[adhanaku]

Summary

Add a self-contained Moonshine Streaming Medium ASR tutorial for OpenVINO.

• Export a dynamic no-cache OpenVINO FP16 IR and run growing-prefix inference.
• Document optional PyTorch XPU source-model validation and direct XPU/OpenVINO logit comparison.
• Include optional 100-utterance WER evaluation plus matched 500-utterance quality results.
• Benchmark static and dynamic fixed-forward latency, document roofline estimates, and demonstrate NNCF INT8 weight-only compression.
• Capture dynamic-shape fusion and ReduceSum optimization findings without requiring custom runtime changes.

Validation

• Validated notebook JSON/schema and Python syntax for all code cells.
• pytest .ci/test_notebooks.py::test_readme -q
• git diff --check
• Verified dynamic OpenVINO export/inference, INT8 compression, static/dynamic benchmark, and PyTorch XPU/OpenVINO one-forward logit comparison on Intel Arc B390.

Notes

PyTorch XPU sections are optional and skip cleanly on systems without a supported Intel GPU. The standard OpenVINO workflow runs with the default PyTorch install.

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB