53 lines
1.2 KiB
JSON
53 lines
1.2 KiB
JSON
|
|
{
|
||
|
|
"vocab_size": 128256,
|
||
|
|
"max_position_embeddings": 262144,
|
||
|
|
"hidden_size": 1536,
|
||
|
|
"intermediate_size": 8960,
|
||
|
|
"moe_intermediate_size": 1280,
|
||
|
|
"num_hidden_layers": 26,
|
||
|
|
"num_nextn_predict_layers": 1,
|
||
|
|
"num_attention_heads": 32,
|
||
|
|
"n_shared_experts": 1,
|
||
|
|
"n_routed_experts": 64,
|
||
|
|
"ep_size": 1,
|
||
|
|
"routed_scaling_factor": 1.0,
|
||
|
|
"kv_lora_rank": 512,
|
||
|
|
"q_lora_rank": null,
|
||
|
|
"qk_rope_head_dim": 64,
|
||
|
|
"v_head_dim": 192,
|
||
|
|
"qk_nope_head_dim": 128,
|
||
|
|
"topk_method": "noaux_tc",
|
||
|
|
"n_group": 1,
|
||
|
|
"topk_group": 1,
|
||
|
|
"num_experts_per_tok": 4,
|
||
|
|
"moe_layer_freq": 1,
|
||
|
|
"first_k_dense_replace": 1,
|
||
|
|
"norm_topk_prob": true,
|
||
|
|
"scoring_func": "sigmoid",
|
||
|
|
"num_key_value_heads": 32,
|
||
|
|
"hidden_act": "silu",
|
||
|
|
"initializer_range": 0.006,
|
||
|
|
"rms_norm_eps": 1e-06,
|
||
|
|
"use_cache": true,
|
||
|
|
"rope_theta": 100000,
|
||
|
|
"rope_scaling": {
|
||
|
|
"beta_fast": 32,
|
||
|
|
"beta_slow": 1,
|
||
|
|
"factor": 64.0,
|
||
|
|
"mscale": 1.0,
|
||
|
|
"mscale_all_dim": 1.0,
|
||
|
|
"original_max_position_embeddings": 4096,
|
||
|
|
"type": "yarn"
|
||
|
|
},
|
||
|
|
"attention_bias": false,
|
||
|
|
"attention_dropout": 0.0,
|
||
|
|
"tie_word_embeddings": false,
|
||
|
|
"architectures": [
|
||
|
|
"DeepseekV3ForCausalLM"
|
||
|
|
],
|
||
|
|
"bos_token_id": 1,
|
||
|
|
"eos_token_id": 2,
|
||
|
|
"transformers_version": "4.57.1",
|
||
|
|
"model_type": "deepseek_v3",
|
||
|
|
"dtype": "bfloat16"
|
||
|
|
}
|