{
  "model_type": "corm",
  "architectures": [
    "CoRMForCausalLM"
  ],
  "auto_map": {
    "AutoConfig": "configuration_corm.CoRMConfig",
    "AutoModelForCausalLM": "modeling_corm.CoRMForCausalLM"
  },
  "hidden_size": 768,
  "num_hidden_layers": 12,
  "num_attention_heads": 12,
  "num_key_value_heads": 4,
  "intermediate_size": 3072,
  "num_local_experts": 8,
  "num_experts_per_tok": 1,
  "use_dynamic_routing": false,
  "top_p": 1.0,
  "expert_attn_size": 64,
  "max_position_embeddings": 1024,
  "vocab_size": 51200,
  "rms_norm_eps": 1e-05,
  "rope_theta": 1000000.0,
  "routing_temperature_init": 1.0,
  "router_type": "corm",
  "router_background_momentum": 0.01,
  "moe_router_use_l2_normalize": true,
  "has_shared_expert": false,
  "learn_output_weights": false,
  "fixed_output_weights": false,
  "torch_dtype": "bfloat16",
  "tie_word_embeddings": false
}