{
  "architectures": [
    "HetmanForCausalLM"
  ],
  "auto_map": {
    "AutoConfig": "configuration_hetman.HetmanConfig",
    "AutoModel": "modeling_hetman.HetmanModel",
    "AutoModelForCausalLM": "modeling_hetman.HetmanForCausalLM"
  },
  "model_type": "hetman",
  "vocab_size": 65536,
  "bos_token_id": 1,
  "eos_token_id": 5,
  "pad_token_id": 3,
  "hidden_size": 1536,
  "intermediate_size": 6144,
  "num_hidden_layers": 34,
  "num_deltanet_layers": 28,
  "num_global_layers": 6,
  "num_attention_heads": 16,
  "deltanet_head_dim": 64,
  "deltanet_conv_kernel": 4,
  "global_head_dim": 128,
  "global_attention_heads": 12,
  "global_kv_heads": 2,
  "hopfield_num_groups": 16,
  "hopfield_codebook_size": 1024,
  "hopfield_val_dim": 32,
  "hopfield_expanded_dim": 128,
  "hopfield_topk_1": 8,
  "hopfield_topk_2": 4,
  "hopfield_shared_attractors": 1,
  "hopfield_subquery_dim": 2048,
  "hopfield_beta": 0.1767767,
  "hopfield_pid_ki": 0.01,
  "hopfield_pid_kd": 0.005,
  "hadamard_block_size": 512,
  "swiglu_limit": 10.0,
  "logit_softcap": 30.0,
  "max_position_embeddings": 262144,
  "qk_norm_bound": 11.3137,
  "rms_norm_eps": 1e-06,
  "rope_theta": 10000000.0,
  "ternary_bits": 1.58,
  "use_hadamard_rotation": true,
  "torch_dtype": "bfloat16",
  "transformers_version": "4.45.0"
}