About 13.5M tokens total of mixed instruct and RP data.

Both RP datasets and the inkstruct include system prompts to help g3 understand the system role (via <start_of_turn>system).

datasets:
  - path: ToastyPigeon/some-rp-extended
    type: customgemma-regex
  - path: allura-org/inkstructmix-v0.2.1a-system-reasoning-separated
    type: customgemma-regex
    data_files: inkstruct-system.json
    split: train[:750]
  - path: ToastyPigeon/unalign-v2
    type: customgemma-regex
    split: train[:50%]
  - path: ToastyPigeon/synth-rp
    split: train[:20%]
    type: customgemma-regex