From 739d5655f0d8e28398da0a2e3ea97eb4d19dc3a5 Mon Sep 17 00:00:00 2001 From: Sam Anklesaria Date: Thu, 4 Sep 2025 14:55:55 +0000 Subject: [PATCH] Note references to torchaudio.info --- examples/dnn_beamformer/README.md | 6 +++++- examples/hubert/README.md | 4 ++++ 2 files changed, 9 insertions(+), 1 deletion(-) diff --git a/examples/dnn_beamformer/README.md b/examples/dnn_beamformer/README.md index ced962290b..bb79c8fb94 100644 --- a/examples/dnn_beamformer/README.md +++ b/examples/dnn_beamformer/README.md @@ -2,7 +2,11 @@ This directory contains sample implementations of training and evaluation pipelines for an DNN Beamforming model. -The `DNNBeamformer` model composes the following componenst: +> [!NOTE] +> The code uses the `torchaudio.info` function which was removed from the library in version 2.9. As such, this code will +> not run as written without modification. It is preserved for reference purposes only. + +The `DNNBeamformer` model composes the following components: + [`torchaudio.transforms.Spectrogram`](https://pytorch.org/audio/stable/generated/torchaudio.transforms.Spectrogram.html#spectrogram) that applies Short-time Fourier Transform (STFT) to the waveform. + ConvTasNet without encoder/decoder that predicts T-F masks for speech and noise, respectively. diff --git a/examples/hubert/README.md b/examples/hubert/README.md index e320649416..f4785e675e 100644 --- a/examples/hubert/README.md +++ b/examples/hubert/README.md @@ -2,6 +2,10 @@ This directory contains sample implementations of pre-training pipeline for [HuBERT: Self-Supervised Speech Representation Learning by Masked Prediction of Hidden Units](https://arxiv.org/abs/2106.07447). +> [!NOTE] +> The code uses the `torchaudio.info` function which was removed from the library in version 2.9. As such, this code will +> not run as written without modification. It is preserved for reference purposes only. + ## Pre-training Usage The Base architecture of HuBERT model requires two iterations of pre-training.