support qwen2 running on ascend npu device (#7022)

Co-authored-by: 刁莹煜 <diaoyingyu1@hisilicon.com>

python/sglang/srt/layers/quantization/compressed_tensors/compressed_tensors_moe.py

@@ -17,11 +17,12 @@ from sglang.srt.layers.quantization.utils import (
     per_tensor_dequantize,
     replace_parameter,
 )
-from sglang.srt.utils import is_cuda, set_weight_attrs
+from sglang.srt.utils import is_cuda, is_npu, set_weight_attrs
 
 _is_cuda = is_cuda()
+_is_npu = is_npu()
 
-if not _is_cuda:
+if not _is_cuda and not _is_npu:
     from vllm import _custom_ops as vllm_ops
     from vllm._custom_ops import scaled_fp8_quant