quantization_info.json

{
  "quantization_type": "16bit",
  "original_dtype": "float32",
  "pruning_applied": true,
  "pruning_amount": 0.2
}