data/sources_eval_base.json

{
  "version": 2,
  "name": "training_v2-teutonic-II-110B",
  "description": "Teutonic-II-110B FineWeb-Edu only. eval_weight synced from DATASETS_MANIFEST_URL / EVAL_MANIFEST_URL. Train weight from train_init_weights.json v5. Shard bytes: prefer R2 via download_finewebedu.py; HF mirror bucket stevendev/corpus for prepare_from_hf.",
  "seq_len": 2048,
  "dtype": "uint32",
  "train_seed": 86,
  "eval_seed": 1234,
  "eval_total": 2000,
  "default_train_sequences": 5000000,
  "eval_manifest_url": "https://pub-fedac496355c4edc9aed57189e6e190f.r2.dev/datasets/manifest.json",
  "datasets_manifest_url": "https://pub-fedac496355c4edc9aed57189e6e190f.r2.dev/datasets/manifest.json",
  "tokenizer": "XiaomiMiMo/MiMo-V2.5-Pro",
  "remote_name_map": {
    "finewebedu": "finewebedu"
  },
  "sources": [
    {
      "name": "finewebedu",
      "weight": 1.0,
      "eval_weight": 1.0,
      "num_shards": 200,
      "eval_num_shards": 80,
      "bucket": "stevendev/corpus",
      "manifest": "dataset/finewebedu/manifest.json",
      "r2_manifest_url": "https://pub-d923bc4e8fcb45f6b703bc750bcf8aa6.r2.dev/finewebedu/manifest.json"
    }
  ]
}