crumpgazonj/sparse-step

Document QA

Layout-aware document question answering. Powered by Hugging Face Transformers.

Features

• —Graceful shutdown with request draining
• —Request queuing with configurable concurrency
• —Model quantization support (INT8/FP16)

Usage

python -c "import requests; print(requests.post('http://localhost:7860/predict', json={'text':'test'}).json())"