Parakeet TDT 0.6B โ€” ONNX

Speech recognition for speech-android.

  • 0.6B params, FastConformer + TDT decoder
  • 25 European languages, 16 kHz input
  • SentencePiece vocabulary (1024 tokens)

Files

File Size Description
parakeet-encoder.onnx (+.data) 2.4 GB Encoder (FP32)
parakeet-encoder-int8.onnx 652 MB Encoder (INT8)
parakeet-decoder-joint.onnx 36 MB Decoder+Joint (FP32)
parakeet-decoder-joint-int8.onnx 9 MB Decoder+Joint (INT8)
vocab.json โ€” SentencePiece vocabulary

Encoder

I/O Name Shape Type
in audio_signal [B, 128, T] float32
in length [B] int64
out outputs [B, 1024, T'] float32
out encoded_lengths [B] int64

Decoder+Joint

I/O Name Shape Type
in encoder_outputs [B, 1024, T'] float32
in targets [B, U] int32
in target_length [B] int32
in input_states_1 [2, B, 640] float32
in input_states_2 [2, B, 640] float32
out outputs [B, U, T', 1030] float32
out output_states_1/2 [2, B, 640] float32

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Collection including aufklarer/Parakeet-TDT-0.6B-ONNX