if self.quantization is not None:
            self.quantization = self.quantization.lower()

        # Parse quantization method from the HF model config, if available.
        quant_cfg = self._parse_quant_hf_config()

        if quant_cfg is not None:
            quant_method = quant_cfg.get("quant_method", "").lower()

            # Detect which checkpoint is it
            for name in QUANTIZATION_METHODS:
                method = get_quantization_config(name)
                quantization_override = method.override_quantization_method(
                    quant_cfg, self.quantization)
                if quantization_override:
                    quant_method = quantization_override
                    self.quantization = quantization_override
                    break

  File "/usr/local/lib/python3.12/dist-packages/vllm/engine/arg_utils.py", line 1034, in create_engine_config
    raise ValueError(
ValueError: BitsAndBytes load format and QLoRA adapter only support 'bitsandbytes' quantization, but got None

        # bitsandbytes quantization needs a specific model loader
        # so we make sure the quant method and the load format are consistent
        if (self.quantization == "bitsandbytes" or
           self.qlora_adapter_name_or_path is not None) and \
           self.load_format != "bitsandbytes":
            raise ValueError(
                "BitsAndBytes quantization and QLoRA adapter only support "
                f"'bitsandbytes' load format, but got {self.load_format}")

        if (self.load_format == "bitsandbytes" or
            self.qlora_adapter_name_or_path is not None) and \
            self.quantization != "bitsandbytes":
            raise ValueError(
                "BitsAndBytes load format and QLoRA adapter only support "
                f"'bitsandbytes' quantization, but got {self.quantization}")

        assert self.cpu_offload_gb >= 0, (
            "CPU offload space must be non-negative"
            f", but got {self.cpu_offload_gb}")

        device_config = DeviceConfig(device=self.device)
        model_config = self.create_model_config()     # <- do auto detection there