{
  "_name_or_path": "/nfs/scistore14/alistgrp/apanfero/models/Llama-2-13b-AQLM-2Bit-1x16-hf/",
  "aqlm": {
    "in_group_size": 8,
    "nbits_per_codebook": 16,
    "num_codebooks": 1,
    "out_group_size": 1
  },
  "architectures": [
    "LlamaForCausalLM"
  ],
  "attention_bias": false,
  "attention_dropout": 0.0,
  "auto_map": {
    "AutoConfig": "configuration_llama_aqlm.LlamaConfig",
    "AutoModelForCausalLM": "modeling_llama_aqlm.LlamaForCausalLM"
  },
  "bos_token_id": 1,
  "eos_token_id": 2,
  "hidden_act": "silu",
  "hidden_size": 5120,
  "initializer_range": 0.02,
  "intermediate_size": 13824,
  "max_position_embeddings": 4096,
  "model_type": "llama_aqlm",
  "num_attention_heads": 40,
  "num_hidden_layers": 40,
  "num_key_value_heads": 40,
  "pretraining_tp": 1,
  "rms_norm_eps": 1e-05,
  "rope_scaling": null,
  "rope_theta": 10000.0,
  "tie_word_embeddings": false,
  "torch_dtype": "float16",
  "transformers_version": "4.38.0",
  "use_cache": true,
  "vocab_size": 32000
}