README.md

metadata

license: apache-2.0
task_categories:
  - visual-question-answering
  - image-to-text
language:
  - en
  - pt
tags:
  - vlm
  - paligemma
  - vision-language-model
  - elementor
  - web-design
  - layout-generation
pretty_name: Elementor Layout VLM Dataset
size_categories:
  - n<1K

Elementor Layout VLM Dataset

📊 Dataset Summary

Dataset para fine-tuning de modelos Vision-Language (VLM) para geração de layouts Elementor a partir de imagens.

• Task: Visual Question Answering (VQA)
• Format: VLM VQA (image, question, answer)
• Total: 30 exemplos
• Training: 24 exemplos
• Validation: 6 exemplos

🎯 Uso Recomendado

AutoTrain Configuration

Task: VLM VQA
Base Model: google/paligemma-3b-pt-448
Dataset: vinicios94/elementor-layout-vlm-dataset

Column Mapping:
  image: image
  prompt: question
  text: answer

Load Dataset

from datasets import load_dataset

dataset = load_dataset("vinicios94/elementor-layout-vlm-dataset")
example = dataset['train'][0]

print(f"Question: {example['question']}")
print(f"Answer: {example['answer'][:200]}...")

📝 Dataset Structure

Cada exemplo contém:

• image: Screenshot de layout (PIL Image)
• question: Prompt para gerar JSON
• answer: JSON válido do Elementor

🚀 Fine-tuning

Otimizado para:

• Modelo: google/paligemma-3b-pt-448
• Hardware: Nvidia L4 (24GB VRAM)
• Batch Size: 2
• LoRA: r=16, alpha=32
• Quantization: int4

📄 License

Apache 2.0