| { |
| "architectures": [ |
| "GlmMoeDsaForCausalLM" |
| ], |
| "attention_bias": false, |
| "attention_dropout": 0.0, |
| "dtype": "bfloat16", |
| "eos_token_id": [ |
| 154820, |
| 154827, |
| 154829 |
| ], |
| "ep_size": 1, |
| "first_k_dense_replace": 2, |
| "head_dim": 64, |
| "hidden_act": "silu", |
| "hidden_size": 6144, |
| "index_head_dim": 128, |
| "index_n_heads": 32, |
| "index_share_for_mtp_iteration": true, |
| "index_skip_topk_offset": 3, |
| "index_topk": 2048, |
| "index_topk_freq": 4, |
| "index_topk_pattern": null, |
| "indexer_rope_interleave": true, |
| "indexer_types": [ |
| "full", |
| "full", |
| "shared", |
| "shared", |
| "shared", |
| "full", |
| "shared", |
| "shared", |
| "shared", |
| "full", |
| "shared", |
| "shared", |
| "shared", |
| "full" |
| ], |
| "initializer_range": 0.02, |
| "intermediate_size": 12288, |
| "kv_lora_rank": 512, |
| "max_position_embeddings": 16384, |
| "mlp_layer_types": [ |
| "dense", |
| "dense", |
| "sparse", |
| "sparse", |
| "sparse", |
| "sparse", |
| "sparse", |
| "sparse", |
| "sparse", |
| "sparse", |
| "sparse", |
| "sparse", |
| "sparse", |
| "sparse" |
| ], |
| "model_type": "glm_moe_dsa", |
| "moe_intermediate_size": 2048, |
| "moe_layer_freq": 1, |
| "n_group": 1, |
| "n_routed_experts": 256, |
| "n_shared_experts": 1, |
| "norm_topk_prob": true, |
| "num_attention_heads": 64, |
| "num_experts_per_tok": 8, |
| "num_hidden_layers": 14, |
| "num_key_value_heads": 64, |
| "num_nextn_predict_layers": 1, |
| "pad_token_id": 154820, |
| "pretraining_tp": 1, |
| "q_lora_rank": 2048, |
| "qk_head_dim": 256, |
| "qk_nope_head_dim": 192, |
| "qk_rope_head_dim": 64, |
| "quantization_config": { |
| "activation_scheme": "dynamic", |
| "fmt": "e4m3", |
| "modules_to_not_convert": [ |
| "lm_head", |
| "model.embed_tokens", |
| "model.layers.0.input_layernorm", |
| "model.layers.0.post_attention_layernorm", |
| "model.layers.0.self_attn.indexer.k_norm", |
| "model.layers.0.self_attn.indexer.weights_proj", |
| "model.layers.0.self_attn.kv_a_layernorm", |
| "model.layers.0.self_attn.q_a_layernorm", |
| "model.layers.1.input_layernorm", |
| "model.layers.1.post_attention_layernorm", |
| "model.layers.1.self_attn.indexer.k_norm", |
| "model.layers.1.self_attn.indexer.weights_proj", |
| "model.layers.1.self_attn.kv_a_layernorm", |
| "model.layers.1.self_attn.q_a_layernorm", |
| "model.layers.10.input_layernorm", |
| "model.layers.10.mlp.gate", |
| "model.layers.10.post_attention_layernorm", |
| "model.layers.10.self_attn.indexer.k_norm", |
| "model.layers.10.self_attn.indexer.weights_proj", |
| "model.layers.10.self_attn.kv_a_layernorm", |
| "model.layers.10.self_attn.q_a_layernorm", |
| "model.layers.11.input_layernorm", |
| "model.layers.11.mlp.gate", |
| "model.layers.11.post_attention_layernorm", |
| "model.layers.11.self_attn.indexer.k_norm", |
| "model.layers.11.self_attn.indexer.weights_proj", |
| "model.layers.11.self_attn.kv_a_layernorm", |
| "model.layers.11.self_attn.q_a_layernorm", |
| "model.layers.12.input_layernorm", |
| "model.layers.12.mlp.gate", |
| "model.layers.12.post_attention_layernorm", |
| "model.layers.12.self_attn.indexer.k_norm", |
| "model.layers.12.self_attn.indexer.weights_proj", |
| "model.layers.12.self_attn.kv_a_layernorm", |
| "model.layers.12.self_attn.q_a_layernorm", |
| "model.layers.13.input_layernorm", |
| "model.layers.13.mlp.gate", |
| "model.layers.13.post_attention_layernorm", |
| "model.layers.13.self_attn.indexer.k_norm", |
| "model.layers.13.self_attn.indexer.weights_proj", |
| "model.layers.13.self_attn.kv_a_layernorm", |
| "model.layers.13.self_attn.q_a_layernorm", |
| "model.layers.14.eh_proj", |
| "model.layers.14.enorm", |
| "model.layers.14.hnorm", |
| "model.layers.14.input_layernorm", |
| "model.layers.14.mlp.gate", |
| "model.layers.14.post_attention_layernorm", |
| "model.layers.14.self_attn.indexer.k_norm", |
| "model.layers.14.self_attn.indexer.weights_proj", |
| "model.layers.14.self_attn.kv_a_layernorm", |
| "model.layers.14.self_attn.q_a_layernorm", |
| "model.layers.14.shared_head.norm", |
| "model.layers.2.input_layernorm", |
| "model.layers.2.mlp.gate", |
| "model.layers.2.post_attention_layernorm", |
| "model.layers.2.self_attn.indexer.k_norm", |
| "model.layers.2.self_attn.indexer.weights_proj", |
| "model.layers.2.self_attn.kv_a_layernorm", |
| "model.layers.2.self_attn.q_a_layernorm", |
| "model.layers.3.input_layernorm", |
| "model.layers.3.mlp.gate", |
| "model.layers.3.post_attention_layernorm", |
| "model.layers.3.self_attn.indexer.k_norm", |
| "model.layers.3.self_attn.indexer.weights_proj", |
| "model.layers.3.self_attn.kv_a_layernorm", |
| "model.layers.3.self_attn.q_a_layernorm", |
| "model.layers.4.input_layernorm", |
| "model.layers.4.mlp.gate", |
| "model.layers.4.post_attention_layernorm", |
| "model.layers.4.self_attn.indexer.k_norm", |
| "model.layers.4.self_attn.indexer.weights_proj", |
| "model.layers.4.self_attn.kv_a_layernorm", |
| "model.layers.4.self_attn.q_a_layernorm", |
| "model.layers.5.input_layernorm", |
| "model.layers.5.mlp.gate", |
| "model.layers.5.post_attention_layernorm", |
| "model.layers.5.self_attn.indexer.k_norm", |
| "model.layers.5.self_attn.indexer.weights_proj", |
| "model.layers.5.self_attn.kv_a_layernorm", |
| "model.layers.5.self_attn.q_a_layernorm", |
| "model.layers.6.input_layernorm", |
| "model.layers.6.mlp.gate", |
| "model.layers.6.post_attention_layernorm", |
| "model.layers.6.self_attn.indexer.k_norm", |
| "model.layers.6.self_attn.indexer.weights_proj", |
| "model.layers.6.self_attn.kv_a_layernorm", |
| "model.layers.6.self_attn.q_a_layernorm", |
| "model.layers.7.input_layernorm", |
| "model.layers.7.mlp.gate", |
| "model.layers.7.post_attention_layernorm", |
| "model.layers.7.self_attn.indexer.k_norm", |
| "model.layers.7.self_attn.indexer.weights_proj", |
| "model.layers.7.self_attn.kv_a_layernorm", |
| "model.layers.7.self_attn.q_a_layernorm", |
| "model.layers.8.input_layernorm", |
| "model.layers.8.mlp.gate", |
| "model.layers.8.post_attention_layernorm", |
| "model.layers.8.self_attn.indexer.k_norm", |
| "model.layers.8.self_attn.indexer.weights_proj", |
| "model.layers.8.self_attn.kv_a_layernorm", |
| "model.layers.8.self_attn.q_a_layernorm", |
| "model.layers.9.input_layernorm", |
| "model.layers.9.mlp.gate", |
| "model.layers.9.post_attention_layernorm", |
| "model.layers.9.self_attn.indexer.k_norm", |
| "model.layers.9.self_attn.indexer.weights_proj", |
| "model.layers.9.self_attn.kv_a_layernorm", |
| "model.layers.9.self_attn.q_a_layernorm", |
| "model.norm" |
| ], |
| "quant_method": "fp8", |
| "weight_block_size": [ |
| 128, |
| 128 |
| ] |
| }, |
| "rms_norm_eps": 1e-05, |
| "rope_interleave": true, |
| "rope_parameters": { |
| "rope_theta": 8000000, |
| "rope_type": "default" |
| }, |
| "routed_scaling_factor": 2.5, |
| "scoring_func": "sigmoid", |
| "tie_word_embeddings": false, |
| "topk_group": 1, |
| "topk_method": "noaux_tc", |
| "torch_dtype": "bfloat16", |
| "transformers_version": "5.12.0", |
| "use_cache": true, |
| "v1c_full_upscale": { |
| "active_attention_heads": 16, |
| "active_hidden_size": 1024, |
| "active_routed_experts": 16, |
| "expansion_strategy": "tiled active subspace with averaged input repeats", |
| "inactive_expert_bias": -1000000000.0, |
| "mtp_source_checkpoint": "/root/inference-v2/mini-glm-5.2/v1b-mtp/checkpoints/mtp-active-decoder-serveddata-align-1k/final", |
| "num_nextn_predict_layers": 1, |
| "source_checkpoint": "/root/inference-v2/mini-glm-5.2/v1a-mini-model/checkpoints/stage-b2-finewebedu16k-cont100m-2gpu/final", |
| "strict_equivalence_caveat": "RoPE 32->64 maps source frequencies into every-other target pair; unused target RoPE pairs are zero.", |
| "target_config": "/root/inference-v2/mini-glm-5.2/v1a-mini-model/config/glm-5.2-original-config.json" |
| }, |
| "v_head_dim": 256, |
| "vocab_size": 154880 |
| } |
|
|
