{
  "_attn_implementation_autoset": false,
  "add_cross_attention": false,
  "architectures": [
    "GlmMoeDsaForCausalLM"
  ],
  "attention_bias": false,
  "attention_dropout": 0.0,
  "bos_token_id": null,
  "cross_attention_hidden_size": null,
  "decoder_start_token_id": null,
  "dtype": "bfloat16",
  "eos_token_id": [
    154820,
    154827,
    154829
  ],
  "ep_size": 1,
  "finetuning_task": null,
  "first_k_dense_replace": 3,
  "head_dim": 16,
  "hidden_act": "silu",
  "hidden_size": 512,
  "index_head_dim": 32,
  "index_n_heads": 4,
  "index_topk": 64,
  "indexer_rope_interleave": true,
  "initializer_range": 0.02,
  "intermediate_size": 256,
  "is_decoder": false,
  "kv_lora_rank": 32,
  "max_position_embeddings": 202752,
  "mlp_layer_types": [
    "dense",
    "dense",
    "dense",
    "sparse"
  ],
  "model_type": "glm_moe_dsa",
  "moe_intermediate_size": 256,
  "moe_layer_freq": 1,
  "n_group": 1,
  "n_routed_experts": 4,
  "n_shared_experts": 1,
  "norm_topk_prob": true,
  "num_attention_heads": 8,
  "num_experts_per_tok": 2,
  "num_hidden_layers": 4,
  "num_key_value_heads": 8,
  "num_nextn_predict_layers": 0,
  "pad_token_id": 154820,
  "prefix": null,
  "pretraining_tp": 1,
  "pruned_heads": {},
  "q_lora_rank": 64,
  "qk_head_dim": 48,
  "qk_nope_head_dim": 32,
  "qk_rope_head_dim": 16,
  "rms_norm_eps": 1e-05,
  "rope_interleave": true,
  "rope_parameters": {
    "rope_theta": 1000000,
    "rope_type": "default"
  },
  "routed_scaling_factor": 2.5,
  "scoring_func": "sigmoid",
  "sep_token_id": null,
  "task_specific_params": null,
  "tf_legacy_loss": false,
  "tie_encoder_decoder": false,
  "tie_word_embeddings": false,
  "tokenizer_class": null,
  "topk_group": 1,
  "topk_method": "noaux_tc",
  "torchscript": false,
  "transformers_version": "5.3.0",
  "use_bfloat16": false,
  "use_cache": true,
  "v_head_dim": 32,
  "vocab_size": 154880
}