MimirQ parsing model bundle
Pinned, build-time model assets used by MimirQ. The application downloads this repository during image construction; request handling never downloads models.
Included components
| Paths | Upstream | Revision | License |
|---|---|---|---|
layout/, table/, ocr/ocr.res |
InfiniFlow/deepdoc |
de0e793dc6d744406c96dabd688ccc969f41b443 |
Apache-2.0 |
ocr/PP-OCRv4/ |
PaddleOCR PP-OCRv4 | 1cfba2e90fc938db55889873735088de210cc173 |
Apache-2.0 |
hf_onnx/ocr_pipeline__paddleocr_onnx__monkt__paddleocr-onnx/ |
monkt/paddleocr-onnx |
7b02d0a30a07ba2b92ad1ff5a8941ae2c633de65 |
Apache-2.0 |
hf/ and hf_onnx/table_structure* |
microsoft/table-transformer-structure-recognition-v1.1-all |
7587a7ef111d9dcbf8ac695f1376ab7014340a0c |
MIT |
xgboost/ |
InfiniFlow/text_concat_xgb_v1.0 |
722ed09a54f23f14fe0279ce6b74ce18e1960f54 |
Apache-2.0 |
layout/layout.onnx differs from the upstream file only by removal of legacy ONNX metadata. The TATR ONNX file is a local ONNX conversion of the pinned MIT-licensed model.
Excluded component
trocr_seal_384 is intentionally excluded because its training provenance and redistribution license have not been established. MimirQ deployments may provide it separately through SEAL_RECOGNITION_MODEL_DIR.
See SHA256SUMS for exact artifact hashes and LICENSES/ for applicable license texts.
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support