README.md

metadata

tags:
  - ocr
  - document-processing
  - smoldocling
  - doctags
  - structured-extraction
  - uv-script
  - generated

Document Processing using SmolDocling-256M-preview

This dataset contains structured document extraction from images in davanstrien/ocr-affordances-pages using SmolDocling.

Processing Details

• Source Dataset: davanstrien/ocr-affordances-pages
• Model: ds4sd/SmolDocling-256M-preview
• Number of Samples: 8
• Processing Time: 1.9 minutes
• Processing Date: 2026-06-05 14:15 UTC

Configuration

• Image Column: image
• Output Column: markdown
• Output Format: markdown
• Dataset Split: train
• Batch Size: 32
• Max Model Length: 8,192 tokens
• Max Output Tokens: 8,192
• GPU Memory Utilization: 80.0%

Model Information

SmolDocling-256M is an ultra-compact multimodal model that excels at:

• 💻 Code Recognition - Detects and formats code blocks with proper indentation
• 🔢 Formula Recognition - Identifies and processes mathematical expressions
• 📊 Tables & Charts - Extracts structured data from tables and charts
• 📐 Layout Preservation - Maintains document structure with bounding boxes
• 🏷️ DocTags Format - Efficient minimal representation for documents
• ⚡ Fast Inference - Only 256M parameters for quick processing

Dataset Structure

The dataset contains all original columns plus:

• markdown: The extracted markdown from each image
• inference_info: JSON list tracking all OCR models applied to this dataset

Usage

from datasets import load_dataset
import json



# Load the dataset
dataset = load_dataset("{output_dataset_id}", split="train")

# Access the extracted content
for example in dataset:
    
    print(example['markdown'])
    
    
    break

# View all OCR models applied to this dataset
inference_info = json.loads(dataset[0]["inference_info"])
for info in inference_info:
    print(f"Column: {info['column_name']} - Model: {info['model_id']}")

Reproduction

This dataset was generated using the uv-scripts/ocr SmolDocling script:

uv run https://huggingface.co/datasets/uv-scripts/ocr/raw/main/smoldocling-ocr.py \
    davanstrien/ocr-affordances-pages \
    <output-dataset> \
    --image-column image \
    --output-format markdown \
    --batch-size 32 \
    --max-model-len 8192 \
    --max-tokens 8192 \
    --gpu-memory-utilization 0.8

Performance

• Processing Speed: ~0.1 images/second
• Model Size: 256M parameters (ultra-compact)
• GPU Configuration: vLLM with 80% GPU memory utilization

Generated with 🤖 UV Scripts