{
  "_attn_implementation_autoset": true,
  "_name_or_path": "SicariusSicariiStuff/Phi-lthy4",
  "architectures": [
    "LlamaForCausalLM"
  ],
  "attention_bias": false,
  "attention_dropout": 0.0,
  "bos_token_id": 100257,
  "eos_token_id": 100265,
  "head_dim": 128,
  "hidden_act": "silu",
  "hidden_size": 5120,
  "initializer_range": 0.02,
  "intermediate_size": 17920,
  "max_position_embeddings": 16384,
  "mlp_bias": false,
  "model_type": "llama",
  "num_attention_heads": 40,
  "num_hidden_layers": 32,
  "num_key_value_heads": 10,
  "original_max_position_embeddings": 16384,
  "pad_token_id": 100351,
  "pretraining_tp": 1,
  "quantization_config": {
    "bits": 4,
    "checkpoint_format": "gptq",
    "damp_percent": 0.01,
    "desc_act": true,
    "group_size": 32,
    "model_file_base_name": "gptq_model-4bit-32g",
    "model_name_or_path": "SicariusSicariiStuff/Phi-lthy4",
    "quant_method": "gptq",
    "static_groups": false,
    "sym": true,
    "true_sequential": true
  },
  "rms_norm_eps": 1e-05,
  "rope_scaling": null,
  "rope_theta": 250000,
  "tie_word_embeddings": false,
  "torch_dtype": "float16",
  "transformers_version": "4.47.1",
  "use_cache": false,
  "vocab_size": 100352
}