app.py

import gradio as gr
from huggingface_hub import InferenceClient
from sae_auto_interp.sae import Sae

"""
For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
"""
client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")


def respond(
    message,
    history: list[tuple[str, str]],
    system_message,
    max_tokens,
    temperature,
    top_p,
):
    messages = [{"role": "system", "content": system_message}]

    for val in history:
        if val[0]:
            messages.append({"role": "user", "content": val[0]})
        if val[1]:
            messages.append({"role": "assistant", "content": val[1]})

    messages.append({"role": "user", "content": message})

    response = ""

    for message in client.chat_completion(
        messages,
        max_tokens=max_tokens,
        stream=True,
        temperature=temperature,
        top_p=top_p,
    ):
        token = message.choices[0].delta.content

        response += token
        yield response


CITATION_BUTTON_TEXT = """
@misc{zhang2024largemultimodalmodelsinterpret,
      title={Large Multi-modal Models Can Interpret Features in Large Multi-modal Models},
      author={Kaichen Zhang and Yifei Shen and Bo Li and Ziwei Liu},
      year={2024},
      eprint={2411.14982},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2411.14982},
}
"""


with gr.Blocks() as demo:
    gr.Markdown(
        """
        # Large Multi-modal Models Can Interpret Features in Large Multi-modal Models

        🔍 [ArXiv Paper](https://arxiv.org/abs/2411.14982) | 🏠 [LMMs-Lab Homepage](https://lmms-lab.framer.ai) | 🤗 [Huggingface Collections](https://huggingface.co/collections/lmms-lab/llava-sae-674026e4e7bc8c29c70bc3a3)
        """
    )

    with gr.Tabs(elem_classes="tab-buttons") as tabs:
        with gr.TabItem("Visualization of Activations", elem_id="visualization", id=0):
            image = gr.Image()

        with gr.TabItem("Steering Model", elem_id="steering", id=2):
            chatbot = gr.Chatbot()

    with gr.Row():
        with gr.Accordion("📙 Citation", open=False):
            gr.Markdown("```bib\n" + CITATION_BUTTON_TEXT + "\n```")


if __name__ == "__main__":
    demo.launch()