{
  "seq_len_cap": 65536,
  "train_rows": 211837,
  "val_rows": 256,
  "train_tokens_total": 2535132827,
  "train_len_mean": 11967.4,
  "train_len_p50": 6030,
  "train_len_p90": 33309,
  "train_len_p99": 59792,
  "train_len_max": 65536,
  "packing_sim": {
    "32768": {
      "bins": 53444,
      "wasted_tokens": 783877001,
      "leftover_tokens": 2834,
      "steps_1ep": 3340,
      "waste_pct": 30.92,
      "tokens_per_step": 524288
    },
    "49152": {
      "bins": 40038,
      "wasted_tokens": 567182217,
      "leftover_tokens": 2834,
      "steps_1ep": 2502,
      "waste_pct": 22.37,
      "tokens_per_step": 786432
    },
    "65536": {
      "bins": 31773,
      "wasted_tokens": 452808294,
      "leftover_tokens": 49205,
      "steps_1ep": 3971,
      "waste_pct": 17.86,
      "tokens_per_step": 524288
    }
  },
  "stats": {
    "nemotron/raw": 204800,
    "nemotron/non_thinking": 541045,
    "nemotron/with_tool_calls": 1170572,
    "nemotron/with_tools_defs": 82481,
    "nemotron/thinking": 885735,
    "benchmaxx/raw": 12397,
    "benchmaxx/non_thinking": 12397,
    "benchmaxx_cot/raw": 15877,
    "benchmaxx_cot/non_thinking": 15877,
    "nemotron/overlong_tokens_dropped": 596520073
  },
  "drops": {
    "nemotron/bad_tools": 13542,
    "nemotron/assistant_without_gen_prompt": 355,
    "nemotron/think_tags_or_bad_tool_call": 21,
    "nemotron/overlong_gt_65536": 7063
  },
  "nemotron_categories": {
    "agentic": 48924,
    "math": 28672,
    "multilingual": 10240,
    "code": 28672,
    "science": 24576,
    "math_proofs": 8192,
    "arc_agi": 10240,
    "search": 4096,
    "swe": 16384,
    "cuda": 2276,
    "chat": 22528
  },
  "rescue_empty_string_tools": {
    "affected": 13542,
    "rescued": 13160,
    "new_train_rows": 224997,
    "new_train_tokens_total": 2592452318,
    "packing_sim_65536": {
      "bins": 32491,
      "wasted_tokens": 463121778,
      "leftover_tokens": 364,
      "steps_1ep": 4061,
      "waste_pct": 17.86
    }
  },
  "monster_filter": {
    "train_kept": 224991,
    "train_dropped": 6,
    "val_kept": 256,
    "val_dropped": 0,
    "train_tokens_total": 2592085287,
    "packing_sim_65536": {
      "bins": 32547,
      "wasted_tokens": 459066386,
      "leftover_tokens": 18709,
      "steps_1ep": 4068
    }
  }
}