minor

Signed-off-by: realAsma <[email protected]>

Author: realAsma

modelopt/torch/quantization/tensor_quant.py

@@ -61,6 +61,7 @@ def scaled_e4m3_impl(
         None if inputs.device.index == torch.cuda.current_device() else inputs.device.index
     ):
         if amax is None:
+            # This adds overhead; however this is not a common use case.
             amax = torch.tensor(448.0, device=inputs.device, dtype=inputs.dtype)
         if amax.numel() == 1:
             outputs = cuda_ext_fp8.fake_e4m3fy(inputs, amax)