config.json

{
  "model_type": "llama",
  "architectures": ["LlamaForCausalLM"],
  "hidden_size": 8192,
  "intermediate_size": 12288,
  "num_attention_heads": 64,
  "num_hidden_layers": 80,
  "rms_norm_eps": 1e-05,
  "max_position_embeddings": 2048,
  "rope_theta": 500000.0,
  "use_scaled_rope": true,
  "tie_word_embeddings": false,
  "transformers_version": "4.31.0",
  "vocab_size": 128256,
  "bos_token_id": 1,
  "eos_token_id": 2,
  "pad_token_id": 0,
  "initializer_range": 0.02,
  "n_kv_heads": 8,
  "ffn_dim_multiplier": 1.3,
  "multiple_of": 4096,
  "vision_chunk_size": 560,
  "vision_max_num_chunks": 4,
  "vision_num_cross_attention_layers": 20
}