config.json

{
  "adaptive_halting": true,
  "architectures": [
    "HyperMiniReasoner"
  ],
  "base_layers": 6,
  "context_length": 1096,
  "dataset": "Plans11/Organized_PreTrain_1k_Context",
  "head_dimension": 30,
  "hidden_size": 240,
  "intermediate_size": 1072,
  "max_position_embeddings": 1096,
  "max_reasoning_steps": 16,
  "memory_tokens": 8,
  "min_reasoning_steps": 2,
  "model_name": "Nano.Deep.Reasoner.11m-HyperMini",
  "model_type": "nano_deep_reasoner_hypermini",
  "num_attention_heads": 8,
  "padding_aware_attention": true,
  "padding_query_zeroing": true,
  "parameter_count": 11094003,
  "reasoning_tokens": [
    "<pad>",
    "<bos>",
    "<eos>",
    "<unk>",
    "<think>",
    "</think>",
    "<thought>",
    "</thought>",
    "<reasoning>",
    "</reasoning>",
    "<answer>",
    "</answer>",
    "<question>",
    "</question>",
    "<analysis>",
    "</analysis>",
    "<tool>",
    "</tool>"
  ],
  "revision_head": true,
  "rope": true,
  "shared_recurrent_reasoner": true,
  "tie_word_embeddings": true,
  "verification_head": true,
  "vocab_size": 16000
}