tagger/fl2cog.py

from transformers import AutoProcessor, AutoModelForCausalLM
import spaces
from PIL import Image 
import torch
import re
import numpy as np

device = "cuda" if torch.cuda.is_available() else "cpu"

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

device = "cuda" if torch.cuda.is_available() else "cpu"
fl_model = AutoModelForCausalLM.from_pretrained('thwri/CogFlorence-2.1-Large', trust_remote_code=True).eval().to("cpu").eval()
fl_processor = AutoProcessor.from_pretrained('thwri/CogFlorence-2.1-Large', trust_remote_code=True)

def modify_caption(caption: str) -> str:
    special_patterns = [
        (r'the image is ', ''),
        (r'the image captures ', ''),
        (r'the image showcases ', ''),
        (r'the image shows ', ''),
        (r'the image ', ''),
    ]
    
    for pattern, replacement in special_patterns:
        caption = re.sub(pattern, replacement, caption, flags=re.IGNORECASE)
    
    caption = caption.replace('\n', '').replace('\r', '')
    caption = re.sub(r'(?<=[.,?!])(?=[^\s])', r' ', caption)
    caption = ' '.join(caption.strip().splitlines())
    
    return caption

@spaces.GPU(duration=30)
def process_image(image):
    if isinstance(image, np.ndarray):
        image = Image.fromarray(image)
    elif isinstance(image, str):
        image = Image.open(image)
    if image.mode != "RGB":
        image = image.convert("RGB")
    
    prompt = "<MORE_DETAILED_CAPTION>"
    
    fl_model.to(device)
    inputs = fl_processor(text=prompt, images=image, return_tensors="pt").to(device)
    generated_ids = fl_model.generate(
        input_ids=inputs["input_ids"],
        pixel_values=inputs["pixel_values"],
        max_new_tokens=1024,
        num_beams=3,
        do_sample=True
    )
    fl_model.to("cpu")
    generated_text = fl_processor.batch_decode(generated_ids, skip_special_tokens=False)[0]
    parsed_answer = fl_processor.post_process_generation(generated_text, task=prompt, image_size=(image.width, image.height))
    return modify_caption(parsed_answer["<MORE_DETAILED_CAPTION>"])

def predict_tags_fl2_cog(image: Image.Image, input_tags: str, algo: list[str]):
    def to_list(s):
        return [x.strip() for x in s.split(",") if not s == ""]
    
    def list_uniq(l):
        return sorted(set(l), key=l.index)
    
    if not "Use CogFlorence-2.1-Large" in algo:
        return input_tags
    tag_list = list_uniq(to_list(input_tags) + to_list(process_image(image) + ", "))
    tag_list.remove("")
    return ", ".join(tag_list)