parakeet-ultra ONNX (browser / WebGPU)
ONNX export of moondream/parakeet-ultra (627M, Parakeet TDT, 25 European languages), prepared for onnxruntime-web on the WebGPU execution provider. Used by the Parakeet Redux WebGPU demo.
Based on the fp32 export in Olicorne/parakeet-tdt-0.6b-v3-ultra-onnx. Changes:
encoder-model.q4.onnx(393 MB): pointwise 1x1 Convs rewritten as MatMul, then 4-bitMatMulNBits(RTN, block 32, asymmetric). Runs on WebGPU.encoder-model.q8.onnx(650 MB): the same, 8-bit (block 64), closest to fp32.decoder_joint-model.int8.onnx: dynamic int8 decoder + joint.decoder_joint-model.onnxis fp32.preprocessor.onnx: NeMo 128-bin log-mel frontend with the STFT cast to float32 instead of float64 so it runs in onnxruntime-web (max abs diff 2e-5 vs the original).
Quality
Measured on 23 clips (JFK, LibriSpeech test-clean, MLS de/fr/es/it/nl/pt/pl), greedy TDT decoding, CPU onnxruntime:
| encoder | decoder | WER en | WER MLS | encoder cosine vs fp32 |
|---|---|---|---|---|
| fp32 | fp32 | 3.56 | 4.73 | 1.0 |
| fp32 | int8 | 3.91 | 4.50 | 1.0 |
| q8 | fp32 | 3.56 | 4.95 | 0.9996 |
| q4 (rtn, b32, asym) | fp32 | 3.20 | 4.50 | 0.977 |
The test set is small, so differences under ~0.5 WER are noise.
I/O
- preprocessor:
waveformsf32 [1,N] (16 kHz mono),waveforms_lensi64 [1] →features[1,128,T],features_lens - encoder:
audio_signal[1,128,T],lengthi64 →outputs[1,1024,T/8],encoded_lengths - decoder_joint:
encoder_outputs[1,1024,1],targetsi32 [1,1],target_lengthi32 [1],input_states_1/2[2,1,640] →outputs[1,1,1,8198] (8193 token logits, blank = 8192, then 5 duration logits for durations 0–4),output_states_1/2
License
CC-BY-4.0, same as moondream/parakeet-ultra, which builds on NVIDIA's parakeet-tdt-0.6b-v3.
- Downloads last month
- 50
Model tree for mrfakename/parakeet-ultra-ONNX
Base model
moondream/parakeet-ultra
// npm i @huggingface/transformers import { pipeline } from '@huggingface/transformers'; // Allocate pipeline const pipe = await pipeline('automatic-speech-recognition', 'mrfakename/parakeet-ultra-ONNX');