LlamaFinetuneBase commited on
Commit
5b38838
1 Parent(s): f1bdb7b

Upload config.json with huggingface_hub

Browse files
Files changed (1) hide show
  1. config.json +60 -0
config.json ADDED
@@ -0,0 +1,60 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "DeepseekV2ForCausalLM"
4
+ ],
5
+ "attention_bias": false,
6
+ "attention_dropout": 0.0,
7
+ "auto_map": {
8
+ "AutoConfig": "configuration_deepseek.DeepseekV2Config",
9
+ "AutoModel": "modeling_deepseek.DeepseekV2Model",
10
+ "AutoModelForCausalLM": "modeling_deepseek.DeepseekV2ForCausalLM"
11
+ },
12
+ "aux_loss_alpha": 0.001,
13
+ "bos_token_id": 100000,
14
+ "eos_token_id": 100001,
15
+ "ep_size": 1,
16
+ "first_k_dense_replace": 1,
17
+ "hidden_act": "silu",
18
+ "hidden_size": 5120,
19
+ "initializer_range": 0.02,
20
+ "intermediate_size": 12288,
21
+ "kv_lora_rank": 512,
22
+ "max_position_embeddings": 163840,
23
+ "model_type": "deepseek_v2",
24
+ "moe_intermediate_size": 1536,
25
+ "moe_layer_freq": 1,
26
+ "n_group": 8,
27
+ "n_routed_experts": 160,
28
+ "n_shared_experts": 2,
29
+ "norm_topk_prob": false,
30
+ "num_attention_heads": 128,
31
+ "num_experts_per_tok": 6,
32
+ "num_hidden_layers": 60,
33
+ "num_key_value_heads": 128,
34
+ "pretraining_tp": 1,
35
+ "q_lora_rank": 1536,
36
+ "qk_nope_head_dim": 128,
37
+ "qk_rope_head_dim": 64,
38
+ "rms_norm_eps": 1e-06,
39
+ "rope_scaling": {
40
+ "beta_fast": 32,
41
+ "beta_slow": 1,
42
+ "factor": 40,
43
+ "mscale": 1.0,
44
+ "mscale_all_dim": 1.0,
45
+ "original_max_position_embeddings": 4096,
46
+ "type": "yarn"
47
+ },
48
+ "rope_theta": 10000,
49
+ "routed_scaling_factor": 16.0,
50
+ "scoring_func": "softmax",
51
+ "seq_aux": true,
52
+ "tie_word_embeddings": false,
53
+ "topk_group": 3,
54
+ "topk_method": "group_limited_greedy",
55
+ "torch_dtype": "bfloat16",
56
+ "transformers_version": "4.39.3",
57
+ "use_cache": true,
58
+ "v_head_dim": 128,
59
+ "vocab_size": 102400
60
+ }