huggingface-vision-trainer

Train and fine-tune vision models on Hugging Face Jobs cloud GPUs.

Updated Mar 8, 2026
One-click install
npx skills add https://github.com/ClyptAI/Clypt-Backend --skill huggingface-vision-trainer-clyptai
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: huggingface-vision-trainer
Source: https://github.com/ClyptAI/Clypt-Backend/tree/main/.agents/skills/hf-skills/skills/huggingface-vision-trainer
Command: npx skills add https://github.com/ClyptAI/Clypt-Backend --skill huggingface-vision-trainer-clyptai

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires transformers, accelerate, albumentations, timm, datasets, torchmetrics, pycocotools, trackio, huggingface_hub, torchvision, monai, evaluate, scikit-learn, and includes scripts (resource) and references (resource) components.

What problem does it solve?

Vision model training at scale can be complex; this skill streamlines end-to-end vision model training on Hugging Face Jobs with managed compute, data prep, and Hub deployment.

Core Features & Use Cases

  • Supports object detection, image classification, and SAM/SAM2 fine-tuning using HF Transformers on cloud GPUs.
  • Includes COCO-format dataset prep, Albumentations augmentations, mAP/mAR evaluation, and Trackio monitoring, with Hub persistence.
  • Suitable for rapid experimentation on datasets like cppe-5, ethz/food101, and MicroMat-mini, and for deploying results to HF Hub.

Quick Start

Run a training job with an HF training script (object detection, image classification, or SAM) on Hugging Face Jobs and push the result to the Hub.

Frequently Asked Questions about huggingface-vision-trainer

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I fine-tune an object detection model using Hugging Face Jobs?

Fine-tuning object detection models on Hugging Face Jobs uses provided scripts supporting DFINE, RT-DETR, DETR, and YOLOS. The process handles COCO-format dataset preparation, Albumentations augmentations, mAP/mAR evaluation, and Hub persistence.

Can I train image classification models with TIMM on cloud GPUs?

Yes, image classification models using TIMM-based classifiers are supported on Hugging Face Jobs cloud GPUs. The workflow applies Albumentations augmentations and torchmetrics evaluation before pushing the trained model to the Hub.

Does this approach support SAM2 segmentation fine-tuning?

SAM and SAM2 segmentation fine-tuning is fully supported on Hugging Face Jobs cloud GPUs. The workflow includes dataset validation and Trackio monitoring throughout the training and Hub push process.

How do I prepare a COCO-format dataset for vision training?

Dataset preparation for vision training uses included scripts to validate COCO-format data and apply Albumentations augmentations. This ensures compatibility with object detection models like RT-DETR and YOLOS before Hub push.

What's the best way to evaluate vision models using mAP and mAR?

Evaluating vision models with mAP and mAR is handled via pycocotools and torchmetrics during Hugging Face Jobs training. This provides standard object detection metrics before pushing the evaluated model to the Hub.

Why does vision model training fail when pushing results to the Hub?

Vision model training Hub push failures can occur if dataset validation or hub-push configurations are missing. The included training scripts and References folder enforce required dependencies, hardware selection, and timeout guidance.