Upload config.json with huggingface_hub
Browse files- config.json +47 -0
config.json
ADDED
@@ -0,0 +1,47 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"architectures": [
|
3 |
+
"LlamaForCausalLM"
|
4 |
+
],
|
5 |
+
"attention_bias": false,
|
6 |
+
"attention_dropout": 0.0,
|
7 |
+
"bos_token_id": 128000,
|
8 |
+
"eos_token_id": [
|
9 |
+
128001,
|
10 |
+
128008,
|
11 |
+
128009
|
12 |
+
],
|
13 |
+
"head_dim": 128,
|
14 |
+
"hidden_act": "silu",
|
15 |
+
"hidden_size": 8192,
|
16 |
+
"initializer_range": 0.02,
|
17 |
+
"intermediate_size": 28672,
|
18 |
+
"max_position_embeddings": 131072,
|
19 |
+
"mlp_bias": false,
|
20 |
+
"model_type": "llama",
|
21 |
+
"num_attention_heads": 64,
|
22 |
+
"num_hidden_layers": 80,
|
23 |
+
"num_key_value_heads": 8,
|
24 |
+
"pretraining_tp": 1,
|
25 |
+
"quantization": {
|
26 |
+
"group_size": 64,
|
27 |
+
"bits": 2
|
28 |
+
},
|
29 |
+
"quantization_config": {
|
30 |
+
"group_size": 64,
|
31 |
+
"bits": 2
|
32 |
+
},
|
33 |
+
"rms_norm_eps": 1e-05,
|
34 |
+
"rope_scaling": {
|
35 |
+
"factor": 8.0,
|
36 |
+
"high_freq_factor": 4.0,
|
37 |
+
"low_freq_factor": 1.0,
|
38 |
+
"original_max_position_embeddings": 8192,
|
39 |
+
"rope_type": "llama3"
|
40 |
+
},
|
41 |
+
"rope_theta": 500000.0,
|
42 |
+
"tie_word_embeddings": false,
|
43 |
+
"torch_dtype": "bfloat16",
|
44 |
+
"transformers_version": "4.40.0",
|
45 |
+
"use_cache": true,
|
46 |
+
"vocab_size": 128256
|
47 |
+
}
|