Clean up imports (#5467)

python/sglang/srt/layers/quantization/compressed_tensors/schemes/compressed_tensors_w8a8_fp8.py

@@ -17,7 +17,6 @@ from sglang.srt.layers.quantization.compressed_tensors.schemes import (
 )
 from sglang.srt.layers.quantization.fp8_utils import (
     Fp8LinearOp,
-    maybe_create_device_identity,
     normalize_e4m3fn_to_e4m3fnuz,
 )
 from sglang.srt.layers.quantization.utils import is_fp8_fnuz, requantize_with_max_scale
@@ -99,8 +98,6 @@ class CompressedTensorsW8A8Fp8(CompressedTensorsScheme):
         weight_loader: Callable,
         **kwargs,
     ):
-        maybe_create_device_identity()
-
         output_size_per_partition = sum(output_partition_sizes)
         layer.logical_widths = output_partition_sizes