{
  "architecture_options": {
    "attention_backend": "flash_2",
    "attention_bias": false,
    "attn_qk_norm": true,
    "d_model": 1536,
    "enable_swa": true,
    "ffn_bias": false,
    "ffn_hidden": 6144,
    "gdn_allow_neg_eigval": true,
    "gdn_conv_bias": false,
    "gdn_conv_size": 4,
    "gdn_expand_v": 2.0,
    "head_dim": 128,
    "layer_norm_bias": false,
    "layer_norm_eps": 1e-05,
    "lm_head_bias": false,
    "n_heads": 12,
    "n_kv_heads": 4,
    "n_layers": 26,
    "rope_full_precision": true,
    "rope_theta": 500000,
    "swa_full_attention_first_layer": false,
    "swa_full_attention_last_layer": false,
    "swa_window": 4096,
    "tie_word_embeddings": true
  },
  "architecture_preset": "main",
  "architectures": [
    "NovaForCausalLM"
  ],
  "auto_map": {
    "AutoConfig": "configuration_nova.NovaConfig",
    "AutoModel": "modeling_nova.NovaModel",
    "AutoModelForCausalLM": "modeling_nova.NovaForCausalLM"
  },
  "bos_token_id": 1,
  "compute_dtype": "bfloat16",
  "dtype": "float32",
  "eos_token_id": 2,
  "fast_decode": true,
  "generation_context_policy": "remaining",
  "hidden_size": 1536,
  "inference_backend": "torch",
  "max_position_embeddings": 8192,
  "model_type": "nova_gdn2",
  "num_hidden_layers": 26,
  "pad_token_id": 0,
  "tokenizer_vocab_size": 65536,
  "transformers_version": "4.57.6",
  "use_cache": false,
  "vocab_size": 65536
}