Modelpublic

OpenVoiceOS/parakeet-tdt-0.6b-polish-coreml-int8

sourceHugging Faceapache-2.0updated 5mo agoView on Hugging Face

0likes4downloads

Model Card

parakeet-tdt-0.6b-polish-coreml-int8

CoreML conversion of yuriyvnv/parakeet-tdt-0.6b-polish — INT8 PER CHANNEL SYMMETRIC quantized.


Architecture	TDT (Token-and-Duration Transducer)
Language	Polish
Sample rate	16000 Hz
Max audio	15.0s
Vocab size	8192
Framework	NVIDIA NeMo → CoreML (coremltools)

Components

File	Component	Best compute
`parakeet_mel_encoder.mlpackage`	mel_encoder	ANE / GPU
`parakeet_decoder.mlpackage`	decoder	CPU only
`parakeet_joint_decision_single_step.mlpackage`	jointdecisionsingle_step	ANE / GPU

Usage

bash

pip install ovos-stt-plugin-coreml

python

from ovos_stt_plugin_coreml import CoremlSTT
from ovos_plugin_manager.utils.audio import AudioFile

stt = CoremlSTT(config={"metadata": "metadata.json"})

with AudioFile("speech.wav") as f:
    audio = f.read()
print(stt.execute(audio))

Source model

yuriyvnv/parakeet-tdt-0.6b-polish