WakeHuBERT tiny
Copyright TigreGotico. Licensed under the Apache License, Version 2.0.

wakehubert.onnx, wakehubert_int8.onnx and config.json are copied unchanged from
the Hugging Face model repository https://huggingface.co/TigreGotico/wakehubert-tiny
at revision e30726f3a1c28bb5dffadf101fe26c97e0e3c3e1.

sha256
  7c08cdb068f16bee9fa4cfe6f2d56b1e78d938d5421296a635920edce1211e1b  wakehubert.onnx
  0b6a70470cf8f8c053b19b0ec14f1098e620eca213c36736018d71681462e064  wakehubert_int8.onnx

A 0.64M-parameter causal speech feature extractor distilled from HuBERT-base
(facebook/hubert-base-ls960): 16 kHz audio in, 128 features at 50 frames per
second out, receptive field 2.5 s. The model card on the Hub describes its
training data and results.
