PaddleOCR-VL-0.9B

Duplicated from https://huggingface.co/PaddlePaddle/PaddleOCR-VL

Example use with transformers:

from transformers import AutoModelForCausalLM, AutoProcessor
import torch

DEVICE="cuda" if torch.cuda.is_available() else "mps" if torch.mps.is_available() else "cpu"
model_id = "pcuenq/PaddleOCR-VL-0.9B"

model = AutoModelForCausalLM.from_pretrained(
    model_id, trust_remote_code=True, dtype=torch.bfloat16
).to(DEVICE).eval()
processor = AutoProcessor.from_pretrained(model_id, trust_remote_code=True)

from transformers.image_utils import load_image
image_url = "https://fiverr-res.cloudinary.com/images/t_main1,q_auto,f_auto,q_auto,f_auto/gigs/154456946/original/41556aac80fc43dcb29ce656d786c0a6f9b4073f/do-handwritten-text-image-or-pdf-to-word-means-typing-form.jpg"
image = load_image(image_url)

messages = [{"role": "user", "content": "OCR"}]
text = processor.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = processor(text=[text], images=[image], return_tensors="pt").to(DEVICE)

generated = model.generate(**inputs, max_new_tokens=200, do_sample=False)

resp = processor.batch_decode(generated, skip_special_tokens=True)[0]
answer = resp.split(text)[-1].strip()
print(answer)