run_meta.json

{
  "base": "ProCreations/booper-pretrain",
  "dataset": "mookiezi/Discord-Dialogues",
  "token_budget": 150000000,
  "batch_size": 256,
  "learning_rate": 0.0001
}