# Coreml Validate

> Validate Kokoro Core ML exports against their frozen PyTorch/reference behavior and audio-quality gates. Use for numerical parity, input/output shape drift, export regressions, CPU-fallback signals, SDK bundles, or release-readiness evidence.

- Skill: `mattmireles/coreml-validate` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add mattmireles/coreml-validate`
- Raw SKILL.md: https://api.skillmd.com/api/skills/mattmireles/coreml-validate/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: mattmireles (https://skillmd.com/u/mattmireles)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/mattmireles/coreml-validate

---


# Validate Kokoro Core ML

1. Freeze the source revision, export configuration, package hashes, fixtures,
   dtypes, shapes, and threshold before observing results.
2. Run `kokoro/coreml_numeric_validate.py` and
   `kokoro/coreml_export_verify.py` or the narrow equivalent for the changed
   export.
3. Test the actual package on the target device when the claim concerns
   scheduling or deployment; separate numerical parity from audio/listening
   quality and performance.
4. Run `scripts/validate_sdk_bundle.mjs` for hosted SDK bundle changes.
5. Fail closed on shape, dtype, output, fallback, package-identity, or fixture
mismatches. Never loosen a preregistered threshold after seeing a result.