37 lines
894 B
JSON
37 lines
894 B
JSON
|
|
{
|
||
|
|
"architectures": [
|
||
|
|
"Qwen3MoeForCausalLM"
|
||
|
|
],
|
||
|
|
"attention_bias": false,
|
||
|
|
"attention_dropout": 0.0,
|
||
|
|
"decoder_sparse_step": 1,
|
||
|
|
"dtype": "bfloat16",
|
||
|
|
"hidden_act": "silu",
|
||
|
|
"hidden_size": 1024,
|
||
|
|
"initializer_range": 0.02,
|
||
|
|
"intermediate_size": 2816,
|
||
|
|
"max_position_embeddings": 32768,
|
||
|
|
"mlp_only_layers": [],
|
||
|
|
"model_type": "qwen3_moe",
|
||
|
|
"moe_intermediate_size": 512,
|
||
|
|
"norm_topk_prob": false,
|
||
|
|
"num_attention_heads": 8,
|
||
|
|
"num_experts": 64,
|
||
|
|
"num_experts_per_tok": 8,
|
||
|
|
"num_hidden_layers": 24,
|
||
|
|
"num_key_value_heads": 4,
|
||
|
|
"output_router_logits": false,
|
||
|
|
"rms_norm_eps": 1e-06,
|
||
|
|
"rope_parameters": {
|
||
|
|
"rope_theta": 1000000.0,
|
||
|
|
"rope_type": "default"
|
||
|
|
},
|
||
|
|
"router_aux_loss_coef": 0.001,
|
||
|
|
"sliding_window": null,
|
||
|
|
"tie_word_embeddings": false,
|
||
|
|
"transformers_version": "5.0.0rc2",
|
||
|
|
"use_cache": true,
|
||
|
|
"use_sliding_window": false,
|
||
|
|
"vocab_size": 151936
|
||
|
|
}
|