docparser-models

Every model file DocParser (a Rust document-parsing engine: Triton serving + an in-process ONNX Runtime backend) loads, flat, one commit per deployment. The checkout's models/MANIFEST.toml pins a commit and maps each file to its place in the Triton model repository; scripts/fetch-models.sh downloads and verifies them against MANIFEST.txt here (sha256 per file). Nothing is trained here: the files are PaddlePaddle's own ONNX exports, scripted derivations of them, and Texo's optimum export.

File	What	Provenance	Licence
layout_fp32.onnx	PP-DocLayoutV3 (RT-DETR-L, 25 classes + reading order), FP32	bit-identical copy of PaddlePaddle/PP-DocLayoutV3_onnx inference.onnx @ 46bbdf18 (sha256 45bf7175…)	Apache-2.0
layout_fp16_batchable.onnx	the same, output reshaped to [B,300,7] and converted to FP16 (GridSample / NMS kept FP32) for Triton batching on TensorRT	scripts/models/batchable_layout.py + convert_fp16.py over the file above (onnx 1.21, onnxconverter-common 1.16)	Apache-2.0
ocr_det.onnx	PP-OCRv6 text detection, medium tier	bit-identical copy of PaddlePaddle/PP-OCRv6_medium_det_onnx inference.onnx @ 61323801 (eb13b44b…)	Apache-2.0
ocr_rec_tiny_fp32.onnx	PP-OCRv6 text recognition, tiny tier (6906-way CTC)	bit-identical copy of PaddlePaddle/PP-OCRv6_tiny_rec_onnx inference.onnx @ 2612ab37 (9ef676d6…)	Apache-2.0
ocr_rec_tiny_fp16.onnx	the same in FP16	convert_fp16.py over the file above	Apache-2.0
table_encoder.onnx	SLANet-Plus encoder (PP-LCNet), the official graph up to the GRU loop's feature input	onnx.utils.extract_model over PaddlePaddle/SLANet_plus_onnx inference.onnx @ 7dbe640e (7790c0c1…) — scripts/models/export_slanet_plus.py::extract_encoder	Apache-2.0
table_decoder.bin	the GRU decoder's 16 parameter tensors as float32 (DocParser's host decoder)	export_slanet_plus.py::dump_decoder over the same file; byte-identical to TurboOCR's SLANeXt_wired_decoder.bin (f4b9f9b2…)	Apache-2.0
formula_encoder.onnx, formula_decoder.onnx, formula_decoder_past.onnx, formula_tokenizer.json	Texo transfer (20M): HGNetV2 encoder, MBart decoder without / with KV cache, tokenizer (687 tokens)	optimum image-to-text-with-past export of alephpi/FormulaNet @ b2668efe (scripts/models/export_texo.py; torch 2.14, transformers 4.57.6, optimum 2.1.0, opset 18)	AGPL-3.0 (LICENSE-texo)

Contracts (inputs, normalisation, post-processing) follow each source's inference.yml; the pre/post-processing code is in the DocParser checkout (crates/docparser-inference/src/preprocess/, deploy/triton_model_repository/). The CTC dictionary of the recogniser is a text file in that checkout, not here. Why these models and not their siblings is measured in docparser-bench/docs/BENCH-2026-09.md (OmniDocBench v1.6 §8, the per-slot ablations §1–7).

Redistribution: the Apache-2.0 files under their licence with this attribution; Texo's files under the AGPL-3.0, whose text is LICENSE-texo.