Add Nanonets OCR script with vLLM support

- UV script for document OCR using Nanonets-OCR-s model
- Features: LaTeX equations, tables, document structure
- Supports batch processing with vLLM
- Includes HF Jobs examples for running on cloud
- Added proper CUDA checks and error handling

README.md

@@ -0,0 +1,134 @@
+---
+viewer: false
+tags: [uv-script, ocr, vision-language-model, document-processing]
+---
+
+# UV Scripts - OCR Collection
+
+This repository contains UV scripts for OCR (Optical Character Recognition) tasks using various models.
+
+## 🚧 Early Testing Version
+
+This is an early version for testing. Documentation and examples will be expanded based on feedback.
+
+## Available Scripts
+
+### 1. Nanonets OCR (`nanonets-ocr.py`)
+
+Converts document images to structured markdown using the Nanonets-OCR-s model.
+
+**Features:**
+- LaTeX equation recognition
+- Table extraction and formatting
+- Document structure preservation
+- Batch processing with vLLM
+
+**Requirements:**
+- GPU with CUDA support
+- Python 3.11+
+
+## Quick Test
+
+To test the script with a sample dataset:
+
+```bash
+# Test with 5 samples from a document dataset
+uv run nanonets-ocr.py \
+    davanstrien/scientific-papers-small \
+    my-test-ocr-output \
+    --max-samples 5
+
+# Or if you have a specific dataset with images
+uv run nanonets-ocr.py \
+    your-username/your-image-dataset \
+    your-username/test-ocr-results \
+    --image-column image \
+    --max-samples 10
+```
+
+## Example Output
+
+The script adds a `markdown` column to your dataset containing the extracted text in markdown format, preserving:
+- Headers and document structure
+- Tables with proper formatting
+- Mathematical equations in LaTeX
+- Lists and other formatting
+
+## GPU Memory
+
+If you encounter GPU memory issues, adjust the batch size and memory utilization:
+
+```bash
+uv run nanonets-ocr.py input output \
+    --batch-size 4 \
+    --gpu-memory-utilization 0.5
+```
+
+## Running on HuggingFace Jobs
+
+Run this script on HF infrastructure without needing your own GPU!
+
+### Command Line
+
+```bash
+# Basic usage
+hf jobs uv run --flavor l4x1 \
+    https://huggingface.co/datasets/uv-scripts/ocr/raw/main/nanonets-ocr.py \
+    input-dataset-id output-dataset-id
+
+# Full example with options
+hf jobs uv run --flavor l4x1 \
+    https://huggingface.co/datasets/uv-scripts/ocr/raw/main/nanonets-ocr.py \
+    NationalLibraryOfScotland/Scottish-School-Exam-Papers \
+    your-username/scottish-exams-ocr \
+    --image-column image \
+    --max-model-len 16384 \
+    --batch-size 16
+
+# With HF token for private repos
+hf jobs uv run --flavor l4x1 --secret HF_TOKEN=$HF_TOKEN \
+    https://huggingface.co/datasets/uv-scripts/ocr/raw/main/nanonets-ocr.py \
+    input-dataset output-dataset \
+    --private
+
+# With vLLM Docker image for optimized performance
+hf jobs uv run \
+    --flavor l4x1 \
+    --image vllm/vllm-openai:latest \
+    https://huggingface.co/datasets/uv-scripts/ocr/raw/main/nanonets-ocr.py \
+    input-dataset output-dataset \
+    --batch-size 32
+```
+
+### Python API
+
+```python
+from huggingface_hub import run_uv_job
+
+# Run the OCR script
+job = run_uv_job(
+    "https://huggingface.co/datasets/uv-scripts/ocr/raw/main/nanonets-ocr.py",
+    args=[
+        "input-dataset-id",
+        "output-dataset-id", 
+        "--image-column", "image",
+        "--max-model-len", "16384"
+    ],
+    flavor="l4x1",
+    secrets={"HF_TOKEN": "your-token"}  # if needed
+)
+```
+
+### Recommended GPU Flavors
+
+- **`l4x1`** (24GB) - Recommended for most OCR tasks
+- **`t4-small`** (16GB) - For smaller batches or lower resolution
+- **`a10g-small`** (24GB) - Alternative to L4
+- **`l40sx1`** (48GB) - For very large batches
+- **`a100-large`** (80GB) - Maximum performance
+
+## Coming Soon
+
+- Additional OCR models (RolmOCR, OlmOCR)
+- Performance benchmarks
+- More examples and use cases