| # Traditional Causal LM document packing (see dataset_document.py). | |
| # FineWeb defaults: config/data/fineweb_document.yaml (data=fineweb_document). | |
| format: document | |
| path: ??? # Override at CLI: data.path=/path/to/prepared_document_data | |
| val_path: null | |
| target_only: false | |