Stage Orukeet INT8 Core ML candidate for OpenWhispr draft integration
#2
by NathanRoll - opened
Existing portable inference archive, uploaded byte-for-byte. INT8 encoder weights, FP16 remaining components, greedy v3 decoding, 16 kHz mono. Archive includes its 400-clip evaluation, attribution and historical experimental provenance. Draft app integration and runtime validation are tracked in https://github.com/Oruk-AI/orukeet/pull/10. No training checkpoint or compiled device cache is included.
Does this model support Hindi or Bengali language?
Not in this release, unfortunately-- Hindi and Bengali aren't among the 25 supported languages.
NathanRoll changed pull request status to merged