{
  "architectures": [
    "LlamaForCausalLM"
  ],
  "model_type": "llama",
  "vocab_size": 151665,
  "hidden_size": 2048,
  "intermediate_size": 5504,
  "num_hidden_layers": 32,
  "num_attention_heads": 16,
  "num_key_value_heads": 4,
  "max_position_embeddings": 512,
  "rope_theta": 500000.0,
  "rms_norm_eps": 1e-06,
  "hidden_act": "silu",
  "tie_word_embeddings": true,
  "torch_dtype": "bfloat16",
  "transformers_version": "4.40.0",
  "_aksarallm_metadata": {
    "model_name": "AksaraLLM-1.55B-Native",
    "total_params": 2039027712,
    "training_data": "293325 sequences",
    "architecture_style": "LLaMA-3 (GQA, RoPE, SwiGLU, RMSNorm)"
  }
}