ultralytics 8.4.207 on Python PyPI

🌟 Summary

Ultralytics 8.4.207 restores reliable PaddlePaddle export and inference, while refreshing several Docker base images.

🐛 Fixed PaddlePaddle setup: Adds the missing six dependency required by x2paddle, which could break exports in fresh environments.
🎮 Improved Paddle GPU handling: A shared setup check now recognizes installed GPU builds and avoids replacing them with CPU builds. Export, inference, and PaddleOCR use this check.
✅ Re-enabled Paddle export testing: CI now tests both exporting a model to Paddle format and running inference with the exported model.
🐳 Updated Docker images: The AMD GPU image moves to Ubuntu 26.04 and Python 3.14; the NVIDIA ARM64 image updates to NVIDIA PyTorch 26.10, CUDA 13.4, and TensorRT 11.4. Other images use refreshed floating Debian tags.
📘 Clarified platform guidance: Documentation explains PaddlePaddle installation requirements and reiterates that TensorRT 11 does not support JetPack; Jetson TensorRT workflows should use a compatible TensorRT 10.x runtime.

🚀 Paddle exports and inference should work more consistently on fresh installations, and automated testing is back to catch regressions.
🛡️ Users with Paddle GPU installations are less likely to have them accidentally replaced during setup.
🔧 Updated Docker bases provide newer operating-system and NVIDIA software releases. These updates do not introduce model architecture changes.

Move Docker base images to the newest Ubuntu, NGC and floating Debian tags by @glenn-jocher in #27085
ultralytics 8.4.207 Fix PaddlePaddle export on new installs and CUDA machines by @glenn-jocher in #27088

Full Changelog: v8.4.206...v8.4.207