FluidInference/parakeet-tdt-0.6b-v3-coreml

Add Encoder_v2.mlmodelc: int8-linear re-quantization fixing token corruption (FluidAudio#760) (#4)

Upload 5 files

Update README.md

Upload 5 files

Upload 3 files

Update README.md

Update README.md

Upload 5 files

Update README.md

Create parakeet_vocab.json

Upload 5 files

Delete Encoder.mlpackage

Delete Encoder6bit.mlmodelc

6bit

Upload 3 files

Upload 3 files

Delete Encoder.mlmodelc

Upload 5 files

Delete Encoder6bit.mlpackage

6bit encoder, potentially faster compile

Upload 5 files

separated model support as well

Upload 11 files

Update README.md

Not quantized - combined logits

15s, moved Transpose. Not quantized

int-8 model

Delete JointDecision.mlmodelc

Delete JointT.mlmodelc

Delete Encoder.mlmodelc

Delete Decoder.mlmodelc

Upload 5 files

JoinT model with Transpose

Decoder, not quantized

Single step, not quantized JD

Upload 5 files

Single Step Joint

New Models

Update README.md

Update README.md

Delete ParakeetEncoder_v2.mlmodelc

Delete Melspectrogram_v2.mlmodelc

Delete Melspectogram.mlmodelc

Upload 10 files

Delete ParakeetEncoder_6bit_15s.mlmodelc

Delete Melspectrogram_15s.mlmodelc

Upload 10 files

Upload 5 files

Upload 4 files

RNNTJoin with masking