config.json

{
    "architectures": [
        "LlamaForConditionalGeneration"
    ],
    "hidden_size": 4096,
    "intermediate_size": 14336,
    "model_type": "llama",
    "num_attention_heads": 32,
    "num_hidden_layers": 40,
    "num_key_value_heads": 8,
    "quantization": {
        "group_size": 64,
        "bits": 8
    },
    "rms_norm_eps": 1e-05,
    "rope_scaling": {
        "factor": 8.0,
        "type": "linear",
        "high_freq_factor": 4.0,
        "low_freq_factor": 1.0,
        "original_max_position_embeddings": 8192,
        "rope_type": "llama3"
    },
    "rope_theta": 500000.0,
    "torch_dtype": "bfloat16",
    "use_bfloat16": false,
    "vocab_size": 128256
}