Files
ModelHub XC 61b848acf9 初始化项目,由ModelHub XC社区提供模型
Model: SamsungSAILMontreal/Qwen3-30B-A3B-Instruct-2507-REAP
Source: Original Platform
2026-05-09 12:20:39 +08:00

53 lines
1.3 KiB
JSON

{
"architectures": [
"Qwen3MoeForCausalLM"
],
"attention_bias": false,
"attention_dropout": 0.0,
"bos_token_id": 151643,
"decoder_sparse_step": 1,
"dtype": "bfloat16",
"eos_token_id": 151645,
"head_dim": 128,
"hidden_act": "silu",
"hidden_size": 2048,
"initializer_range": 0.02,
"intermediate_size": 6144,
"max_position_embeddings": 262144,
"max_window_layers": 48,
"merge_args": {
"balance_group_size": 0,
"dataset": "c4+math+the-stack-smol",
"expert_saliency": "reap",
"gate_softmax": false,
"group": "freq_logits",
"merge": "none",
"merge_size": 96,
"merger_bs": 1024,
"merger_seq_len": 512,
"pca_dim": 64,
"precompute_input": true,
"use_gate_output": false
},
"mlp_only_layers": [],
"model_type": "qwen3_moe",
"moe_intermediate_size": 768,
"norm_topk_prob": true,
"num_attention_heads": 32,
"num_experts": 96,
"num_experts_per_tok": 8,
"num_hidden_layers": 48,
"num_key_value_heads": 4,
"output_router_logits": false,
"rms_norm_eps": 1e-06,
"rope_scaling": null,
"rope_theta": 10000000,
"router_aux_loss_coef": 0.001,
"sliding_window": null,
"tie_word_embeddings": false,
"transformers_version": "4.57.6",
"use_cache": true,
"use_sliding_window": false,
"vocab_size": 151936
}