Review key module learnings

Which recorded observation would most strongly suggest that backend optimization can be critical for transformer edge deployment on CPU?

XNNPACK delegation significantly reduced inference latency Portable ExecuTorch export produced a smaller .pte artifact PyTorch eager execution completed successfully Input tensor shapes remained fixed during export

Check answer

The lab showed that portable ExecuTorch execution did not outperform PyTorch eager execution. Which explanation is most consistent with the observed results?

Portable execution prioritizes broad operator portability over backend-specific kernel optimization ExecuTorch executes transformer models using Python rather than compiled operators Portable runtimes disable graph optimizations that are available in PyTorch Transformer inference performance is determined primarily by model file size rather than operator implementation

Check answer

Answer every question correctly to complete this page.

Review key module learnings

Nice work