project:
  seed: 1945
  era_start: 1868
  era_end: 1945
  reports_dir: reports
data:
  index_csv: data/index/list_person_all_extended_utf8.csv
  aozora_repo: data/raw/aozora
  raw_manifest: data/processed/aozora/selection.jsonl
  holdout_manifest: data/processed/aozora/holdout.jsonl
  clean_dir: data/clean/aozora
  clean_manifest: data/processed/aozora/clean.jsonl
  aozora_split_manifest: data/processed/aozora/splits.jsonl
  split_manifest: data/processed/aozora/splits.jsonl
  min_clean_chars: 200
split:
  train: 0.96
  validation: 0.02
  test: 0.02
tokenizer:
  vocab_size: 8192
  min_frequency: 2
  directory: artifacts/tokenizer
tokenized:
  directory: data/processed/tokenized
huggingface:
  namespace: "trtd56"
  private: true
  dataset_repos:
    phase1: "trtd56/TimeCapsuleLLM-ja-corpus-v0"
    phase2: "trtd56/TimeCapsuleLLM-ja-corpus-v1-small"
    phase3: "trtd56/TimeCapsuleLLM-ja-corpus-v1"
    phase4: "trtd56/TimeCapsuleLLM-ja-corpus-v2"
  model_repos:
    phase1: "trtd56/TimeCapsuleLLM-ja-v0"
    phase2: "trtd56/TimeCapsuleLLM-ja-v1-small"
    phase3: "trtd56/TimeCapsuleLLM-ja-v1"
    phase4: "trtd56/TimeCapsuleLLM-ja-v2"
  licenses:
    dataset: other
    model: apache-2.0