app.py

import spaces
import torch
import re
import gradio as gr
from threading import Thread
from transformers import TextIteratorStreamer, AutoTokenizer, AutoModelForCausalLM
from PIL import ImageDraw
from torchvision.transforms.v2 import Resize

import subprocess
subprocess.run('pip install flash-attn --no-build-isolation', env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"}, shell=True)

model_id = "vikhyatk/moondream2"
revision = "2024-05-20"
tokenizer = AutoTokenizer.from_pretrained(model_id, revision=revision)
moondream = AutoModelForCausalLM.from_pretrained(
    model_id, trust_remote_code=True, revision=revision,
    torch_dtype=torch.bfloat16, device_map={"": "cuda"},
    attn_implementation="flash_attention_2"
)
moondream.eval()

control_vectors = torch.load("control_vectors.pt", map_location="cpu")
control_vectors = [t.to('cuda', dtype=torch.bfloat16) for t in control_vectors]

class LayerWrapper(torch.nn.Module):
    def __init__(self, og_layer, control_vectors, scale=4.2):
        super().__init__()
        self.og_layer = og_layer
        self.control_vectors = control_vectors
        self.scale = scale

    def forward(self, *args, **kwargs):
        layer_outputs = self.og_layer(*args, **kwargs)
        layer_outputs = (layer_outputs[0] + self.scale * self.control_vectors, *layer_outputs[1:])
        return layer_outputs

moondream.text_model.transformer.h = torch.nn.ModuleList([
    LayerWrapper(layer, vector, 4.2) for layer, vector in zip(og_h, control_vectors)
])

@spaces.GPU(duration=10)
def answer_question(img, prompt):
    image_embeds = moondream.encode_image(img)
    streamer = TextIteratorStreamer(tokenizer, skip_special_tokens=True)
    thread = Thread(
        target=moondream.answer_question,
        kwargs={
            "image_embeds": image_embeds,
            "question": prompt,
            "tokenizer": tokenizer,
            "streamer": streamer,
        },
    )
    thread.start()

    buffer = ""
    for new_text in streamer:
        buffer += new_text
        yield buffer.strip()

def extract_floats(text):
    # Regular expression to match an array of four floating point numbers
    pattern = r"\[\s*(-?\d+\.\d+)\s*,\s*(-?\d+\.\d+)\s*,\s*(-?\d+\.\d+)\s*,\s*(-?\d+\.\d+)\s*\]"
    match = re.search(pattern, text)
    if match:
        # Extract the numbers and convert them to floats
        return [float(num) for num in match.groups()]
    return None  # Return None if no match is found


def extract_bbox(text):
    bbox = None
    if extract_floats(text) is not None:
        x1, y1, x2, y2 = extract_floats(text)
        bbox = (x1, y1, x2, y2)
    return bbox

def process_answer(img, answer):
    if extract_bbox(answer) is not None:
        x1, y1, x2, y2 = extract_bbox(answer)
        draw_image = Resize(768)(img)
        width, height = draw_image.size
        x1, x2 = int(x1 * width), int(x2 * width)
        y1, y2 = int(y1 * height), int(y2 * height)
        bbox = (x1, y1, x2, y2)
        ImageDraw.Draw(draw_image).rectangle(bbox, outline="red", width=3)
        return gr.update(visible=True, value=draw_image)

    return gr.update(visible=False, value=None)

with gr.Blocks() as demo:
    gr.Markdown(
        """
        # 🌔 contemplative moondream
        a demo of [moondream](http://moondream.ai) steered to discuss the meaning of life using [activation vectors](https://github.com/vikhyat/moondream/blob/main/notebooks/RepEng.ipynb)
        """
    )
    with gr.Row():
        prompt = gr.Textbox(label="Input", value="Describe this image.", scale=4)
        submit = gr.Button("Submit")
    with gr.Row():
        img = gr.Image(type="pil", label="Upload an Image")
        with gr.Column():
            output = gr.Markdown(label="Response")
            ann = gr.Image(visible=False, label="Annotated Image")

    submit.click(answer_question, [img, prompt], output)
    prompt.submit(answer_question, [img, prompt], output)
    output.change(process_answer, [img, output], ann, show_progress=False)

demo.queue().launch()