gjdeboer commited on
Commit
131ea73
·
verified ·
1 Parent(s): 19836d5

Upload config.json with huggingface_hub

Browse files
Files changed (1) hide show
  1. config.json +48 -0
config.json ADDED
@@ -0,0 +1,48 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_fused_kernels_backend": "flash_attn",
3
+ "_use_fused_kernels": true,
4
+ "architectures": [
5
+ "LlamaForCausalLM"
6
+ ],
7
+ "attention_bias": false,
8
+ "attention_dropout": 0.0,
9
+ "bos_token_id": 128000,
10
+ "eos_token_id": 128001,
11
+ "head_dim": 128,
12
+ "hidden_act": "silu",
13
+ "hidden_size": 4096,
14
+ "initializer_range": 0.02,
15
+ "intermediate_size": 14336,
16
+ "max_position_embeddings": 131072,
17
+ "mlp_bias": false,
18
+ "model_type": "llama",
19
+ "num_attention_heads": 32,
20
+ "num_hidden_layers": 32,
21
+ "num_key_value_heads": 8,
22
+ "pad_token_id": 128275,
23
+ "pretraining_tp": 1,
24
+ "quantization": {
25
+ "group_size": 64,
26
+ "bits": 4,
27
+ "mode": "affine"
28
+ },
29
+ "quantization_config": {
30
+ "group_size": 64,
31
+ "bits": 4,
32
+ "mode": "affine"
33
+ },
34
+ "rms_norm_eps": 1e-05,
35
+ "rope_scaling": {
36
+ "factor": 8.0,
37
+ "high_freq_factor": 4.0,
38
+ "low_freq_factor": 1.0,
39
+ "original_max_position_embeddings": 8192,
40
+ "rope_type": "llama3"
41
+ },
42
+ "rope_theta": 500000.0,
43
+ "tie_word_embeddings": false,
44
+ "torch_dtype": "bfloat16",
45
+ "transformers_version": "4.51.3",
46
+ "use_cache": true,
47
+ "vocab_size": 128276
48
+ }