| dataset_info: | |
| features: | |
| - name: input_ids | |
| sequence: int32 | |
| splits: | |
| - name: train | |
| num_bytes: 5881094452.0 | |
| num_examples: 5720909 | |
| - name: test | |
| num_bytes: 309531828.0 | |
| num_examples: 301101 | |
| download_size: 3050952379 | |
| dataset_size: 6190626280.0 | |
| configs: | |
| - config_name: default | |
| data_files: | |
| - split: train | |
| path: data/train-* | |
| - split: test | |
| path: data/test-* | |
| The first 15% of Openwebtext2 tokenized with LLamaTokenizer for 256 context length. Total of ~1.5 Billion tokens. |