{
  "_name_or_path": "/tmp/tmphei5qhy9",
  "architectures": [
    "GPTRefactForCausalLM"
  ],
  "attention_bias_in_fp32": true,
  "attention_softmax_in_fp32": true,
  "auto_map": {
    "AutoConfig": "configuration_gpt_refact.GPTRefactConfig",
    "AutoModelForCausalLM": "modeling_gpt_refact.GPTRefactForCausalLM"
  },
  "do_sample": true,
  "eos_token_id": 0,
  "initializer_range": 0.02,
  "layer_norm_epsilon": 1e-05,
  "model_type": "gpt_refact",
  "multi_query": true,
  "n_embd": 2048,
  "n_head": 32,
  "n_inner": null,
  "n_layer": 32,
  "n_positions": 4096,
  "quantization_config": {
    "bnb_4bit_compute_dtype": "bfloat16",
    "bnb_4bit_quant_type": "fp4",
    "bnb_4bit_use_double_quant": true,
    "llm_int8_enable_fp32_cpu_offload": false,
    "llm_int8_has_fp16_weight": false,
    "llm_int8_skip_modules": [
      "lm_head"
    ],
    "llm_int8_threshold": 6.0,
    "load_in_4bit": false,
    "load_in_8bit": true,
    "quant_method": "bitsandbytes"
  },
  "scale_attention_softmax_in_fp32": true,
  "torch_dtype": "float16",
  "transformers_version": "4.37.1",
  "use_cache": true,
  "vocab_size": 49216
}