training_metadata.json

{
  "model": "Mini.Deep.Thinker.11m",
  "architecture": "decoder-only",
  "parameters": 11004896,
  "context_length": 1096,
  "strict_example_limit": 1095,
  "vocab_size": 4096,
  "dataset": "Plans11/Organized_PreTrain_1k_Context",
  "global_step": 1250,
  "unique_examples_trained": 40000,
  "last_session_size": 20000,
  "learning_rate": 0.0003,
  "batch_size": 8,
  "gradient_accumulation": 4,
  "weight_decay": 0.1,
  "seed": 1147,
  "reasoning_tokens": [
    "<|think|>",
    "<|thought|>",
    "<|reasoning|>",
    "<|answer|>"
  ],
  "causal_objective": "next_token_prediction",
  "tokenizer_immutable": true,
  "content_hash_deduplication": true,
  "shuffled_deterministic_scan": true,
  "best_session_loss": 1.889919606000185
}