from transformers import AutoProcessor, PaliGemmaForConditionalGeneration
from PIL import Image
import requests
import torch

model_id = "gokaygokay/PaliGemma-PixelProse"

url = "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/tasks/car.jpg?download=true"
image = Image.open(requests.get(url, stream=True).raw)

model = PaliGemmaForConditionalGeneration.from_pretrained(model_id).to('cuda').eval()
processor = AutoProcessor.from_pretrained(model_id)

## prefix
prompt = "caption en"
model_inputs = processor(text=prompt, images=image, return_tensors="pt").to('cuda')
input_len = model_inputs["input_ids"].shape[-1]

with torch.inference_mode():
    generation = model.generate(**model_inputs, repetition_penalty=1.05, max_new_tokens=512, do_sample=False)
    generation = generation[0][input_len:]
    decoded = processor.decode(generation, skip_special_tokens=True)
    print(decoded)

Citation and attribution

This model release is maintained by Gökay Aydoğan. If you reference this repository in academic work, please cite it as follows and also cite the upstream models, datasets, or projects it builds upon.

bibtex

@software{aydogan2024paligemma_pixelprose,
  author = {Aydoğan, Gökay},
  title = {{PaliGemma-PixelProse}},
  year = {2024},
  publisher = {Hugging Face},
  url = {https://huggingface.co/gokaygokay/PaliGemma-PixelProse},
  note = {Model repository; cite the base model and upstream datasets as required.}
}