Update README.md

README.md

@@ -50,8 +50,6 @@ with torch.amp.autocast(
 print(probs)
 ```
 
-Published GPU benchmarks use FP16 mixed precision for model execution. SDM respects the caller's active PyTorch autocast context but does not enable autocasting automatically, so use the context shown above for comparable GPU throughput.
-
 To learn more, visit [structured-data-models](https://github.com/NVIDIA/structured-data-models).
 
 ## License