cstr/OXR-1.0-DFINE-layout-GGUF

OXR-1.0 D-FINE layout GGUF (CrispEmbed)

GGUF conversions of SoMarkAI/OXR-1.0 by SoMark Team, for CrispEmbed's native ggml runtime. All credit for the model belongs to the original authors; these files only change the storage format. It targets CrispEmbed's layout_detect D-FINE architecture (layout.arch = dfine) and is the layout stage of OXR-1.0's native document pipeline (recognition model: cstr/OXR-1.0-GGUF).

License and attribution

• OXR License: Apache-2.0 with additional terms (LICENSE, verbatim; base text in LICENSE-APACHE-2.0). The additional terms take precedence:
  • commercial use without a separate SoMark commercial license is permitted only while you and your affiliates stay below USD 2,000,000 in total funding and in annual revenue;
  • an online service built on OXR or its derivatives must prominently identify SoMark OXR;
  • offering third parties a hosted API that competes with SoMark's document parsing / OCR / layout services is not permitted at any size without SoMark's written license;
  • violating these terms terminates the license automatically.
• These GGUF files are derivative works and are redistributed under the same terms; NOTICE carries the attribution.
• CrispEmbed only loads them after explicit acknowledgement (CRISPEMBED_OXR=1 and CRISPEMBED_ACCEPT_OXR_LICENSE=1); its model registry requires --accept-license oxr to download them, and crispembed-server warns about sections 2 and 3 when an OXR stage is configured.
• UPSTREAM_README.md is the unmodified upstream model card at revision c708407f9d2b4df658f2543d9132b62bb7221df4 (SHA256 a3241e175099ea57a4491be25cce8e965945889e41ed25c3c01911fd17a2236f).

Statement of changes (Apache-2.0 section 4(b))

These files were modified from the original by CrispEmbed: the checkpoint weights of revision c708407f9d2b4df658f2543d9132b62bb7221df4 were format-converted to GGUF (tensor renaming, casting, tokenizer and preprocessing metadata embedded). No weights were trained, fine-tuned, merged or otherwise altered.

Files

File	Precision	Bytes	SHA256
oxr-1.0-dfine-layout-f32.gguf	F32	78,118,336	f092601e1571c57b62e0ce8ae4e1ed8e91334405cbb8ca95e20dd5c012fbf7ef

• F32: lossless conversion of layout_analysis/last.pth

How the files were produced

1. Downloaded SoMarkAI/OXR-1.0 at the pinned revision c708407f9d2b4df658f2543d9132b62bb7221df4; the converter (models/convert-oxr-dfine-to-gguf.py) checks the source files, tensor names and shapes.
2. Quantized variants were produced with CrispEmbed's own crispembed-quantize from the listed source GGUF.
3. Built and uploaded by CrispEmbed commit 893293336fd51c2d4c2086bd93b986c8b8a45624 (workflow ocr-gguf-publish.yml, protocol docs/oxr-dfine-publication-v1.json); manifest.json lists every recipe, input pin, receipt and file hash.

Parity status (CrispEmbed native vs. the original PyTorch detector, CPU)

• The reference arms (f32) return the original service's final selected regions (same labels in the same order) on all 5 frozen pages (receipt (21), german (17), scan (9), table (2), japanese (1) regions), CrispEmbed commit 7f6dd6b1, reference revision c708407f9d2b.
• Strict per-stage budgets fail at decoder layers 2-3 on two pages (native is closer than PyTorch FP32 to an F64 yardstick on 56/56 stages); CPU speed is ~2.6x PyTorch; F16, GPU and representative layout AP are not qualified. Weight lineage: closest public checkpoint is D-FINE-M Objects365 (cosine ~0.94), unproven.

Usage

Requires acknowledging the OXR License:

CRISPEMBED_OXR=1 CRISPEMBED_ACCEPT_OXR_LICENSE=1 crispembed-server --layout oxr-1.0-dfine-layout-f32.gguf