Hugging Face's logo

tencent
/
Tencent-Hunyuan-Large

Text Generation
Transformers
Safetensors
English
Inference Endpoints
Model card Files Community
8
woodchen7 commited on
Commit
c9e801e
•
1 Parent(s): 21f2deb

Upload config.json with huggingface_hub

Browse files
Files changed (1) hide show
  1. config.json +119 -0
config.json ADDED
@@ -0,0 +1,119 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "/apdcephfs_gy2/share_302508627/woodchenwu/hunyuan-L-instruct-128K-1102online-HF/",
3
+ "architectures": [
4
+ "HunYuanForCausalLM"
5
+ ],
6
+ "attention_bias": false,
7
+ "attention_dropout": 0.0,
8
+ "auto_map": {
9
+ "AutoConfig": "configuration_hunyuan.HunYuanConfig",
10
+ "AutoModel": "modeling_hunyuan.HunyuanModel",
11
+ "AutoModelForCausalLM": "modeling_hunyuan.HunYuanForCausalLM"
12
+ },
13
+ "bos_token_id": 1,
14
+ "capacity_factor": 1.0,
15
+ "cla_share_factor": 2,
16
+ "eos_token_id": 2,
17
+ "hidden_act": "silu",
18
+ "hidden_size": 6400,
19
+ "initializer_range": 0.02,
20
+ "intermediate_size": 18304,
21
+ "max_position_embeddings": 131072,
22
+ "model_type": "hunyuan",
23
+ "moe_drop_tokens": false,
24
+ "moe_random_routing_dropped_token": false,
25
+ "moe_topk": 1,
26
+ "num_attention_heads": 80,
27
+ "num_experts": 16,
28
+ "num_hidden_layers": 64,
29
+ "num_key_value_heads": 8,
30
+ "num_shared_expert": 1,
31
+ "pad_token_id": 0,
32
+ "pretraining_tp": 1,
33
+ "quantization_config": {
34
+ "activation_scheme": "static",
35
+ "ignored_layers": [
36
+ "model.layers.38.mlp.gate.wg",
37
+ "model.layers.60.mlp.gate.wg",
38
+ "model.layers.20.mlp.gate.wg",
39
+ "model.layers.11.mlp.gate.wg",
40
+ "model.layers.37.mlp.gate.wg",
41
+ "model.layers.51.mlp.gate.wg",
42
+ "model.layers.54.mlp.gate.wg",
43
+ "model.layers.49.mlp.gate.wg",
44
+ "model.layers.59.mlp.gate.wg",
45
+ "lm_head",
46
+ "model.layers.61.mlp.gate.wg",
47
+ "model.layers.50.mlp.gate.wg",
48
+ "model.layers.26.mlp.gate.wg",
49
+ "model.layers.34.mlp.gate.wg",
50
+ "model.layers.29.mlp.gate.wg",
51
+ "model.layers.2.mlp.gate.wg",
52
+ "model.layers.58.mlp.gate.wg",
53
+ "model.layers.23.mlp.gate.wg",
54
+ "model.layers.16.mlp.gate.wg",
55
+ "model.layers.4.mlp.gate.wg",
56
+ "model.layers.31.mlp.gate.wg",
57
+ "model.layers.45.mlp.gate.wg",
58
+ "model.layers.56.mlp.gate.wg",
59
+ "model.layers.3.mlp.gate.wg",
60
+ "model.layers.19.mlp.gate.wg",
61
+ "model.layers.14.mlp.gate.wg",
62
+ "model.layers.8.mlp.gate.wg",
63
+ "model.layers.41.mlp.gate.wg",
64
+ "model.layers.53.mlp.gate.wg",
65
+ "model.layers.46.mlp.gate.wg",
66
+ "model.layers.7.mlp.gate.wg",
67
+ "model.layers.25.mlp.gate.wg",
68
+ "model.layers.6.mlp.gate.wg",
69
+ "model.layers.1.mlp.gate.wg",
70
+ "model.layers.10.mlp.gate.wg",
71
+ "model.layers.12.mlp.gate.wg",
72
+ "model.layers.15.mlp.gate.wg",
73
+ "model.layers.18.mlp.gate.wg",
74
+ "model.layers.57.mlp.gate.wg",
75
+ "model.layers.63.mlp.gate.wg",
76
+ "model.layers.55.mlp.gate.wg",
77
+ "model.layers.40.mlp.gate.wg",
78
+ "model.layers.32.mlp.gate.wg",
79
+ "model.layers.39.mlp.gate.wg",
80
+ "model.layers.62.mlp.gate.wg",
81
+ "model.layers.28.mlp.gate.wg",
82
+ "model.layers.30.mlp.gate.wg",
83
+ "model.layers.27.mlp.gate.wg",
84
+ "model.layers.22.mlp.gate.wg",
85
+ "model.layers.5.mlp.gate.wg",
86
+ "model.layers.21.mlp.gate.wg",
87
+ "model.layers.36.mlp.gate.wg",
88
+ "model.layers.48.mlp.gate.wg",
89
+ "model.layers.42.mlp.gate.wg",
90
+ "model.layers.44.mlp.gate.wg",
91
+ "model.layers.52.mlp.gate.wg",
92
+ "model.layers.0.mlp.gate.wg",
93
+ "model.layers.13.mlp.gate.wg",
94
+ "model.layers.17.mlp.gate.wg",
95
+ "model.layers.24.mlp.gate.wg",
96
+ "model.layers.33.mlp.gate.wg",
97
+ "model.layers.35.mlp.gate.wg",
98
+ "model.layers.43.mlp.gate.wg",
99
+ "model.layers.47.mlp.gate.wg",
100
+ "model.layers.9.mlp.gate.wg"
101
+ ],
102
+ "quant_method": "fp8"
103
+ },
104
+ "rms_norm_eps": 1e-05,
105
+ "rope_scaling": {
106
+ "alpha": 1000.0,
107
+ "factor": 1.0,
108
+ "type": "dynamic"
109
+ },
110
+ "rope_theta": 10000.0,
111
+ "tie_word_embeddings": true,
112
+ "torch_dtype": "bfloat16",
113
+ "transformers_version": "4.44.2",
114
+ "use_cache": true,
115
+ "use_cla": true,
116
+ "use_mixed_mlp_moe": true,
117
+ "use_qk_norm": true,
118
+ "vocab_size": 129024
119
+ }