{
  "release_format": "native-safetensors-v1",
  "architectures": [
    "smollm2_135m_trigger_v2_quadorbit_lm"
  ],
  "model_type": "smollm2_135m_trigger_v2_quadorbit_lm",
  "model_config": {
    "vocab_size": 49152,
    "d_model": 576,
    "n_layers": 30,
    "max_seq_len": 8192,
    "n_heads": 9,
    "n_kv_heads": 3,
    "head_dim": 64,
    "d_ffn": 1536,
    "rope_theta": 100000.0,
    "qk_norm": false,
    "native_gqa": false,
    "rms_eps": 1e-05,
    "init_std": 0.041666666666666664,
    "tie_embeddings": true,
    "grad_checkpoint": true,
    "loss_chunk_size": 2048,
    "variant": "v2",
    "trigger_token_ids": [
      1510,
      3039,
      6545,
      12528,
      16597,
      16817,
      17400,
      18322,
      27048,
      35682,
      37406,
      46481
    ],
    "trigger_text": [
      " think",
      " thinking",
      " Think",
      " thinks",
      " thinkers",
      "Think",
      "think",
      " Thinking",
      "thinking",
      " thinker",
      " rethink",
      "Thinking"
    ],
    "orbit_width": 32,
    "beta_init": 0.1,
    "beta_max": 0.5,
    "branch_strength_init": 0.25,
    "branch_strength_max": 1.0,
    "context_training": true,
    "context_hint_tokens": 8,
    "branch_context_scale": 1.0,
    "context_contrastive_weight": 0.5,
    "general_trigger_training": true,
    "trigger_training_stride": 128
  },
  "checkpoint_step": 1999,
  "parameter_count": 134939649,
  "branch_parameter_count": 424641,
  "torch_dtype": "float32",
  "library_name": "pytorch"
}