Regarding the open-source license of ONNX models

[dpnotnull]

Apologies for the doubt, but from what I've seen on Netron, the layout model and tsr model used by DeepDoc are under the AGPL-3.0 license.

Does this mean that I cannot freely use DeepDoc in my company's commercial projects?

Looking forward to your response, many thanks🙏

[KevinHuSh]

Which parts are under AGPL-3.0?

[dpnotnull]

Dear Team: Many thanks for your reply, and I apologize for the late email response. When I check the metadata of the .onnx models , found that the tsr.onnx and layout.xxx are based on AGPL-3.0 license: $ onnx inspect -m ragflow/ragflow/rag/res/deepdoc/tsr.onnx  Meta information

…

--------------------------------------------------------------------------------   IR Version: 8   Opset Import: [version: 17 ]   Producer name: pytorch   Producer version: 2.2.0   Domain:    Doc string:    meta.{} = {} description Ultralytics best model trained on PubTables-1M-yolo-datasets/data.yaml   meta.{} = {} author Ultralytics   meta.{} = {} license AGPL-3.0 https://ultralytics.com/license   meta.{} = {} date 2024-03-25T04:36:35.872713   meta.{} = {} version 8.1.18   meta.{} = {} stride 32   meta.{} = {} task detect   meta.{} = {} batch 1   meta.{} = {} imgsz [640, 640]   meta.{} = {} names {0: 'table', 1: 'table column', 2: 'table row', 3: 'table column header', 4: 'table projected row header', 5: 'table spanning cell', 6: 'table grid cell'} onnx inspect -m ragflow/ragflow/rag/res/deepdoc/layout.onnx  Meta information

-------------------------------------------------------------------------------- &nbsp; IR Version: 8 &nbsp; Opset Import: [version: 17 ] &nbsp; Producer name: pytorch &nbsp; Producer version: 2.2.0 &nbsp; Domain:&nbsp; &nbsp; Doc string:&nbsp; &nbsp; meta.{} = {} description Ultralytics best model trained on /yifan/CDLA-yolo/data.yaml &nbsp; meta.{} = {} author Ultralytics &nbsp; meta.{} = {} license AGPL-3.0 https://ultralytics.com/license &nbsp; meta.{} = {} date 2024-02-29T01:29:38.132690 &nbsp; meta.{} = {} version 8.1.18 &nbsp; meta.{} = {} stride 32 &nbsp; meta.{} = {} task detect &nbsp; meta.{} = {} batch 1 &nbsp; meta.{} = {} imgsz [640, 640] &nbsp; meta.{} = {} names {0: 'none', 1: 'text', 2: 'title', 3: 'figure', 4: 'figure_caption', 5: 'table', 6: 'table_caption', 7: 'header', 8: 'footer', 9: 'reference', 10: 'equation'} Does this mean that in our company's commercial projects, at least we cannot use these two types of models freely? Would you please help to check this? Thanks again🙂 BR, Yunpeng Original From:"Kevin Hu"< ***@***.*** &gt;; Date:2024/9/11 9:40 To:"infiniflow/ragflow"< ***@***.*** &gt;; CC:"dpnotnull"< ***@***.*** &gt;;"Author"< ***@***.*** &gt;; Subject:Re: [infiniflow/ragflow] Regarding the open-source license of ONNXmodels (Discussion #2257) Which parts are under AGPL-3.0? — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: ***@***.***&gt;

[dpnotnull]

I git clone https://github.com/infiniflow/ragflow yesterday, and run below deepdoc demo script:

python deepdoc/vision/t_ocr.py --inputs XXXX --output_dir XXXX

Then downloading the model file:

[dpnotnull]

And when I check the metadata of some onnx model file

onnx inspect -m tsr.onnx

Still display :

...
meta.{} = {} description Ultralytics best model trained on PubTables-1M-yolo-datasets/data.yaml
meta.{} = {} author Ultralytics
meta.{} = {} license AGPL-3.0 https://ultralytics.com/license
meta.{} = {} date 2024-03-25T04:36:35.872713
meta.{} = {} version 8.1.18
meta.{} = {} stride 32
meta.{} = {} task detec
...

[dpnotnull]

Did I misconfigure any relevant environment variables, script parameters, or overlook any other detailed settings?

[KevinHuSh]

You could download from huggingface.
https://huggingface.co/InfiniFlow/deepdoc

[dpnotnull]

It's apache-2.0 now.

meta.{} = {} license Apache-2.0 http://www.apache.org/licenses/LICENSE-2.0

Many thanks~ 🙏🙏🙏