config.json

{
  "architectures": [
    "GlmMoeDsaForCausalLM"
  ],
  "attention_bias": false,
  "attention_dropout": 0.0,
  "dtype": "float32",
  "ep_size": 1,
  "first_k_dense_replace": 1,
  "head_dim": 16,
  "hidden_act": "silu",
  "hidden_size": 64,
  "intermediate_size": 128,
  "max_position_embeddings": 256,
  "model_type": "glm_moe_dsa",
  "moe_intermediate_size": 32,
  "moe_layer_freq": 1,
  "moe_router_dtype": "float32",
  "n_group": 1,
  "n_routed_experts": 8,
  "n_shared_experts": 1,
  "norm_topk_prob": true,
  "num_attention_heads": 4,
  "num_experts_per_tok": 2,
  "num_hidden_layers": 4,
  "num_key_value_heads": 4,
  "num_nextn_predict_layers": 0,
  "pad_token_id": 0,
  "rms_norm_eps": 1e-05,
  "rope_theta": 10000.0,
  "tie_word_embeddings": false,
  "vocab_size": 256
}