|
import spaces |
|
import torch |
|
import re |
|
import gradio as gr |
|
from threading import Thread |
|
from transformers import TextIteratorStreamer, AutoTokenizer, AutoModelForCausalLM |
|
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 = "2024-04-02" |
|
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() |
|
|
|
|
|
@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() |
|
|
|
|
|
with gr.Blocks(theme="Monochrome") as demo: |
|
gr.Markdown( |
|
""" |
|
# AskMoondream: Moondream 2 Demonstration Space |
|
Moondream2 is a 1.86B parameter model initialized with weights from SigLIP and Phi 1.5. |
|
Modularity AI presents this open source huggingface space for running fast experimental inferences on Moondream2. |
|
""" |
|
) |
|
|
|
|
|
chatbot = gr.Chatbot() |
|
|
|
|
|
with gr.Row(): |
|
img = gr.Image(type="pil", label="Upload an Image") |
|
prompt = gr.Textbox(label="Your Question", placeholder="Ask something about the image...", show_label=False) |
|
|
|
|
|
send_btn = gr.Button("Send") |
|
|
|
|
|
def send_message(history, prompt): |
|
history.append((prompt, None)) |
|
response = answer_question(img.value, prompt) |
|
history.append((None, response)) |
|
return history, "" |
|
|
|
send_btn.click(send_message, [chatbot, prompt], [chatbot, prompt]) |
|
prompt.submit(send_message, [chatbot, prompt], [chatbot, prompt]) |
|
|
|
demo.queue().launch() |