TheBloke commited on
Commit
c97a618
1 Parent(s): be85875

Update for Transformers AWQ support

Browse files
Files changed (1) hide show
  1. config.json +32 -25
config.json CHANGED
@@ -1,27 +1,34 @@
1
  {
2
- "_name_or_path": "./",
3
- "architectures": [
4
- "LlamaForCausalLM"
5
- ],
6
- "bos_token_id": 1,
7
- "eos_token_id": 2,
8
- "hidden_act": "silu",
9
- "hidden_size": 5120,
10
- "initializer_range": 0.02,
11
- "intermediate_size": 13824,
12
- "max_length": 512,
13
- "max_position_embeddings": 4096,
14
- "model_type": "llama",
15
- "num_attention_heads": 40,
16
- "num_hidden_layers": 40,
17
- "num_key_value_heads": 40,
18
- "pad_token_id": 0,
19
- "pretraining_tp": 1,
20
- "rms_norm_eps": 1e-05,
21
- "rope_scaling": null,
22
- "tie_word_embeddings": false,
23
- "torch_dtype": "float32",
24
- "transformers_version": "4.30.0.dev0",
25
- "use_cache": true,
26
- "vocab_size": 32000
 
 
 
 
 
 
 
27
  }
 
1
  {
2
+ "_name_or_path": "./",
3
+ "architectures": [
4
+ "LlamaForCausalLM"
5
+ ],
6
+ "bos_token_id": 1,
7
+ "eos_token_id": 2,
8
+ "hidden_act": "silu",
9
+ "hidden_size": 5120,
10
+ "initializer_range": 0.02,
11
+ "intermediate_size": 13824,
12
+ "max_length": 512,
13
+ "max_position_embeddings": 4096,
14
+ "model_type": "llama",
15
+ "num_attention_heads": 40,
16
+ "num_hidden_layers": 40,
17
+ "num_key_value_heads": 40,
18
+ "pad_token_id": 0,
19
+ "pretraining_tp": 1,
20
+ "rms_norm_eps": 1e-05,
21
+ "rope_scaling": null,
22
+ "tie_word_embeddings": false,
23
+ "torch_dtype": "float32",
24
+ "transformers_version": "4.30.0.dev0",
25
+ "use_cache": true,
26
+ "vocab_size": 32000,
27
+ "quantization_config": {
28
+ "quant_method": "awq",
29
+ "zero_point": true,
30
+ "group_size": 128,
31
+ "bits": 4,
32
+ "version": "gemm"
33
+ }
34
  }