ultralytics 8.3.201

v8.3.201 - `ultralytics 8.3.201` Expand TensorFlow `ai-edge-litert≥1.2` support (non-macOS) (#22110)

on Python PyPI

🌟 Summary

Ultralytics 8.3.201 focuses on smoother TensorFlow exports, more reliable Docker-based exports, improved CLI help for Ultralytics HUB, and helpful docs refinements — all aimed at making training and deployment more stable and user-friendly. 🚀

📊 Key Changes

• TensorFlow export compatibility expanded (primary change) 🔧

  • Relaxed ai-edge-litert requirement to >=1.2.0 for non-macOS platforms; macOS remains pinned to <1.4.0 to avoid known issues.
  • Minor typing improvement: check_requirements() now accepts tuple[str] as well as Path | str | list[str].
  • See details in the current PR: Expand TensorFlow ai-edge-litert≥1.2 support (non-macOS) by @glenn-jocher.

• Export robustness and Docker reliability 🐳

  • Exporter now validates non-zero file size and correctly logs tuple-returning exporters for accurate success reporting: Assert non-zero filesize in Exporter by @glenn-jocher.
  • Docker export images install TensorRT on-demand (no pre-install) to match runtime CUDA, reducing version conflicts: Use AutoInstall for TensorRT package by @Y-T-G.
  • Dockerfiles export using a local temp path (tmp/yolo11n.pt) for EdgeTPU/NCNN/IMX steps, improving stability and reproducibility: Export in Dockerfile /tmp dir by @glenn-jocher.

• Better CLI help for Ultralytics HUB 🧭

  • yolo -h now shows the correct context-aware help for HUB subcommands by adjusting help precedence: Display help message with yolo -h by @Y-T-G.

• Training flexibility for multimodal (YOLOE) workflows 🧠

  • Removed training-mode assertion to allow get_text_pe() during training: Remove training mode assertion in get_text_pe() by @Y-T-G.

• Dataset reliability fix 📦

  • Correct CSV parsing for SKU-110K (no-header CSVs) to prevent misaligned columns: Fix CSV read error in SKU-110K.yaml by @Y-T-G.

• Documentation improvements 📚

  • New in-depth guide: Model YAML Configuration Guide — architecture structure, scaling, TorchVision backbones, and custom modules: Docs PR by @Y-T-G.
  • Augmentation docs now show a “Supported Tasks” column to clarify applicability across detect/segment/pose/OBB/classify: Add Supported Tasks column by @Y-T-G.

🎯 Purpose & Impact

• Smoother TensorFlow export workflows ✅

  • Fewer dependency conflicts and better forward-compatibility for ONNX→TensorFlow exports on Linux/Windows.
  • macOS stability preserved with a safe upper bound on ai-edge-litert. 🛡️

• More reliable and portable Docker builds 🔁

  • On-demand TensorRT avoids CUDA/TensorRT mismatches.
  • Temp-path exports reduce flakiness and speed up CI/CD builds.

• Clearer, more helpful CLI and logs 🗺️

  • yolo -h behaves as users expect, especially for Ultralytics HUB commands.
  • Exporter now catches empty outputs early and logs the correct paths and sizes.

• Enhanced training flexibility and dataset setup 🧪

  • Enables text positional embeddings retrieval during training for multimodal YOLOE scenarios.
  • SKU-110K dataset downloads parse cleanly, reducing setup hiccups.

• Easier customization and configuration 📘

  • The Model YAML guide and updated augmentation reference make it simpler to design, scale, and deploy custom YOLO11 models with confidence.

Tip: To try TensorFlow SavedModel export with the updated dependency handling:

pip install ultralytics
yolo export model=yolo11n.pt format=saved_model

What's Changed

• Fix CSV read error in SKU-110K.yaml download script by @Y-T-G in #22094
• Remove training mode assertion in get_text_pe() by @Y-T-G in #22076
• Update Dockerfile-export to use AutoInstall for TensorRT package by @Y-T-G in #22084
• Export in Dockerfile /tmp dir by @glenn-jocher in #22107
• Display help message with yolo -h by @Y-T-G in #22102
• Assert non-zero filesize in Exporter by @glenn-jocher in #22108
• Add "Supported Tasks" column to augmentation arguments table by @Y-T-G in #21900
• Docs: Add comprehensive model YAML configuration guide by @Y-T-G in #21830
• ultralytics 8.3.201 Expand TensorFlow ai-edge-litert≥1.2 support (non-macOS) by @glenn-jocher in #22110

Full Changelog: v8.3.200...v8.3.201