config.json

{
  "architectures": [
    "CubitForCausalLM"
  ],
  "attention_bias": false,
  "auto_map": {
    "AutoConfig": "configuration_cubit.CubitConfig",
    "AutoModel": "modeling_cubit.CubitModel",
    "AutoModelForCausalLM": "modeling_cubit.CubitForCausalLM"
  },
  "bos_token_id": null,
  "dtype": "bfloat16",
  "eos_token_id": 100257,
  "head_dim": 128,
  "hidden_act": "silu",
  "hidden_size": 2048,
  "initializer_range": 0.02,
  "intermediate_size": 8192,
  "krr_block_size": 128,
  "krr_implementation": "streaming",
  "layer_types": [
    "sliding_attention",
    "sliding_attention",
    "sliding_attention",
    "full_attention",
    "sliding_attention",
    "sliding_attention",
    "sliding_attention",
    "full_attention",
    "sliding_attention",
    "sliding_attention",
    "sliding_attention",
    "full_attention",
    "sliding_attention",
    "sliding_attention",
    "sliding_attention",
    "full_attention"
  ],
  "max_position_embeddings": 65536,
  "mlp_bias": false,
  "model_type": "cubit",
  "num_attention_heads": 16,
  "num_hidden_layers": 16,
  "num_key_value_heads": 16,
  "pad_token_id": 100277,
  "reference_norm_eps": 1e-06,
  "rms_norm_eps": 1e-06,
  "rope_theta": 500000,
  "share_reference": false,
  "sliding_window": 4096,
  "tie_word_embeddings": false,
  "transformers_version": "5.4.0",
  "use_cache": false,
  "vocab_size": 100352
}