MACHINE LEARNING / huggingface
Hugging Face Vision Training
Prepare dataset-validated detection, classification or segmentation training jobs with explicit Hub persistence and job-secret handling.
“Trains and fine-tunes vision models for object detection (D-FINE, RT-DETR v2, DETR, YOLOS), image classification (timm models — MobileNetV3, MobileViT, ResNet, ViT/DINOv3 — plus any”
01 / THE REASONING
Why this made the selection.
- It submits object-detection, classification, or SAM training to Hugging Face Jobs and saves results on the Hub.
- Job submission is an MCP tool rather than a Python import, and the Hub token must be passed explicitly.
- Separates dataset contracts for boxes/classes, image labels and segmentation masks/prompts.
- Includes task-specific training scripts, validation guidance and monitoring/persistence paths.
02 / THE REVIEW RECORD
What we actually inspected.
Source review has boundaries.
A clear record is more useful than a “safe” badge.
Material inspected
- skills/huggingface-vision-trainer/SKILL.md
- skills/huggingface-vision-trainer/scripts/object_detection_training.py
- README.md
- LICENSE
Our findings
- Training is aimed at Hugging Face Jobs so a local GPU is not required, and outputs are saved to the Hub.
- hf_jobs is an MCP tool, not a Python function, and the skill says not to import it from huggingface_hub.
- Jobs must receive an explicit hub token; implicit resolution is called unreliable.
- The skill does not poll job status on its own; it waits for the user to ask.
- Object-detection script samples boxes, clips/sanitizes them and injects HF_TOKEN before Trainer creation.
- Its inspected branches initialize Trackio and push trained artifacts when requested; these operations were not run.
Not established by this review
- The resource was not executed and its tools were not called.
- No skill execution, authenticated service operation, or outcome benchmark was performed.
The review applies to the material and revision named here. A newer upstream release can change its behavior.
03 / PUT IT TO WORK
Add a skill to your workflow.
DOCUMENTED COMMAND
npx skills add huggingface/skills/skills/huggingface-vision-trainer Run in your terminal to choose your AI tools, project or global scope, and installation method. This short command follows the current upstream version; the review covers the revision shown above.
Optional: install the reviewed revision
Keeps the original source and installer version where recorded, while leaving your AI tools and installation scope for you to choose.
npx skills add https://github.com/huggingface/skills/tree/c3ff942be3e07116b475253f0a8fc4f54682c1a8/skills/huggingface-vision-trainer - Install the pinned skill directory interactively, retaining its bundled references and scripts.
- Validate a sample and the complete dataset schema before selecting a task script.
- Set job timeout, secrets, target repository and privacy, then use a small run before a full GPU job.
Before you start
- HF Jobs entitlement and write-scoped Hub token
- Licensed training data/model and an appropriate compute budget
04 / FOLLOW THE EVIDENCE
The source trail.
Our notes are separate from the original resource.
Check upstream before adopting a new version.
- Pinned skill definition ↗Checked
https://github.com/huggingface/skills/blob/c3ff942be3e07116b475253f0a8fc4f54682c1a8/skills/huggingface-vision-trainer/SKILL.md
Supports: summary, whySelected, bestFor, limitations, install, access, review
https://github.com/huggingface/skills/blob/c3ff942be3e07116b475253f0a8fc4f54682c1a8/README.md
Supports: install
- Apache-2.0 licence ↗Checked
https://github.com/huggingface/skills/blob/c3ff942be3e07116b475253f0a8fc4f54682c1a8/LICENSE
Supports: license
https://github.com/huggingface/skills/blob/c3ff942be3e07116b475253f0a8fc4f54682c1a8/skills/huggingface-vision-trainer/scripts/object_detection_training.py
Supports: review, limitations, access, install