tokenization_stats.json

{
  "raw_rows_seen": 7085839,
  "invalid_rows": 1716489,
  "over_cutoff_rows": 365789,
  "kept_rows_before_packing": 5003561,
  "packed_rows_written": 1085226,
  "cutoff_len": 20479,
  "template": "qwen3",
  "enable_thinking": null,
  "truncate_mode": "drop",
  "packing": true,
  "neat_packing": true
}