MLX
Safetensors
llama
8-bit precision
File size: 839 Bytes
fa2a437
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
{
    "architectures": [
        "LlamaForCausalLM"
    ],
    "attention_bias": false,
    "attention_dropout": 0.0,
    "bos_token_id": 1,
    "eos_token_id": 2,
    "hidden_act": "silu",
    "hidden_size": 768,
    "initializer_range": 0.02,
    "intermediate_size": 2048,
    "max_position_embeddings": 2048,
    "model_type": "llama",
    "num_attention_heads": 12,
    "num_hidden_layers": 12,
    "num_key_value_heads": 12,
    "pretraining_tp": 1,
    "quantization": {
        "group_size": 64,
        "bits": 8
    },
    "quantization_config": {
        "group_size": 64,
        "bits": 8
    },
    "rms_norm_eps": 1e-05,
    "rope_scaling": null,
    "rope_theta": 10000.0,
    "tie_word_embeddings": false,
    "torch_dtype": "float32",
    "transformers_version": "4.36.2",
    "use_cache": true,
    "vocab_size": 32000
}