ldwang commited on
Commit
2d2020f
1 Parent(s): 356733c

Upload config.json with huggingface_hub

Browse files
Files changed (1) hide show
  1. config.json +15 -9
config.json CHANGED
@@ -1,25 +1,31 @@
1
  {
2
  "architectures": [
3
- "LlamaForCausalLM"
4
  ],
5
- "bos_token_id": 100006,
6
- "eos_token_id": 100007,
 
 
 
 
 
 
7
  "hidden_act": "silu",
8
  "hidden_size": 6144,
9
  "initializer_range": 0.02,
10
  "intermediate_size": 24576,
11
- "max_position_embeddings": 4096,
12
- "model_type": "llama",
13
  "num_attention_heads": 48,
14
  "num_hidden_layers": 60,
15
  "num_key_value_heads": 8,
16
- "pad_token_id": 0,
17
  "pretraining_tp": 1,
18
  "rms_norm_eps": 1e-05,
19
  "rope_scaling": null,
 
20
  "tie_word_embeddings": false,
21
  "torch_dtype": "bfloat16",
22
- "transformers_version": "4.31.0",
23
  "use_cache": true,
24
- "vocab_size": 100008
25
- }
 
1
  {
2
  "architectures": [
3
+ "AquilaForCausalLM"
4
  ],
5
+ "attention_bias": false,
6
+ "attention_dropout": 0.0,
7
+ "auto_map": {
8
+ "AutoConfig": "configuration_aquila.AquilaConfig",
9
+ "AutoModelForCausalLM": "modeling_aquila.AquilaForCausalLM"
10
+ },
11
+ "bos_token_id": 143717,
12
+ "eos_token_id": 143718,
13
  "hidden_act": "silu",
14
  "hidden_size": 6144,
15
  "initializer_range": 0.02,
16
  "intermediate_size": 24576,
17
+ "max_position_embeddings": 8192,
18
+ "model_type": "aquila",
19
  "num_attention_heads": 48,
20
  "num_hidden_layers": 60,
21
  "num_key_value_heads": 8,
 
22
  "pretraining_tp": 1,
23
  "rms_norm_eps": 1e-05,
24
  "rope_scaling": null,
25
+ "rope_theta": 1000000.0,
26
  "tie_word_embeddings": false,
27
  "torch_dtype": "bfloat16",
28
+ "transformers_version": "4.39.3",
29
  "use_cache": true,
30
+ "vocab_size": 143973
31
+ }