Spaces:
Runtime error
Runtime error
File size: 2,485 Bytes
51f704e 2e92739 51f704e 2e92739 51f704e 2e92739 51f704e 2e92739 51f704e 2e92739 51f704e b8fcf34 51f704e 2e92739 51f704e 2e92739 51f704e 2e92739 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 |
import spaces
import torch
import re
import gradio as gr
from threading import Thread
from transformers import TextIteratorStreamer, AutoTokenizer, AutoModelForCausalLM
import subprocess
# Install flash-attn for faster inference
subprocess.run('pip install flash-attn --no-build-isolation', env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"}, shell=True)
# Model and tokenizer setup
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()
# Function to generate responses
@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()
# Create the Gradio interface
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 layout
chatbot = gr.Chatbot()
# Image upload and prompt input
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 message button
send_btn = gr.Button("Send")
# Function to send message and get response
def send_message(history, prompt):
history.append((prompt, None))
response = answer_question(img.value, prompt)
history.append((None, response))
return history, "" # Clear the input box
send_btn.click(send_message, [chatbot, prompt], [chatbot, prompt])
prompt.submit(send_message, [chatbot, prompt], [chatbot, prompt])
demo.queue().launch() |