{
  "_name_or_path": "/mnt/iem-nas/home/qinzhen/qinzhen/experiment/nanogpt/ckpts_hf/train_xmixers_llama_medium_50b/",
  "ape_type": "sincos",
  "architectures": [
    "LLaMAForCausalLM"
  ],
  "base": 10000,
  "bias": false,
  "bos_token_id": 1,
  "causal": true,
  "embed_dim": 1024,
  "eos_token_id": 2,
  "ffn_activation": "gelu",
  "glu_activation": "silu",
  "init_std": 0.02,
  "kv_heads": -1,
  "linear_activation": "silu",
  "lrpe_type": 1,
  "max_position_embeddings": 1024,
  "mid_dim": 2816,
  "model_type": "llama_",
  "norm_type": "layernorm",
  "num_heads": 16,
  "num_layers": 24,
  "pad_token_id": null,
  "tie_word_embeddings": true,
  "torch_dtype": "bfloat16",
  "transformers_version": "4.44.2",
  "use_ape": true,
  "use_cache": true,
  "use_lrpe": true,
  "use_output_gate": true,
  "vocab_size": 50304
}