| { |
| "architectures": [ |
| "BailingMoeV2ForCausalLM" |
| ], |
| "attention_dropout": 0.0, |
| "auto_map": { |
| "AutoConfig": "configuration_bailing_moe_v2.BailingMoeV2Config", |
| "AutoModel": "modeling_bailing_moe_v2.BailingMoeV2Model", |
| "AutoModelForCausalLM": "modeling_bailing_moe_v2.BailingMoeV2ForCausalLM" |
| }, |
| "eos_token_id": 156892, |
| "pad_token_id": 156892, |
| "num_hidden_layers": 32, |
| "hidden_size": 4096, |
| "intermediate_size": 9216, |
| "first_k_dense_replace": 1, |
| "hidden_act": "silu", |
| "initializer_range": 0.006, |
| "max_position_embeddings": 32768, |
| "model_type": "bailing_moe_v2", |
| "moe_intermediate_size": 1024, |
| "norm_topk_prob": true, |
| "num_experts_per_tok": 8, |
| "norm_head": false, |
| "num_attention_heads": 32, |
| "num_experts": 256, |
| "num_key_value_heads": 4, |
| "rope_theta": 600000, |
| "pretraining_tp": 1, |
| "rope_scaling": null, |
| "tie_word_embeddings": false, |
| "torch_dtype": "bfloat16", |
| "transformers_version": "4.45.0", |
| "use_bias": false, |
| "use_rmsnorm": true, |
| "rms_norm_eps": 1e-06, |
| "head_dim": 128, |
| "num_shared_experts": 1, |
| "use_cache": true, |
| "use_qkv_bias": false, |
| "embedding_dropout": 0.0, |
| "norm_softmax": false, |
| "output_dropout": 0.0, |
| "partial_rotary_factor": 0.5, |
| "vocab_size": 157184, |
| "rotary_dim": 64, |
| "output_router_logits": false, |
| "using_split_qkv_in_self_attention": false, |
| "router_dtype": "fp32", |
| "moe_router_enable_expert_bias": true, |
| "routed_scaling_factor": 2.5, |
| "n_group": 8, |
| "topk_group": 4, |
| "use_qk_norm": true, |
| "score_function": "sigmoid", |
| "quantization_config": { |
| "activation_scheme": "dynamic", |
| "fmt": "e4m3", |
| "quant_method": "fp8", |
| "weight_block_size": [ |
| 128, |
| 128 |
| ], |
| "modules_to_not_convert": [ |
| "model.layers.2.mlp.gate", |
| "model.layers.3.mlp.gate.expert_bias", |
| "model.layers.20.attention.key_layernorm", |
| "model.layers.18.attention.query_layernorm", |
| "model.layers.7.post_attention_layernorm", |
| "model.layers.6.attention.query_layernorm", |
| "model.layers.16.attention.query_layernorm", |
| "model.layers.28.input_layernorm", |
| "model.layers.15.mlp.gate.expert_bias", |
| "model.layers.12.attention.query_layernorm", |
| "model.layers.13.attention.query_layernorm", |
| "model.layers.19.mlp.gate.expert_bias", |
| "model.layers.9.post_attention_layernorm", |
| "model.layers.27.mlp.gate.expert_bias", |
| "model.layers.11.attention.key_layernorm", |
| "model.layers.6.attention.key_layernorm", |
| "model.layers.12.input_layernorm", |
| "model.layers.1.mlp.gate.expert_bias", |
| "model.layers.18.input_layernorm", |
| "model.layers.11.input_layernorm", |
| "model.layers.31.mlp.gate", |
| "model.layers.7.attention.query_layernorm", |
| "model.layers.22.post_attention_layernorm", |
| "model.layers.13.input_layernorm", |
| "model.layers.30.attention.query_layernorm", |
| "model.layers.20.post_attention_layernorm", |
| "model.layers.11.mlp.gate", |
| "model.layers.24.attention.query_layernorm", |
| "model.layers.27.mlp.gate", |
| "model.layers.26.mlp.gate.expert_bias", |
| "model.layers.7.attention.key_layernorm", |
| "model.layers.23.post_attention_layernorm", |
| "model.layers.30.mlp.gate", |
| "model.layers.28.attention.key_layernorm", |
| "model.layers.11.mlp.gate.expert_bias", |
| "model.layers.3.input_layernorm", |
| "model.layers.4.attention.query_layernorm", |
| "model.layers.7.mlp.gate.expert_bias", |
| "model.word_embeddings", |
| "model.layers.22.mlp.gate", |
| "model.layers.2.input_layernorm", |
| "model.layers.4.mlp.gate", |
| "model.layers.26.post_attention_layernorm", |
| "model.layers.27.attention.key_layernorm", |
| "model.layers.29.mlp.gate.expert_bias", |
| "model.layers.20.mlp.gate", |
| "model.layers.10.mlp.gate.expert_bias", |
| "model.layers.16.input_layernorm", |
| "model.layers.1.post_attention_layernorm", |
| "model.layers.15.mlp.gate", |
| "model.layers.28.mlp.gate.expert_bias", |
| "model.layers.25.post_attention_layernorm", |
| "model.layers.4.attention.key_layernorm", |
| "model.layers.22.attention.query_layernorm", |
| "model.layers.21.mlp.gate.expert_bias", |
| "model.layers.31.input_layernorm", |
| "model.layers.4.input_layernorm", |
| "model.layers.21.mlp.gate", |
| "model.layers.16.mlp.gate.expert_bias", |
| "model.layers.28.post_attention_layernorm", |
| "model.layers.27.input_layernorm", |
| "model.layers.23.mlp.gate.expert_bias", |
| "model.layers.31.attention.key_layernorm", |
| "model.layers.21.post_attention_layernorm", |
| "model.layers.1.attention.query_layernorm", |
| "model.layers.5.post_attention_layernorm", |
| "model.layers.8.post_attention_layernorm", |
| "model.layers.26.attention.key_layernorm", |
| "model.layers.0.attention.key_layernorm", |
| "model.layers.20.input_layernorm", |
| "model.layers.14.input_layernorm", |
| "model.layers.26.attention.query_layernorm", |
| "model.layers.5.attention.key_layernorm", |
| "model.layers.18.mlp.gate.expert_bias", |
| "model.layers.17.mlp.gate", |
| "model.layers.12.post_attention_layernorm", |
| "model.layers.10.post_attention_layernorm", |
| "model.layers.14.mlp.gate.expert_bias", |
| "model.layers.15.attention.key_layernorm", |
| "model.layers.8.attention.query_layernorm", |
| "model.layers.9.mlp.gate", |
| "model.norm", |
| "model.layers.26.input_layernorm", |
| "model.layers.19.input_layernorm", |
| "model.layers.21.input_layernorm", |
| "model.layers.9.input_layernorm", |
| "model.layers.3.attention.query_layernorm", |
| "model.layers.19.post_attention_layernorm", |
| "model.layers.31.mlp.gate.expert_bias", |
| "model.layers.11.attention.query_layernorm", |
| "model.layers.24.attention.key_layernorm", |
| "model.layers.9.attention.query_layernorm", |
| "model.layers.0.input_layernorm", |
| "model.layers.1.attention.key_layernorm", |
| "model.layers.25.attention.key_layernorm", |
| "model.layers.23.mlp.gate", |
| "model.layers.23.attention.query_layernorm", |
| "model.layers.10.mlp.gate", |
| "model.layers.31.attention.query_layernorm", |
| "model.layers.15.attention.query_layernorm", |
| "model.layers.5.mlp.gate", |
| "model.layers.27.post_attention_layernorm", |
| "model.layers.18.mlp.gate", |
| "model.layers.22.attention.key_layernorm", |
| "model.layers.17.post_attention_layernorm", |
| "model.layers.6.mlp.gate.expert_bias", |
| "model.layers.28.mlp.gate", |
| "model.layers.12.mlp.gate.expert_bias", |
| "model.layers.13.attention.key_layernorm", |
| "model.layers.17.attention.key_layernorm", |
| "model.layers.1.input_layernorm", |
| "model.layers.4.mlp.gate.expert_bias", |
| "model.layers.25.mlp.gate", |
| "model.layers.30.attention.key_layernorm", |
| "model.layers.3.post_attention_layernorm", |
| "model.layers.6.input_layernorm", |
| "model.layers.13.post_attention_layernorm", |
| "model.layers.19.attention.key_layernorm", |
| "model.layers.21.attention.query_layernorm", |
| "model.layers.7.input_layernorm", |
| "model.layers.24.input_layernorm", |
| "model.layers.8.input_layernorm", |
| "model.layers.10.input_layernorm", |
| "model.layers.16.attention.key_layernorm", |
| "model.layers.16.post_attention_layernorm", |
| "model.layers.2.post_attention_layernorm", |
| "model.layers.24.post_attention_layernorm", |
| "model.layers.15.input_layernorm", |
| "model.layers.0.post_attention_layernorm", |
| "model.layers.18.attention.key_layernorm", |
| "model.layers.29.attention.key_layernorm", |
| "model.layers.23.input_layernorm", |
| "model.layers.0.attention.query_layernorm", |
| "model.layers.3.attention.key_layernorm", |
| "model.layers.30.mlp.gate.expert_bias", |
| "model.layers.14.attention.query_layernorm", |
| "model.layers.2.attention.query_layernorm", |
| "model.layers.5.input_layernorm", |
| "model.layers.14.mlp.gate", |
| "model.layers.29.post_attention_layernorm", |
| "model.layers.30.input_layernorm", |
| "model.layers.16.mlp.gate", |
| "model.layers.25.input_layernorm", |
| "model.layers.29.input_layernorm", |
| "model.layers.15.post_attention_layernorm", |
| "model.layers.5.attention.query_layernorm", |
| "model.layers.24.mlp.gate.expert_bias", |
| "model.layers.19.attention.query_layernorm", |
| "model.layers.20.attention.query_layernorm", |
| "model.layers.29.mlp.gate", |
| "model.layers.25.attention.query_layernorm", |
| "model.layers.22.input_layernorm", |
| "model.layers.26.mlp.gate", |
| "model.layers.27.attention.query_layernorm", |
| "model.layers.13.mlp.gate.expert_bias", |
| "model.layers.2.mlp.gate.expert_bias", |
| "model.layers.18.post_attention_layernorm", |
| "model.layers.12.attention.key_layernorm", |
| "model.layers.28.attention.query_layernorm", |
| "model.layers.14.post_attention_layernorm", |
| "model.layers.19.mlp.gate", |
| "model.layers.2.attention.key_layernorm", |
| "model.layers.23.attention.key_layernorm", |
| "model.layers.30.post_attention_layernorm", |
| "model.layers.20.mlp.gate.expert_bias", |
| "model.layers.4.post_attention_layernorm", |
| "model.layers.21.attention.key_layernorm", |
| "lm_head", |
| "model.layers.31.post_attention_layernorm", |
| "model.layers.17.attention.query_layernorm", |
| "model.layers.24.mlp.gate", |
| "model.layers.8.mlp.gate", |
| "model.layers.11.post_attention_layernorm", |
| "model.layers.12.mlp.gate", |
| "model.layers.17.mlp.gate.expert_bias", |
| "model.layers.9.attention.key_layernorm", |
| "model.layers.6.post_attention_layernorm", |
| "model.layers.14.attention.key_layernorm", |
| "model.layers.10.attention.query_layernorm", |
| "model.layers.5.mlp.gate.expert_bias", |
| "model.layers.13.mlp.gate", |
| "model.layers.8.mlp.gate.expert_bias", |
| "model.layers.25.mlp.gate.expert_bias", |
| "model.layers.10.attention.key_layernorm", |
| "model.layers.6.mlp.gate", |
| "model.layers.29.attention.query_layernorm", |
| "model.layers.8.attention.key_layernorm", |
| "model.layers.3.mlp.gate", |
| "model.layers.22.mlp.gate.expert_bias", |
| "model.layers.9.mlp.gate.expert_bias", |
| "model.layers.7.mlp.gate", |
| "model.layers.1.mlp.gate", |
| "model.layers.17.input_layernorm" |
| ] |
| } |
| } |
