README.md

metadata

title: Image Describer
emoji: 🏃
colorFrom: purple
colorTo: red
sdk: gradio
sdk_version: 6.1.0
app_file: app.py
pinned: false

Image Describer (vit-gpt2) — Hugging Face Space

This Space runs nlpconnect/vit-gpt2-image-captioning with an optional T5 rewriter.

• CPU-friendly and returns model outputs uncensored (no filter implemented).
• Optional prompt box to bias the paraphrase step.

To run locally:

1. Create a Python 3.8+ venv and install requirements: pip install -r requirements.txt
2. Run: python app.py

Notes:

• For production/throughput on CPU, reduce max_length/num_beams or convert to ONNX + quantize.
• This Space uses the public HF model weights; no HF tokens are required to run the demo.