{
  "_name_or_path": "./outputs/phi-reasoning-instruct-fixed",
  "architectures": [
    "PhiForLogicalReasoning"
  ],
  "attention_dropout": 0.0,
  "attn_implementation": "eager",
  "bos_token_id": 50256,
  "consistency_loss_weight": 0.001,
  "embd_pdrop": 0.0,
  "eos_token_id": 50256,
  "gating_threshold": 0.3,
  "halting_threshold": 0.8,
  "head_dim": 80,
  "hidden_act": "gelu_new",
  "hidden_size": 2560,
  "initializer_range": 0.02,
  "intermediate_size": 10240,
  "layer_norm_eps": 1e-05,
  "max_position_embeddings": 2048,
  "max_reasoning_steps": 16,
  "min_reasoning_steps": 2,
  "model_type": "phi_reasoning",
  "num_attention_heads": 32,
  "num_hidden_layers": 32,
  "num_key_value_heads": 32,
  "num_reasoning_layers": 4,
  "num_reasoning_steps": 8,
  "num_reasoning_tokens": 32,
  "pad_token_id": null,
  "partial_rotary_factor": 0.4,
  "ponder_loss_weight": 0.01,
  "qk_layernorm": false,
  "reasoning_dropout": 0.1,
  "reasoning_injection_point": 16,
  "reasoning_intermediate_size": 2560,
  "resid_pdrop": 0.1,
  "rope_parameters": {
    "partial_rotary_factor": 0.4,
    "rope_theta": 10000.0,
    "rope_type": "default"
  },
  "rope_scaling": null,
  "rope_theta": 10000.0,
  "share_reasoning_layers": false,
  "tie_word_embeddings": false,
  "torch_dtype": "bfloat16",
  "transformers_version": "4.46.0",
  "use_adaptive_halting": true,
  "use_cache": false,
  "use_input_gating": true,
  "vocab_size": 51200,
  "auto_map": {
    "AutoConfig": "configuration.PhiReasoningConfig",
    "AutoModelForCausalLM": "modeling.PhiForLogicalReasoning"
  }
}