Hands-on tutorials for curating interleaved multimodal data — documents that contain text, images, and metadata interlaced in reading order — with NeMo Curator.
| Tutorial | Description | Files |
|---|---|---|
| Getting Started | Load, explore, filter, and save interleaved data from MINT-1T PDF shards | interleaved_data_quickstart.ipynb, interleaved_pipeline.py |
| PDF Extraction Pipeline (Nemotron-Parse) | Convert PDFs into structured interleaved Parquet using Nemotron-Parse v1.2 | main.py |
New to interleaved multimodal curation? Start with the Getting Started notebook, or run the pipeline script directly on local data:
python tutorials/interleaved/getting-started/interleaved_pipeline.py \
--input-path /path/to/shard-0/ \
--output-path /path/to/output/ \
--on-materialize-error drop_row \
--mode overwrite| Category | Links |
|---|---|
| Concepts | Core Concepts |
| API Reference | API Docs |
Documentation: Main Docs • GitHub Discussions