model_gpu_mlx0.32/mlx_config.json

{
  "quantize": "none",
  "format": "mlx-safetensors",
  "prep_version": 2,
  "head": "mean+proj",
  "num_attention_heads": 12
}