License clarification for official PP-YOLOE+_S COCO checkpoint and ONNX redistribution

Hello PaddleDetection maintainers,

This is closely related to #9493, but we are specifically seeking clarification for the official PP-YOLOE+_S COCO checkpoint distributed through the PaddleDetection model zoo.

We are evaluating this checkpoint for use in a commercial, closed-source media management product.

We understand that the PaddleDetection source repository is licensed under Apache-2.0, but we have not been able to find an explicit license statement covering the distributed pretrained checkpoint itself.

Could you please confirm whether the following uses are permitted for the official PP-YOLOE+_S COCO checkpoint:

1. Commercial use of the official pretrained checkpoint.
2. Conversion of the checkpoint to ONNX.
3. Server-side inference using the converted ONNX model.
4. Bundling or distributing the converted ONNX model with a closed-source commercial application or Docker image.
5. Fine-tuning or modifying the model in the future.
6. Redistributing resulting derivative weights.

If these uses are permitted, could you also clarify:

• whether Apache License 2.0 applies to the official pretrained checkpoint and converted/derivative weights;
• which LICENSE, NOTICE, attribution, copyright or model-card information must accompany the model;
• whether PaddlePaddle requires any additional obligations arising from the COCO training data.

We are specifically asking about the checkpoint distributed from the official PaddleDetection PP-YOLOE model table, not a third-party mirror.

Thank you for the clarification.