Skip to content

Latest commit

 

History

History
37 lines (32 loc) · 3.23 KB

File metadata and controls

37 lines (32 loc) · 3.23 KB

Third-party model notice

docling.rs's pipelines use third-party models of two kinds: ONNX graphs that are format conversions of docling-project's own PyTorch models (not weights docling.rs trains or modifies), and a few re-hosted, unmodified public releases of models docling itself builds on (OCR, ASR, the chunk tokenizer). All are licensed separately from docling.rs's own MIT code (see LICENSE) under their upstream terms:

Model Source License
RT-DETR layout model (layout_heron.onnx) docling-project/docling-layout-heron Apache-2.0
TableFormer (tableformer/{encoder,decoder,bbox}.onnx) docling-project/docling-models (model_artifacts/tableformer/accurate) CDLA-Permissive-2.0 / Apache-2.0
DocumentFigureClassifier (picture_classifier.onnx) docling-project/DocumentFigureClassifier-v2.5 (upstream's own ONNX, re-hosted unmodified) Apache-2.0
CodeFormulaV2 (cf_{vision,embed,decoder_kv}.onnx + cf_tokenizer.json; cf_decoder_kv_int8.onnx is a post-training quantization of the same export) docling-project/CodeFormulaV2 Apache-2.0
PP-OCRv3 recognition model + dictionary (ocr_rec.onnx, ppocr_keys_v1.txt) PaddleOCR's ch_PP-OCRv3_rec_infer, taken as the ready ONNX conversion re-hosted by SWHL/RapidOCR (unmodified) Apache-2.0
Whisper tiny (asr/{encoder_model,decoder_model}.onnx, vocab.json) onnx-community/whisper-tiny, the community ONNX export of openai/whisper-tiny (fetched directly, unmodified) Apache-2.0
Hybrid-chunker tokenizer (chunk/tokenizer.json) sentence-transformers/all-MiniLM-L6-v2's tokenizer.json (a tokenizer definition, no weights) Apache-2.0

(The pdfium shared library re-hosted alongside the models is not a model but carries its own terms: bblanchon/pdfium-binaries builds of Google's PDFium, BSD-3-Clause / Apache-2.0.)

scripts/install/export_layout.py, scripts/install/export_tableformer.py and scripts/install/export_code_formula.py do the conversion (PyTorch → ONNX via torch.onnx.export); no weights are retrained, fine-tuned, or otherwise altered. .github/workflows/publish-models.yml runs that conversion (and re-hosts pdfium + the OCR model alongside it) and publishes everything as GitHub Release assets on this repo (tag models-v1), fetched by scripts/install/download_dependencies.sh — see that script and crates/docling-node/deps.js — purely to spare downstream users the PyTorch/transformers/docling_ibm_models toolchain needed to export them locally.

All the upstream licenses permit redistribution with attribution; this file is that attribution. See the linked model cards for the full license text and any additional upstream terms.