fix: install rfdetr train extras in RF-DETR finetuning notebooks

[asfandiyar-rf]

Problem

On a fresh Colab run, the training cell in both RF-DETR finetuning notebooks fails:

ImportError: RF-DETR training dependencies are missing. Install them with `pip install "rfdetr[train,loggers]"` and try again.

Two causes:

1. Newer rfdetr releases (current: 1.8.3) moved training dependencies (pytorch_lightning, torchmetrics, pycocotools, etc.) into the optional [train] extra, so the plain rfdetr install can no longer train.
2. The existing install line !pip install -q rfdetr>=1.4.0 ... is unquoted, so the shell treats >=1.4.0 as an output redirect — the version constraint was never actually applied and Colab always pulled the latest release.

Fix

!pip install -q "rfdetr[train,loggers]>=1.4.0" roboflow "supervision==0.29.1"

• [train] restores the training dependencies; [loggers] matches the package's own error message and keeps default TensorBoard logging working outside Colab.
• Quoting makes the >=1.4.0 floor actually apply.
• supervision is pinned to 0.29.1, following the pattern in the newer rf-detr-keypoint-detection.ipynb, to prevent the same class of drift from the annotator API.

Notebooks changed

• notebooks/how-to-finetune-rf-detr-on-detection-dataset.ipynb
• notebooks/how-to-finetune-rf-detr-on-segmentation-dataset.ipynb

Both verified end-to-end on fresh Colab runs (training now completes).

Note: nvidia-launchables/how-to-finetune-rf-detr-on-segmentation-dataset-a100.ipynb still pins rfdetr==1.3.0 (pre-extras split, so not broken today) — left untouched since updating it would need an A100 retest.

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[Copilot]

Pull request overview

Updates the RF-DETR finetuning notebooks’ dependency install cells so fresh Colab runs reliably install the correct RF-DETR version and the optional training/logging dependencies needed for training.

Changes:

• Quote the rfdetr...>=... spec so >= is not misinterpreted by the shell as redirection.
• Install rfdetr with [train,loggers] extras to include training dependencies required by newer rfdetr releases.
• Pin supervision to 0.29.1 to reduce notebook drift and match the pattern used elsewhere in the repo.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
notebooks/how-to-finetune-rf-detr-on-detection-dataset.ipynb	Fixes dependency installation to include RF-DETR training extras and pins supervision for stable Colab runs.
notebooks/how-to-finetune-rf-detr-on-segmentation-dataset.ipynb	Same dependency installation fix as detection notebook to ensure training works on fresh Colab.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.