config.json
| 1 | { |
| 2 | "architectures": [ |
| 3 | "Glm4MoeForCausalLM" |
| 4 | ], |
| 5 | "attention_bias": true, |
| 6 | "attention_dropout": 0.0, |
| 7 | "auto_map": { |
| 8 | "AutoConfig": "configuration_glm4_moe.Glm4MoeConfig", |
| 9 | "AutoModel": "modeling_glm4_moe.Glm4MoeModel", |
| 10 | "AutoModelForCausalLM": "modeling_glm4_moe.Glm4MoeForCausalLM" |
| 11 | }, |
| 12 | "dtype": "bfloat16", |
| 13 | "eos_token_id": [ |
| 14 | 151334, |
| 15 | 151329 |
| 16 | ], |
| 17 | "first_k_dense_replace": 1, |
| 18 | "head_dim": 128, |
| 19 | "hidden_act": "silu", |
| 20 | "hidden_size": 4096, |
| 21 | "initializer_range": 0.02, |
| 22 | "intermediate_size": 10944, |
| 23 | "max_position_embeddings": 131072, |
| 24 | "model_type": "glm4_moe", |
| 25 | "moe_intermediate_size": 1408, |
| 26 | "n_group": 1, |
| 27 | "n_routed_experts": 128, |
| 28 | "n_shared_experts": 1, |
| 29 | "norm_topk_prob": true, |
| 30 | "num_attention_heads": 96, |
| 31 | "num_experts_per_tok": 8, |
| 32 | "num_hidden_layers": 46, |
| 33 | "num_key_value_heads": 8, |
| 34 | "num_nextn_predict_layers": 1, |
| 35 | "pad_token_id": 151329, |
| 36 | "partial_rotary_factor": 0.5, |
| 37 | "rms_norm_eps": 1e-05, |
| 38 | "rope_scaling": null, |
| 39 | "rope_theta": 1000000, |
| 40 | "routed_scaling_factor": 1.0, |
| 41 | "tie_word_embeddings": false, |
| 42 | "topk_group": 1, |
| 43 | "transformers_version": "4.56.1", |
| 44 | "use_cache": false, |
| 45 | "use_grouped_mm": true, |
| 46 | "use_qk_norm": false, |
| 47 | "vocab_size": 151552 |
| 48 | } |
| 49 | |