{
  "model_type": "mistral",
  "architectures": ["MistralForCausalLM"],
  "hidden_size": 4096,
  "intermediate_size": 11008,  // Size of the feedforward layer
  "num_attention_heads": 32,
  "num_hidden_layers": 32,
  "initializer_range": 0.02,
  "max_position_embeddings": 2048,
  "use_cache": true,
  "pad_token_id": 0,
  "bos_token_id": 1,
  "eos_token_id": 2,
  "tie_word_embeddings": false,
  "attention_probs_dropout_prob": 0.1,  // Dropout for attention layers
  "hidden_dropout_prob": 0.1,  // Dropout for fully connected layers
  "layer_norm_eps": 1e-5,  // Epsilon for layer normalization
  "transformers_version": "4.33.0",  // Specify the Transformers version
  "quantization_config": {
    "load_in_4bit": true,
    "bnb_4bit_compute_dtype": "bfloat16",
    "bnb_4bit_quant_type": "nf4",
    "bnb_4bit_use_double_quant": true
  }
}