config.json

{
    "model_type": "llama",
    "architectures": ["LlamaForCausalLM"],
    "vocab_size": 32000,
    "embedding_size": 4096,
    "num_layers": 32,
    "num_heads": 32,
    "hidden_size": 5120,
    "intermediate_size": 20480,
    "max_position_embeddings": 2048,
    "pad_token_id": 0,
    "eos_token_id": 1,
    "bos_token_id": 2,
    "layer_norm_eps": 1e-5,
    "initializer_range": 0.02,
    "activation_function": "gelu",
    "num_attention_heads": 32,
    "num_hidden_layers": 32
}