Move triton import inside function to work on Windows

Signed-off-by: Keval Morabia <[email protected]>

Author: kevalmorabia97

modelopt/torch/quantization/qtensor/nvfp4_tensor.py

@@ -19,7 +19,6 @@
 
 from ..backends.utils import fp4_compatible
 from ..qtensor.base_qtensor import BaseQuantizedTensor
-from ..triton.fp4_kernel import fp4_dequantize
 from ..utils import reduce_amax, reduce_block_amax, reduce_block_padding
 
 # Define conversion tables
@@ -266,6 +265,8 @@ def _unpack_tensor(input: torch.Tensor):
                 ) from e
 
         if fast:
+            from ..triton.fp4_kernel import fp4_dequantize
+
             return fp4_dequantize(
                 self._quantized_data,
                 kwarg["scale"],