Merge branch 'main' of https://huggingface.co/spaces/nvidia/nemotron-ocr-v2 into main

push latest whl

Upload Screenshot 2026-04-09 at 2.59.12 PM.png

Upload 2 files

Delete mosaic_raw.png

Delete examples_collage.png

Upload 3 files

batched pipeline

batched pipeline

v1 whl

debug whl

reverted whl

updated whl

added v1 pipeline

Merge branch 'main' of https://huggingface.co/spaces/nvidia/nemotron-ocr-v2 into main

updated whl

Update README.md

moved disclaimer

moved disclaimer

added disclaimer

updated whl with relational fix

fix v1 missing config

Merge branch 'main' of https://huggingface.co/spaces/nvidia/nemotron-ocr-v2 into main

Rebuild wheel with relational model changes, add v1 model option

Update README.md

Create LICENSE

Update app.py

added layout aware

aggregate switch

aggregated at sentence level

updated app.py

updated app.py

updated app.py

updated app.py

updated visualization

Rebuild wheel with torch 2.9.1 to match HF Space runtime

Rebuild wheel with torch 2.9.1 to match HF Space runtime

Rebuild wheel with torch 2.9.1 to match HF Space runtime

Fix pillow version conflict with gradio

Fix wheel installation for Gradio SDK

Switch to Gradio SDK with pre-built cu128 wheel

Pre-built wheel replaces in-image CUDA compilation

Space: Docker + NGC PyTorch, prebuild CUDA extension

Build C++ extension at startup for torch ABI compatibility

Set python_version to 3.12 to match compiled .so

Initial Space: Nemotron OCR v2 demo with ZeroGPU