio as gr2from huggingface_hub import InferenceClient3 4"""5For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference6"""7client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")8# client = InferenceClient("infly/OpenCoder-8B-Instruct")9 10 11def respond(12    message,13    history: list[tuple[str, str]],14    system_message,15    max_tokens,16    temperature,17    top_p,18):19    messages = [{"role": "system", "content": system_message}]20 21    for val in history:22        if val[0]:23            messages.append({"role": "user", "content": val[0]})24        if val[1]:25            messages.append({"role": "assistant", "content": val[1]})26 27    messages.append({"role": "user", "content": message})28 29    response = ""30 31    for message in client.chat_completion(32        messages,33        max_tokens=max_tokens,34        stream=True,35        temperature=temperature,36        top_p=top_p,37    ):38        token = message.choices[0].delta.content39 40        response += token41        yield response42 43 44"""45For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface46"""47demo = gr.ChatInterface(48    respond,49    additional_inputs=[50        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),51        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),52        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),53        gr.Slider(54            minimum=0.1,55            maximum=1.0,56            value=0.95,57            step=0.05,58            label="Top-p (nucleus sampling)",59        ),60    ],61)62 63 64if __name__ == "__main__":65    demo.launch()66