{
  "architectures": [
    "MistralForCausalLM"
  ],
  "max_position_embeddings": 2048,
  "hidden_size": 4096,
  "num_attention_heads": 32,
  "num_hidden_layers": 32,
  "vocab_size": 50257,
  "model_type": "mistral",   # <- Important key!
  "use_cache": true,
  "n_head": 32,
  "n_layer": 32,
  "n_positions": 2048,
  "initializer_range": 0.02,
  "layer_norm_epsilon": 1e-5,
  "attention_probs_dropout_prob": 0.1
}