SPES-9B / config.json
zjr2000's picture
Upload 14 files
b5d59cd verified
raw
history blame contribute delete
928 Bytes
{
"architectures": [
"Qwen3MoeForCausalLM"
],
"attention_bias": false,
"attention_dropout": 0.0,
"clip_qkv": null,
"decoder_sparse_step": 1,
"dtype": "bfloat16",
"eos_token_id": 151643,
"hidden_act": "silu",
"hidden_size": 2048,
"initializer_range": 0.02,
"intermediate_size": 6144,
"max_position_embeddings": 4096,
"mlp_only_layers": [],
"model_type": "qwen3_moe",
"moe_intermediate_size": 6144,
"norm_topk_prob": true,
"num_attention_heads": 16,
"num_experts": 8,
"num_experts_per_tok": 2,
"num_hidden_layers": 28,
"num_key_value_heads": 8,
"output_router_logits": false,
"pad_token_id": 151643,
"rms_norm_eps": 1e-06,
"rope_scaling": null,
"rope_theta": 1000000,
"router_aux_loss_coef": 0.001,
"sliding_window": null,
"tie_word_embeddings": false,
"transformers_version": "4.57.3",
"use_cache": true,
"use_sliding_window": false,
"vocab_size": 151936
}