Spaces:
Runtime error
Runtime error
File size: 3,510 Bytes
848ce1e 8bef29c a553a03 8bef29c 8132ec4 8bef29c 8132ec4 8bef29c 8132ec4 8bef29c 838b9d1 43ba66f 8bef29c 848ce1e 43ba66f 838b9d1 43ba66f 848ce1e 8bef29c |
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 |
import gradio as gr
import requests
from PIL import Image
import torch
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("q-future/co-instruct-preview",
trust_remote_code=True,
torch_dtype=torch.float16,
use_flash_attention_2=True,
device_map={"":"cuda:0"})
def chat(message, history, image_1, image_2):
print(history)
if history:
if image_1 is not None and image_2 is None:
past_message = "USER: The image: <|image|> " + history[0][0] + " ASSISTANT:" + history[0][1]
for i in range((len(history) - 1)):
past_message += "USER:" +history[i][0] + " ASSISTANT:" + history[i][1] + "</s>"
message = past_message + "USER:" + message + " ASSISTANT:"
images = [image_1]
if image_1 is not None and image_2 is not None:
past_message = "USER: The first image: <|image|>\nThe second image: <|image|>" + history[0][0] + " ASSISTANT:" + history[0][1] + "</s>"
for i in range((len(history) - 1)):
past_message += "USER:" + history[i][0] + " ASSISTANT:" + history[i][1] + "</s>"
message = past_message + "USER:" + message + " ASSISTANT:"
images = [image_1, image_2]
else:
if image_1 is not None and image_2 is None:
message = "USER: The image: <|image|> " + message + " ASSISTANT:"
images = [image_1]
if image_1 is not None and image_2 is not None:
message = "USER: The first image: <|image|>\nThe second image: <|image|>" + message + " ASSISTANT:"
images = [image_1, image_2]
print(message)
return model.tokenizer.batch_decode(model.chat(message, images, max_new_tokens=150).clamp(0, 100000))[0].split("ASSISTANT:")[-1]
with gr.Blocks(title="img") as demo:
title_markdown = ("""
<div align="center">*Preview Version (v1)! Now we support two images as inputs! Try it now!*</div>
<h1 align="center"><a href="https://github.com/Q-Future/Q-Instruct"><img src="https://github.com/Q-Future/Q-Instruct/blob/main/q_instruct_logo.png?raw=true", alt="Q-Instruct (mPLUG-Owl-2)" border="0" style="margin: 0 auto; height: 85px;" /></a> </h1>
<h2 align="center">Q-Instruct: Improving Low-level Visual Abilities for Multi-modality Foundation Models</h2>
<h5 align="center"> Please find our more accurate visual scoring demo on <a href='https://huggingface.co/spaces/teowu/OneScorer'>[OneScorer]</a>!</h2>
<div align="center">
<div style="display:flex; gap: 0.25rem;" align="center">
<a href='https://github.com/Q-Future/Q-Instruct'><img src='https://img.shields.io/badge/Github-Code-blue'></a>
<a href="https://Q-Instruct.github.io/Q-Instruct/fig/Q_Instruct_v0_1_preview.pdf"><img src="https://img.shields.io/badge/Technical-Report-red"></a>
<a href='https://github.com/Q-Future/Q-Instruct/stargazers'><img src='https://img.shields.io/github/stars/Q-Future/Q-Instruct.svg?style=social'></a>
</div>
</div>
""")
gr.Markdown(title_markdown)
with gr.Row():
input_img_1 = gr.Image(type='pil', label="Image 1 (The first image)")
input_img_2 = gr.Image(type='pil', label="Image 2 (The second image)")
gr.ChatInterface(fn = chat, additional_inputs=[input_img_1, input_img_2])
demo.launch(share=True) |