quant_4bit_depthformer_base_decoder_step.json

{
  "quantization": "4bit_nf4",
  "base_model": "Content/MLModels/quant_fp16_depthformer_base_decoder_step.onnx",
  "block_size": 64,
  "double_quant": true,
  "note": "4-bit quantization requires runtime dequantization"
}