README.md

metadata

license: apache-2.0
base_model: Qwen/Qwen2.5-VL-3B-Instruct
language:
  - sr
pipeline_tag: image-to-text
tags:
  - art-curation
  - accessibility
  - multilingual
  - vision-language
  - LoRA
  - cultural-heritage
library_name: transformers
datasets:
  - bsr_train_mm.jsonl
  - bsr_test_mm.jsonl
metrics: []

SHIFT-D3.5-Qwen2.5-VL-3B-ArtCurator-DE

This LoRA-fine-tuned model generates detailed descriptions of cultural-heritage artworks for blind users in Serbian.

Usage

from transformers import QwenForCausalLM, QwenProcessor

processor = QwenProcessor.from_pretrained("JoseferEins/SHIFT-D3.5-Qwen2.5-VL-3B-ArtCurator-SR")
model     = QwenForCausalLM.from_pretrained("JoseferEins/SHIFT-D3.5-Qwen2.5-VL-3B-ArtCurator-SR")

inputs = processor(
    images="path/to/artwork.jpg",
    text="Опиши ову слику детаљно / Opiši ovu sliku detaljno.",
    return_tensors="pt"
)
out = model.generate(**inputs)
print(processor.decode(out[0], skip_special_tokens=True))